Fragility implies more to lose than to gain, equals more downside than upside, equals (unfavourable) asymmetry. - Nassim Taleb - Antifragility
When you add uncertainty to projects, they tend to cost more and take longer to complete. This applies to many, if not all projects. Projects take longer because the estimates are too optimistic. We have evidence of such bias, called overconfidence. - Nassim Taleb - Antifragility
An unbiased appreciation of uncertainty is a cornerstone of rationality—but it is not what people and organisations want. - Daniel Kahneman - Thinking Fast, and Slow
Harp burro paper on lean solutions
- Identifying Uncertainty
- Where does uncertainty come from?
- Conclusion
Introduction
The CarThe Dreaded Car PortIf you were to find the most sustainable solution to a structural engineering challenge, it will, by definition, be a solution that sits at the edge of a cliff. It will be dangerously close to a solution that will fail structurally. One slight error, inconsistency or deviation and we will tumble off the edge of the cliff into structural failure.
As engineers we are tempted to play it safe and move away from the edge. But as we move away from the edge of the cliff, the solution will be less sustainable; it will require more structure, it will be over-engineered. Every engineer’s goal should be to find this ideal solution, right at the edge of the cliff. But there is a driving force hiding deep inside engineering solutions which will ensure we never achieve the most sustainable and ideal solution: uncertainty.
We design buildings that don’t yet exist, using materials that we have limited knowledge of, for uses that are not clearly defined, using techniques that are over simplifications, done by people with limited skill and who make mistakes, in a commercial environment where every iteration costs money.
In this way it can be said that an engineer’s primary challenge is not to provide a solution for a building that ensures it stands up, but to reconcile this uncertainty.
Well, the way we do it, and the way we must do it, is to deliberately over-engineer. Over-engineering is not inherently bad, despite the negative connotation it has in our industry. But the question remains, what are the consequences of over-engineering?
The one significant consequence is that the structural solution has an impact on sustainability. And our primary goal as engineers should be to produce sustainable solutions.
We are going to explore the concept of uncertainty in structural engineering design and how it is a key contributor to sustainability in the design of a building.
We are going to explain why engineering is not black and white, why some uncertainties are well defined and prescriptive while others hide in the shadows. It is these uncertainties which lead to a variety of solutions to the same challenge by different engineers. And, perhaps more significantly, by the same engineers in different environments. We will explore how human nature, our knowledge, our personalities, and our experiences, influences certain types of uncertainty. This variety is the difference between a solution that is sustainable and one that is not.
We will start by identifying a key principle of structural engineering design: the ideal structural solution, one that has the highest levels of sustainability and how this solution lives at a tipping point. A small move to the left of this point will result in the solution ending in structural failure; a tiny move to the right lands the solution in the structurally ideal. In the words of Taleb, our solutions are asymmetrical. We call this tipping point the cliff. We will explore what defines this point, why we want to get as close to it as possible, what can encourage us to play close to the cliff, and how we can get closer to it.
We are going to show how we make decisions, how our decision making process is flawed, how we can break problems down into smaller parts, how to implement a system of how to identify the level of uncertainty to better understand it. We will then show you the benefits this brings to our solutions.
As an experienced engineer, playing in the space close to the cliff is where the fun is. Through this play, I will share two types of innovation, one small and meaningful and another that is a game changer that result from this play. We will explore where these innovations live, how to find them, how to tap into them, and the different ways to look at them. We will connect these innovations to uncertainty and ultimately link this back to sustainability.
What is engineering?
“I am worried my building is collapsing. Can you help me?”
One of my first projects as a structural engineer out in real world of consulting was to do a review of a suspended concrete slab for a school in Sydney. After they removed the formwork, the slab deflected excessively and collections of horizontal cracks appeared at the tops of all the columns. The builders were worried. We were called in to do a review.
Cracks in concrete are normal and would not necessarily cause alarm. But there was something about these cracks and the performance of the slab more generally which didn’t feel right.
Under the supervision of my manager, I did the analysis and came up with some results. The analysis was pretty standard stuff and there wasn’t anything out of the ordinary that was beyond any junior engineers capabilities.
At least, that is what I thought.
Somewhat surprisingly, I determined that the concrete slab was under-engineered. I wasn’t overly concerned, it seemed a simple engineering mistake and I concluded that we needed to undertake some remedial work. When I look back now, I shake my head and think about how naïve I was.
My manager, however, wasn’t convinced about my analysis and asked me to have another look at it. I determined later, that he actually wanted the slab to work, he had no interest in disgracing the other engineer. And no interest in undertaking expensive remedial work. I wish I could remember his reasoning, but it was a while ago now and I would only be guessing. As an experienced engineer now, I could give a list of reasons why saying they had made a mistake is a bad idea.
The reason my manager wasn’t happy with my conclusion was that the concrete “wants” to find a solution that works. Provided it is engineered to reasonable level (in other words, no gross errors in the design or construction), concrete slab have the potential to find a way to distribute the forces so that it doesn’t fail. The normal consequence of this redistribution is cracking, or what we generically call serviceability. And that was what we were observing in the concrete.
Specifically, my manager was not happy that I did a single analysis, showed that it didn’t work, and drew my conclusion. He wanted me to explore the idea of looking at the analysis a different way. He wanted me to try and try again, until I found a solution that worked. But isn’t there just one way to analyse the slab? Isn’t the solution black and white, it either works or doesn’t work? With his help, I eventually found a way to analyse the concrete slab so that it worked. It came down to the way in which the columns were modelled in my analysis. It turns out you could consider the columns to be 100% ‘active’, all the way down to 0% ‘active’. This choice changes the way a concrete slab distributes loads.
Once we found a solution that worked, we looked at the detailing of the reinforcement and determined what the consequences would be. Sure enough we determined that with this solution that I eventually managed to get to work, the stresses in the columns were high enough that cracking was likely to occur at the tops of the columns. We concluded that the cracking was a simple consequence of the design, had no detrimental structural impact, and the concrete slab was structurally adequate. No remedial work was required, and the structural engineers were cleared of any mistakes.
I learned a few things that day, besides the fact that concrete can be quite accommodating. I started to realise that there are multiple ways to look at structures, even ones that are not complex. I concluded that engineering is not black and white.
What is most fascinating about engineering is how it is riddled with uncertainty, some known and some unknown, but yet we are somehow able to reconcile it. I define successful engineers as those that are able to reconcile uncertainty. It is the difference between those who can “publish” and those that cannot, as my old boss, Roger Hooke once said.
In the above example, I used words like “analysed”, “modelled” and “under-engineered”. These are all words that structural engineers take for granted. What do these words mean? What am I referring to?
I wasn’t out on site testing the actual concrete slab and columns. I was in a small office, in front of a small computer running a piece of commercially available software. I was using the software to undertake a theoretical test of a virtual concrete slab. In this virtual space, everything is idealised; everything is,well, perfect. It is the perfect bit of concrete and steel built to millimeter perfection. This is quite different to what actual was built.
There is a clear disconnect between the slab that is out there, existing in the real world and the slab that exists virtually in the computer. And we hope that what lives virtually inside the computer, the model, is a reasonably good prediction of the real thing. And the answers that we get from the computer are based on an analysis that the computer undertakes on this idealised virtual model I have created. But that is just not the case at all.
Structural modelling
One of the tasks for structural engineers is to design structures that are going to be built. We do this by creating a structural model that attempts to represent the real structure once it is constructed. It is impossible to create a model that represents reality, and every structural model is a simplification, often a “gross simplification,” of what will exist in real life in the future.
The process of simplifying and idealising the real structure is referred to as modelling, although this term is also applied to the full process of creating the idealisation and then analysing it with an appropriate method of analysis. The idealisation is a gross simplication of the real building structure, and is referred to as a structural model. A further important step in structural modelling is to take the results of the analysis of the simplified model, and interpret them in relation to the behaviour of the real structure. This requires insight, and an appreciation and mature understanding of structural behaviour on the part of the structural engineer. It is an essential step in the analysis, and hence design, of a real structure. Design and Analysis of Concrete Structures
To the untrained this could induce concern. This could easily be interpreted to mean that no-one is able to interpret a structural model, even if you have time, advanced software, experience, skill and expertise. What is an experienced structural engineer basing their “insight” on? Structures don’t get tested after they are constructed to determine if a solution is correct. Or to test how accurate the solution is to one that wouldn’t work. The only conclusion that can really be drawn from a building that has been built, is that the solution worked for this one particular case that actually got built. The loads, the material and the particulars of the analysis still remain unknown. The authors are saying that a model can be interpreted at the discretion of the engineer. There is no insight gained from a solution that does not fail. We require our structural engineers to be mavericks. There is a better way. And we will show you methodology that defines a better way to address uncertainty that is beyond a single engineers personal insight.
Defining Uncertainty
What is uncertainty? Before we begin lets briefly introduce the concept of uncertainty as it relates to structural engineer solutions.
Uncertainty by definition
Although other words are often substituted with uncertainty we will try to use it within this paper in its true meaning. The difference between uncertainty and risk was defined by Knight in 1921 and has largely been uncontested. We will also highlight the subtle difference between unpredictability and uncertainty.
Uncertainty and risk
Risk and uncertainty are often used to represent the same characteristic of a problem, however, it is important to distinguish the two concepts. In both cases the outcome is unknown. If you role two six sided dice, the outcome is unknown and could be anywhere between two and twelve. This is true for both uncertainty and risk. The difference between the two lies in your ability to predict the answer. With two dice that are not biased, there is a probability that you will get seven more often than any other number (there are more combinations that add to seven). If the dice are biased then the probabilities change. If our two dice are both biased to roll a three, and this is not known to us, then rolling a six might be just as likely as rolling a seven. The probability distribution has changed. This situation is characterised by uncertainty. Knight 1921
The Theory Behind Decision-Making Under Uncertainty Versus RiskIn this paper we will maintain this distinction between risk and uncertainty. In general, our discussions will be about uncertainty, not risk, as it is the variations in the probability distribution which is our primary concern.
Uncertainty and unpredictability
Unpredictability is to not have the ability to say or estimate that a specific thing will occur at some point in the future and uncertainty is to not be able to say that a specific thing is currently absolutely true. For our purposes we will bundle both unpredictability and uncertainty into one and call them both uncertainty. We will do this because, for our purposes, we do not need to make any distinction in our discussions about when a specific thing will occur. We believe this is justified because structural engineering is about analysing a structure that does not yet exist.
Each individual structural engineering solution will have an amount of uncertainty. The amount of uncertainty will be influenced by a number of factors. We are going to explore these factors later.
Uncertainty, knowledge and the design process
In this paper they define ‘uncertainty’ as a lack of knowledge about the final solution. And the design process can be viewed as a set of inter-related activities which are performed to increase knowledge, and in so doing, reduce uncertainty about the design solution.
We will come back to this idea when we discuss managing uncertainty within the context of the maturity of a design.
The presence of uncertainty
At this point it is suffice to say that structural engineering is not black and white because there is always uncertainty. In other words, it is because there is uncertainty that engineering is not black and white. Lets take a simple example. Take a steel bar and rest it between two chairs that are spaced at 1m apart. Now apply a 1.0kg weight in the middle of the bar. Lets analyse this using some hightech piece of software, or do the calucation by hand. Using all the information available we can calculate the theoretical deflection. If this theoretical deflection in the middle of the bar is exactly, say, 1.0mm, the chances that a specific piece of steel with exactly these parameters will, in reality, deflect 1.0mm is pretty good. But that is as good a prediction that we can make: “pretty good.” We have the mathematical know-how and the standard theoretical mechanical properties of the steel to make this theoretical prediction. But the real world is never exactly like this. The deflection will vary. Put another way, the actual value will be different if we were to use two different pieces of steel; the actual value will be different if we use two different 1.0kg weights; the actual value will vary if the steel is part of a more complex structure or the span is not exactly 1000mm. There are number of different variables for which we can not absolutely know the answer. All these variables carry with them a certain amount of uncertainty. Collectively the carry even more uncertainty.
What causes this variation? How much is the actual value different from this theoretical value? How can we engineer structures if we don’t actually know what these values will be in a real life situation? These are questions we will address.
When a single solution is not a single solution
DeterministicThere is a space that exists between the theoretical solution and the actual solution. A liminality. As described by Arnold van Gennep in 1908, this space, this liminal, has advantages and disadvantages. It can trigger positive emotions and can trigger creative solutions. Conversely, liminality can trigger fear and discomfort. Everyone has different responses to these triggers. This is important. We will show how important later.
Let’s describe this space using some graphics.
Lets represent a single solution as a point. It will represent the theoretical singular value actual calculated answer. We represent this single solution as a black dot.
This black dot sits in space with no scale and no rating or ranking system. It is simply a dot.
We know that this theoretically calculated value will (almost) never really occur because it is a theoretical value for a structure that does not yet exist, with unknown materials and loads using a theoretically predictive analysis: it has some variability. What is more important is that we cannot rely on this answer. It is an approximation of the result.
So instead of a point, lets start by representing this value as a circle.
And lets represent the amount of uncertainty by lines either side of this circle.
These arrows can change in length with their length representing the amount of potential variation in the solution. The longer the arrows, the more uncertainty. The shorter the arrows, the more certainty we have about the solution. The longer the arrows, the larger the range in which the solution could land. Due to this uncertainty, a singular solution could land anywhere within a range. We will explore how this extent is determined.
We are going to adopt the term to define this range of the solution as a measure of uncertainty and use the term “the horizon of uncertainty.”
A normal distribution
An actual solution, one that really occurs in the structure, could land within the circle (the black dot represents the solution as experienced in reality - the actual measured value).
The solution could land away from the circle, but within the extent of the arrows, the horizon of uncertainty.
There is also a possibility, even if it is highly unlikely, it will land outside of the horizon of uncertainty.
The likelihood that the solution will not land where one would expect (right in the middle of the circle) reduces as you move aware from the theoretically calculated value. Considering our steel bar example, the likelihood the deflection will be 1.01mm will be higher than, say, 1.36mm.
We can plot these possibilities on a graph. In order to determine this graph, we would need to do this exact same experiment with a number, say 1000, pieces of steel.
The answer is that the actual solution, the measured deflection of the steel bar, will be distributed from the expected location by a normal distribution. If we plot the answers from our 1000 samples, it will form a bell curve. What this graph is telling us, is that more often than not, the value will likely land close to the middle, but the answer could be anywhere within the curve.
If we plot this curve against our circle and arrows, the extent of the graph will match the extent of the arrows. In other words, the arrows show the extent of the possible solutions that may actually occur.
The shape of the bell curve is important. Is there more of a chance of landing in the middle? Is there a 90% chance of the answer landing within a few points of the middle, or is it more like 20%. This makes a difference. We will discuss the different shapes this normal distribution can take for a given solution.
Types of uncertainty
Lets reduce structural engineering to its most basic form: structural engineering is about determining the applied loads that are applied to an individual structural element and comparing this to the ability of this element to support the loads. If the ability (we refer to this as the capacity) of the element is enough to support the loads, then the element will not fail. We are examining structures at a micro-structural level here. Lets go bigger, lets go macro-structural. Structures are an interlinking collection of these individual structural elements. Lets transmit all of the loads through all of the structure. If every individual element in a structure can carry the loads applied to it then the whole structure will not fail. And we do the analysis by using our knowledge and modelling an idealised structure.
On the surface of it, when described in this way, structural engineering is very straight forward. So where does uncertainty come from?
I have italicized four phrases: determining the applied loads, ability of this element, interlinking collection of these individual structural elements, and analysis by using our knowledge and modelling an idealised structure. Uncertainty lives in all four of these concepts. What are the applied loads? What is the ability of the element? Where and how are the individual structural elements interlinked to form the structure? How does an idealised structure match the real structure and how is our knowledge used.
We will bundle the first two concepts together and collectively call them one form of uncertainty: prescriptive uncertainty. The third concept we will think about separately and call it another form of uncertainty: embedded uncertainty. The fourth concept we will call: abstraction uncertainty
Lets talk about each of these separately.
Prescriptive Uncertainty - The Known Unknowns
Lets consider the first two concepts: determining the applied loads, and the ability of this element. Firstly, in reality we cannot precisely predict what the applied loads will be for the life of the building. Decades of research around the world have been spent determining what the loads inside an office space might be, the loads inside a residence, the wind and earthquake loads. Importantly, we have also explored how likely it is that these loads are going to be exceeded during the life of a building. Not to mention Black Swan events that might occur.
Secondly, in reality we cannot precisely specify the capacity of a material. We have a pretty good idea about these values, but they are not exactly what we predict. Again, there has been decades of empirical research undertaken to determine the variability within any given standardised material such as timber, steel, concrete, glass, etc.
Collectively, we have done enough research and data collection on these two concepts that we have been able to develop an entire philosophy around them. And it is now standardised such that every structural engineer around the world uses it to determine if a solution is structurally adequate. In essence, we have developed a highly prescriptive approach on how to deal with these two uncertainties. These uncertainties are known unknowns. They are stochastic and are generally defined by normal distributions.
The philosophy is called Limit States Design. Using LSD, the structural adequacy is a function of the risk associated with the unpredictability of the materials and the applied loads. LSD says that if a material has a known specific value for a material property, there is a known chance that it will not be this value. Similarly, LSD says that if the loads applied to a building have a statistically known specific value, there is a known chance that these loads may be exceeded during the lifetime of the building. When you compare the variability of the material (resistance) to the applied loads (actions), you want to make sure that there is only a small chance, or small risk, that the structure will fail. The way that LSD determines this chance is by comparing the normalised graphs of the resistance and the actions and ensuring that the overlap of these graphs is small.
In this way, LSD is highly prescriptive. The uncertainties, and unpredictability’s, are well established and standardised in highly prescriptive rules or codes. Each country or geographic region has codes that define the factors of variance. And although they are different from region to region, based on locally acquired statistical data, they all follow the same basic formula.
But not all uncertainty within structural engineering design is this highly prescriptive. While these prescriptive uncertainties are codified belts, the next category of uncertainties are personal braces. These are not stochastic as there are no statistical analysis undertaken. We call them embedded uncertainty.
Embedded Uncertainty - The Unknown Unknowns
One of my first internships at a structural engineering consulting company, I was given the task of engineering a simple steel structure. I engineered the structure using limit states design to about 90% of the capacity of the steel. I was using a pretty rudimentary engineering software package and then engineering each element by hand—determining the size of the element. I was inexperienced but was confident as an engineer and, to me, I was pretty happy with the design. When the design was reviewed by my supervisor he shocked me with a slight change. He suggested that every steel member could be reduced in size. At first, I questioned my calculations, thinking I was over-engineering and being too conservative. I re-engineered the entire structure with the smaller member sizes and it resulted in almost all of the steel being overstressed by about 10%. He agreed that we will run with these smaller sizes.
I thought he was being reckless. Why did he think it was acceptable to breach the fundamentals of engineering? What gave him such confidence? Why was he so indifferent to the rules?
I couldn’t help myself and asked why. His response gave me no additional comfort, “I am confident with the solution. I am more than happy to take all the structural steel elements to 110%. Don’t worry about it.”
This made no sense to me. He seemed to be making a mockery of my 7 years of education.
I remember talking to others about this and asking other people their opinion. Even several years after I was still talking about it. Some strongly agreed, others didn’t. There was a swathe of opinions (probably representing something like a normal distribution).
This gives us some insight into the third concept: interlinking collection of these individual structural elements. The challenge with structures—a collection of interlinking structural elements—is that how the structures link together, their connections, and the interactions between the elements, makes a difference. It makes a difference to how the element responds to the loads, but also makes a difference to how the loads are transferred through the structure.
Take our steel beam example, we specifically said that the answer is theoretically known using mathematics. This gives us great comfort. The mathematics sheds certainty on the challenge. It is deterministic. However there is actually no situation exactly like this, our example is idealized. In particular, what we call the boundary conditions, the supports or links to the adjacent elements, are never exactly this idealized situation. In other words, there is no engineering problem that can be comprehensively proven mathematically. This beam cannot exist in reality. This is true for all elements and all structures. So what we do is manipulate these idealized scenarios and transpose them into real life structures. It is this process of transposing that generates uncertainty. The more we move away from the ideal, the more our deterministic mathematical representation no longer represents our structure, the more complex the structure, the more doubt we have that the solution is correct. It is this doubt that causes uncertainty.
In addition to this, particularly with complex structures, there is always a chance that the engineer has not thought through every scenario. Have they considered the settlement in the ground, have they considered the loads during construction, have they considered the interaction between different materials. A structural engineers job is primarily to think of every situation. I often define engineeringas finding a way a structure can fail and stopping it from failing in that way. In a way this is the definition of the premortem which we will discuss later. But it is fools errand in perfection to think that we can think of every scenario, and almost every project will have some, often small, situation that they could not have foreseen. There are always uncertainties that are unknown unknowns.
Embedded uncertainty arises because engineering problems associated with entire interlinked structures do not have direct mathematical solutions. There is no prescriptive approach on dealing with these types of uncertainties. Importantly, the level to which an engineer associates uncertainty to the solution that they have created is dealt with on a case by case basis by the individual using their engineering judgement.
What we are saying is that three separate engineers, producing exactly the same solutions for a given problem, will, often unconsciously, associate different levels of uncertainty to their solution.
How does this uncertainty play out? Lets say three engineers model a complex structure. Using exactly the same technique they arrive at an answer. Lets say one of the solutions is that a concrete slab is determined to need to be 186mm thick in order to be structurally adequate. One engineer may specifiy 185mm, another 190 and another 200. All things being equal, the choice is primarily driven by the level of uncertainty they place on the solution. The engineer who chooses 185 is clearly confident with the solution, while the engineer who chooses 200 is not.
Adding to the complexity is that there is often more than one technique to determining a solution. Think of the concrete slab in the example in the introduction. This can add additional doubt into the minds of the engineer; have they chosen the correct methodology for determining the solution. Uncertainty directly correlates to doubt.
We will look at the sources of this uncertainty later.
Abstraction Uncertainty
The 500 deep concrete beam
If an engineer is designing a concrete beam and they are making a decision between using a 500 deep beam and a 600 deep beam, their decision on the solution has much less to do with the engineering and much more to do with uncertainty.
Daniel Kahneman introduced us to the concept of noise. Some people are more comfortable with risk than others, which might lean them towards the 500 deep beam. Kahneman calls this level noise. It is just a personal comfort with uncertainty.
If I was promoted yesterday and in a good mood, I am might be more likely to lean towards the 500 deep beam. Kahneman calls this occasion noise. It is decisions that change with the personal occasion that you are current in at the exact moment that you are making the decision.
There is another actor at play here. Kahneman says that uncertainly and doubt lives in System 2 thinking. System 2 requires effort and hard work. Effort and work that we would prefer to avoid. We are inherently lazy. So if there is some level of doubt that the 500 deep beam is adequate, and if we have undertaken an analysis and found that a 600 deep beam is adequate, what would encourage us to explore the option of the 500 deep beam? Being inherently lazy we would naturally drift towards 600.
So how do we encourage engineers to break through that laziness and work harder to choose the 500 deep beam? Is there a way to set up an environment where an engineer will choose a better option? There is. And we will explore it later.
This is it, or is it?
Unlike scientists that conduct experiments on real objects. Engineering is about analysing things that don’t yet exist. The building is an abstraction, it is an idea that we have imagined in our heads and explored through idealised perfection. There are two principles within this abstraction: how closely will the real structure match the structure that we have imagined, and how does my engineering knowledge match the requirements to successfully imagine the virtual structure that I have modelled. This is abstraction uncertainty—it is uncertainty that relates to our knowledge and how we use this knowledge. It is uncertainty linked to who I am as a person. It is uncertainty linked to how I am feeling at the time, and the environment in which I work. What is unique about abstraction uncertainty is that it can manifests itself in unusual ways.
How closely does the image I have formulated in my head and the analysis that I have undertaken based on that image, predict the actual behaviour if we were to empirically check the real built building? There is skill involved in formulating the image of the structure of a building, and there is additional skill in translating this image into something that can be modelled and analysed.
Adding to the uncertainty, is that the solution is not verifiable. It is rare for a building to be checked after construction to see if the analysis was a good predictor of the actual building. The only conclusion that can be made when a building is finally built, and the structure is still standing, is that the solution must have had some indeterminate level of over-engineering, otherwise it would have failed. But to what extent? And could the solution have been leaner and more sustainable?
Structural engineers need to analyse the actual building using gross simplification techniques. How well does this analysis predict the real behaviour of the structure. As we will discuss further, it is impossible to analyse an exact structure. As good as we are at predicting the behaviour of a real steel beam, for example, the connections at each end of the beams cannot be predicted with absolute certainty. Assumptions, rules of thumb, personal experience and randomness all influence the level of simplification of the analysis. It is also worth noting that even if we could predict structures to their infinite exactness, there is an element of commerciality to engineering. There is a balance between the time spent to develop a solution that is good-enough, and the additional time and effort to get closer to the exact solution; one that is “at the edge”.
What are the chances that my knowledge is not sufficient to undertake the engineering task? It is impossible for any single engineer to know everything there is to know about structural engineering. This results in a number of different approaches by engineers to tackle the demands of the job. Many engineers specialise in one particular type of engineering, such as concrete, and then only hold a basic knowledge of other types of engineering, like timber or steel. Others attempt to have a broad knowledge of a multitude of different type of structural engineering. But almost without exception, every engineer would have a type of engineering of which they would have limited to no knowledge. This could be as simple as a type of material like glass, aluminium, brass or timber. For others, it could be load types, such as earthquake, vibrations or cyclonic winds. For others, it could be limited in their conceptual understanding, their ability to collaborate or their ability to model certain structures.
With this limitations of our knowledge we can also see that uncertainty can come from human error. Most human errors are accidental. Some are due to lack of knowledge, others from lack of experience. Errors can arise due to a particular aspect of the design that the engineer did not know was important for the building. When the Millennium Bridge opened across the Thames in London, it was quickly shut down due to excessive vibration. The answers to these questions are not relevant, but the questions worth exploring are, did the engineers turn their mind to the concept of vibration, did the engineers use the wrong knowledge to undertake the vibration calculations, was the correct knowledge applied incorrectly, or were there human error in the calculations?
Errors can also arise from the use of customs that could be wrong. For your entire career you may believe that something is correct, you have made it a custom. Only to find out that the belief is wrong. These customs can even be embedded in codes. Shear in concrete beams has only recently been developed to a point where there is a true understanding of its behaviour. Before now the codes were based on empirical formulas that were known to be poor predictors of shear. Vibration in floors is still not readily understood and, again, customs that we know are incorrect remain in place to predict vibrations.
And, of course, routine errors are common during calculations. These are not just errors of rounding, but errors of incorrect transfer of data, or even errors in conceptual understanding of structures. Codes are often filled with complicated formulas and concepts, and small errors of transferring of formulas or numbers, or a misunderstanding of a concept is possible.
A builders response
“I know engineers need to over engineer everything. I get it.”
I paused. Something didn’t sit right with me. I had just spent a few minutes describing to a builder why we needed to embed the timber joists into a brick wall, rather than use a wall plate. Its complicated and I didn’t do a good job at describing why we cannot use wall plates. And I could see him loose interested during the conversation.
“I have been using wall plates for 30 years,” he continued.
I realised something on that day. The builder didn’t actually think I over engineered everything. After all, there is nothing inherently wrong with over engineering, and it is likely to benefit builders too. I realised what he was actually saying. He was actually saying, “I don’t understand.” This is my understanding of word ‘ostensibly.’ And it is common when we discuss complex concepts to people outside of our domain of expertise.
I have since become heavily focused on coming up with simple explanations, or methodologies for explanations, that a normal person could understand. I listen to clients and architects language that they use to describe structures. Like the use of the word ‘stick’ to mean a column. Or ‘house of cards’ to describe instability due to lateral stability. And when I am chatting with a client or architect, I pick up on their language and then adopt it and use it within my own language. I also do this with people that I mentor. I listen to the specific words they use. It was during a presentation that I did with some school age children that, despite my own children being familiar with them, I learned that “cantilever”, “loads”, “deflections” and “displacements” for example (I was clever enough to avoid “bending moments” and “shear”) were new words for a 6 year old.
Identifying Uncertainty
I want to now expand our diagrams and take one step further and explore uncertainty in structural engineering by placing our solution on a graph. The graph will correlate Structural Adequacy(1) to Sustainability(2).
This simple graph will represent several complex aspects of structural engineering design, including the uncertainty of a solution, structural adequacy, an ideal solution and how these relate to sustainability.
The graph will at first show a level of optimisation of the structure with the idea that optimisation of the structure is the ideal use of a material such that is the least amount of material that could be used and still remain functional and not fail. It is therefore, by definition, the most sustainable solution. But sustainability is more than just the optimisation of a material. Sustainability is also about the choice of material, or substitution, and its impact on the environmental, it is about reduction and even potential elimination of materials.
Innovation plays a key role in sustainability. With each aspect of sustainability (optimisation, substitution, reduction and elimination), innovation can play a role. We will explore the idea of innovation in more depth.
The Graph
So lets describe how we are going to develop this graph by plotting structural adequacy against sustainability.
On the Y-Axis we have the level of structural adequacy, the amount to which a specific design is structurally adequate. Above the zero axis the design is structurally adequate, with more structurally adequate designs higher up the axis. More structural adequacy means that the structure has some level of redundancy such as it could carry more loads than the design requires. Below the zero line, the design is not structurally adequate. We might call these designs below the zero line those designs that do not meet the structural requirements of the project, or more strictly, failure of the structural requirements. As with structurally adequate, structurally inadequate has levels, with more structural inadequacy meaning the structure is less able to meet the structural requirements(3).
On the X-Axis we have a level of sustainability of the structure of the building. For simplification, this may be a measure of the embedded carbon of the structure, although it could be considered using more complex criteria if desired. In the positive area of the graph, with the lowest numbers, we have the highest sustainable design rating and as we increase the value along the X-axis the solution becomes less sustainable.
As mentioned, the point at which the highest level of sustainability can be achieved is at the zero point on the X-Axis. Below zero point is a special case. It is an extreme outlier, it is “Not Sustainable”. Structural failure, which occurs in the entire negative range, is, by definition, not sustainable. We hope that this label which lands in this negative range will become self evident in a moment.
On the Y-Axis, we are going to plot an additional line. This line will represent the Limit States Design (LSD) Adequacy Line. We will explore the meaning of LSD later, but for now it is suffice to say that it represents the factors of safety embedded into the design to cater for known uncertainty. Below these safety levels the design is said to fail the LSD criteria.
Solutions that are above the LSD Adequacy line are deemed to be “acceptable.” They have appropriate amounts of safety to account for the potential events that are unpredictable during the lifespan of the building.
On our graph, solutions that are structurally adequate , above the zero point on the Y-Axis but are below the LSD Adequacy line are deemed to be unacceptable in accordance with the principles of LSD. In other words, the factors of safety embedded in these solutions are not high enough to account for the potential unpredictable events that may occur during the lifespan of the building.
These are the criteria that sets up the graph which we will use to determine whether a solution is structurally acceptable and the level of sustainability for the solution.
What is the “best” solution?
For any single engineering decision, there is an infinite number of solutions that could potentially land within the acceptable design zone. What we want to explore is where exactly a solution should land in order to form the “best” solution. The criteria we are going to use to define the best solution is the one that is structurally adequate and with the highest level of sustainability. Although we will present an alternative case for the best solution at the end of this paper, for our discussion we will use sustainability as the only criteria of importance to judge one solution as being superior to an alternative solution.
A Look at Specifics
When we examine a specific singular engineering solution, the solution could land anywhere on this graph. Depending on where it lands we can judge the solution as being acceptable or not acceptable. Further to this, we will also be able to judge whether one solution is better than another solution, depending on where the two solutions land on the graph.
A solution needs to sit above the LSD line to deemed to be acceptable. An example of a specific solution which is acceptable would lay in the upper right quadrant such as shown below.
The above solution is in the upper right quadrant, a zone we have defined as being acceptable; it is above the LSD adequate line on the X-Axis of adequacy and to the right of the zero line on the Y-Axis of sustainability. Note that it is some distance to the right of the zero line on the Y-Axis. This suggests that there may be different solutions that would land closer to the zero Y-Axis which, using our criteria, is more sustainable.
An example of a specific design which is not acceptable would lay in the lower left quadrant.
The solution that we have plotted above is a solution that fails. It is structurally not adequate. The structure would either collapse, or not comply with basic serviceability criteria of the design.
The black and white (or not) of structural engineering
We have specifically shown the singular design solution on the graph as an enlarged O, and specifically shown the design solution to be within a circle. We are using these to indicate that even within a specific singular solution, there is always small amounts of variability; engineering is never black and white. This premise of structural engineering not being black and white may surprise some.
“But it is just science, everyone will come up with the same answer.” This a common quote from a client. This is simply not true, and how not true this statement is, is the essense of our discussion. And even more than this it is not true on a multitude of levels.
We have discussed the philosophy of LSD, and how it deals with the unpredictability of the materials and the unpredictability of the applied loads. But uncertainty lives in other places too. The uncertainty of LSD is well documented and, although factors vary throughout the world due to local experience, highly prescriptive. But there are other areas of uncertainty which hide in the shadows. And many even hide in the shadows of our mind. Skill, biases, prejudices, mood, environment, risk-aversion and many other factors influence a designers decisions and cloud our decision making process. These factors have a direct impact on the sustainability of a building.
Design Trajectories
Now lets track the possible trajectory of a specific design solution. This will represent one of hundreds, possibly thousands of individual design solutions that exist within a single project. It might be best to use a specific example to describe the design trajectory.
Lets say that we have a specific concrete floor plate that is supported and bounded by four columns on an 6m by 6m grid. We are going to choose the slab thickness as the variable that we will modify to try to achieve adequacy. There are other variables that we could choose, but, as we will explain later, thickness may be considered the most appropriate. The potential choice for the thickness of the slab ranges from zero to infinite. Clearly these extremes are non-sensical but we will use them here to define the boundaries of the possible. If we were to plot this range of thicknesses on the graph, the resultant line would follow something like this:
Continuing with our concrete slab example to plot the possible options for the design. As we move from the lower boundary of a slab thickness of zero, to a certain thickness of say 200mm thick, the concrete slab would not be structurally adequate. It would collapse and fail perhaps, or less extreme would be that it does not perform its function; perhaps it is too bouncy for its design use. In relation to Limit States Design it either fails the strength of fails the serviceability design state. The solution is in the lower left quadrant of the graph.
As we increase the structural adequacy of the design solution, in our example by increasing the slab thickness above 200mm, it is structurally adequate, however, due to the risk criteria outlined in Limit States Design it is said not to be “acceptable.” This essentially means that, if we consider the potential life of the building, and the potential variations that may occur during this lifespan, there may be a moment, perhaps even a fleeting second, during an earthquake perhaps, when the building is susceptible to failure.
If we increase the concrete slab thickness even more, it reaches a point where it crosses the LSD Adequate line and lands in the upper right quadrant: the “acceptable” zone. A solution above the LSD Adequate line means that the solution is structurally adequate and is able to support the loads when the appropriate risks of unpredictability of materials and loads are considered. If we are to go further above the level of LSD adequate line, the solution will increase in structural adequacy.
It may be worth mentioning at this point that the more structural adequacy we achieve, which may at first appear appealing—why would we not want a structure to be more safe?—the less sustainable the solution and the solution will start to move away from our optimal design solution. There are other consequences including cost, client requirements and architectural merit which we are going to ignore here but will undoubtedly form part of the success matrix for the project.
In our concrete slab example, we could continually increase the thickness of the concrete slab. At the extreme boundary condition we could potentially increase the slab thickness such that it is infinitely thick. So how do we choose the solution? How thin can we make the concrete slab?
The answer lies within the realms of design uncertainty.
The cliff or the ideal?
There is a location on the graph where the solution will land at the exact point of the ideal. A point where there is a perfect balance. Where a solution “will just work” from an adequacy criteria. This point also has the highest level of sustainability because it uses the least amount of structural material. It sits on the precipice. It is the ideal.
But it is a dangerous place to be: a tiny move to the left and you fall off the cliff into the land of the inadequate.
So how do we find the perfect balance?
The asymmetry
The challenge here is the asymmetry. The consequence of being on the wrong side of the line can be devastating—literally. We don’t really want to play close to the edge unless we are certain that we will not fall off. This is the driving factor for engineers being risk-averse. Stay away from the edge, play where it is safe. If you fall off that cliff, there is no coming back.
The more we move away from the cliff edge; in our example, the more we increase the slab thickness, the more we move away from the ideal. We call this over-engineering. Over-engineered is where the structure has higher levels of safety than is require for the ideal solution.
Over-engineering has negative connotations, and for good reason. Why would we add thickness to the slab, adding costs, adding materials, adding height and most importantly reducing the sustainability of the design, when it is not required to produce a solution that is adequate?
We are going to talk about over-engineering, but before we do that lets talk about optionality.
Optionality
If you remember, in our example of the concrete slab we chose the thickness to be the primary variable. There are other variables which we could consider. One such variable is the concrete strength. For example, this might revolve around different material choices or balancing of the different materials within the design. This, in itself, could create a more sustainable solution through an innovative solution using less material perhaps, or material that has better efficiency of its embodied carbon generally, a term we call carbon efficiency.
If we plot these options on the graph, the zone could look something like this:
What might become apparent is the idea that one of these paths are better than the other. There exists an optimal and a sub-optimal solution choice. Where the optimal solution provides a higher level of adequacy with the same level of sustainability. Or conversely, where the sub-optimal solution crosses the LSD Adequate line at a level of sustainability lower than the optimal solution.
So now we see the graph showing a range of potential solutions for a single engineering challenge. With the shaded area showing the zone of the solution that are structurally acceptable. This brings us to the concept of over-engineering.
Why do we over-engineer?
Over-engineering typically results in a less sustainable solution. The more you over-engineer, the less sustainable the solution. So why would we over-engineer? This is the primary discussion we are having here. Why would over-engineering ever occur?
Over-engineering—a solution that is not the most sustainable—occurs as a consequence of the inherent relationship between uncertainty and asymmetry.
Over-engineering—a solution that is not the most sustainable—occurs as a consequence of the inherent relationship between uncertainty and asymmetry.
Lets start by introducing the idea of engineering judgement.
We argue that there is, in fact, a “zone” to which the design is adequate. This zone represents multiple options available. It is at the discretion of the individual as to where the chosen solution will land. We will refer to this discretion as engineering judgement.
So using our engineering judgement, lets revisit our previous question, how do we choose a solution which is the best?
The key to the answer lies in the uncertainty embedded in the solution. This uncertainty drives a variability in determining if the solution will work. As previously discussed, if we plot the variability as a standard deviation (SD) we can see where the solution must be in order to be deemed to be structurally acceptable.
Lets plot a single potential solution on the graph. The graph will have its optimal and suboptimal options (shown as lines) and the standard deviation resulting from the uncertainty engaged in the solution (using the arrow convention mentioned previously) the solution will sit within the design zone potentially something like this:
- Lets review the methodology for the graphing of the solution. The exact solution is represented as a circle. It could be replaced by the O we used earlier that sits in the middle of the circle representing a singular solution. The arrows that radiate from this circle is the standard deviation (SD) resulting from the uncertainty about the solution. The longer the arrows, the more the uncertainty. Plotting all the dimensions of the solution creates a zone within which the solution could land. The solution could land within the zone in the Y-Axis as a result of the variations to the solution (we will come to this later); the solution could land within the zone of the X-Axis as a result of the uncertainty. The more the uncertainty, the larger the extent in the X direction.
As we can see, the left edge of this zone is some distance away of the cliff. In other words, our design is adequate, and would appear and have an appropriate amount of safety even considering if the unknowns land the singular solution at the lectern edge of the zone. But the solution is over-engineered; we could “reduce” the solution and still ensure that the design is acceptable and, importantly, achieve a higher level of sustainability.
Lets look at a different situation. The same solution could have a greater amount of uncertainty. This would increase the extent of the variance or standard deviation of the solution in the X direction.
We can see now, that with the same solution, but with a larger amount of uncertainty, the potential solution suddenly comes perilously close to the edge of the cliff.
Reducing uncertainty is beneficial for achieving a more sustainable solution. So, how can we reduce uncertainty?
Reduce the uncertainty
Reducing uncertainty and minimizing variation is beneficial. Doing so allows us to move the solution as far left as possible.
To reduce uncertainty or standard deviation, we need a better understanding of the variables. This will result in a more sustainable design, allowing us to reduce the thickness of the concrete slab, for example. By plotting this on a chart, we can see that we have not only reduced the X direction extent of the design, but also moved it to the left by increasing certainty.
The key here is that we cannot move the design any more to the left than the left extent of the design solution plus the standard deviation. If we move the solution more to the left, then the solution has the potential to fall off the cliff (land to the left of the zero point on the X-axis). The only way that we can move it more to the left without falling off the cliff is to reduce the standard deviation by reducing the uncertainty.
So the key question becomes, how can we reduce the uncertainty. Before we look at this, lets take a look at the ideal.
The ideal
Lets take a quick recap of what we have discovered so far in determining the ideal solution. From our discussions so far, the ideal solution consists of three criteria:
- A solution that has the least amount of uncertainty as possible,
- A solution that is as far to the left as possible, but
- A solution that does not have the potential to fall off the cliff.
Where Does Uncertainty Come From?
During the design stage of a building, I will build the building, demolish the building, and then build the building dozens of times—in my head. During feasilibity, during preliminary, and then during detailed design, I will live inside the building again and again. I do this so many times that it is often surprising to actually see it getting built.
A structural engineers job is to design a structure that is yet to be built with loads that are yet to be determined using materials that are unknown and for an unknown time frame. Uncertainty exists in all aspects of this process because the building does not yet exist. Unlike scientists, who undertake experiments on things that exist, engineers are analysing something that does not yet exist. This is riddled with uncertainty. Uncertainty about the loads, uncertainty about the materials, uncertainty about the durability, and uncertainty about the accuracy of the analysis all being completed by humans with all their fallible traits.
There are numerous sources of uncertainty within the design, and we cannot and should not assume that every design is able to provide the ideal solution. In fact, it is impossible to determine what the ideal design is, let alone attempt to achieve it.
If we strive for perfection we are not only kidding ourselves and opening ourselves up to make mistakes, but unlikely to look for innovations and unlikely to look for solutions that are close to the cliff.
We are going to explore the sources of uncertainty.
We believe there are good correlations between noise that occurs while making judgments that Kahneman identifies in his book Noise. We will highlight these correlations.
Uncertainty and noise in judgments
Skill and experience are a significant source of uncertainty. Particularly skill and experience directly relating to the solution. Generally it would be safe to assume that the more experience a person has in a particular area, the more certainty they will associate with their solutions. This would be particularly relevant if they had several of their solutions successfully built, and have not collapsed in real life. And also because, the more you do a particular task the more confident you become with the process, including the common errors or subtleties of the work.
But this can also be a hindrance too. If there has been a bad experience, it is often that the engineer will associate more uncertainty to the solution. Kahneman talks about how expertise affects judgement, and generally makes the link between higher levels of expertise generally provides better judgments. He takes expertise one step further and talks about respect-experts. These are experts undertaking singular problems that are not verifiable who you must trust are giving you appropriate advice.
Skill makes a difference too. Different levels of skill a person has will result in a different assessment of the risk associated with a specific uncertainty. And it is not necessarily that more skill equals less risk, often a designer will assess a specific uncertainty as containing more risk when they have experienced the specific situation previously and had a bad experience. Once bitten, twice shy, as the saying goes.
Another factor is a person’s general tolerance to uncertainty and their personal aversion to risk. If you are more risk averse, your uncertainty would be greater. Kahneman refers to this as level noise.
Some engineers are more confident with one particular material over another, or feel that some materials are more predicable than others. This is often independent of their experience with this material. This translates into uncertainty. Kahneman refers to this as stable pattern noise.
Another factor is the engineers general disposition at the moment when they make the final decision. The disposition could be good, due to a personal events such as a resent promotion perhaps. Disposition also relates to the environment in which they work, and the environment of the team they are collaborating with. Kahneman refers to this as occasion noise which is a subset of pattern noise.
Specific Scenarios
Lets explore a few situations where you can justify to yourself, as the engineer, for moving away from the cliff edge because the situations adds to the uncertainty.
- Scenario: Your experience shows that on site, when these types of things are actually built, the solution cannot be built exactly to the specification—some allowance should be made due to deviation in the preciseness, or tolerance for perfection. This is possibly a result of your limited ability to see the full extent of the challenges of actually building what is designed. It could also be true that some builders are very capable of doing particular tasks while others are not as skilled, or would potentially charge more to get a skilled tradesperson in to complete the task. This is very much the case with a high finish in off-form concrete, for example.
- Scenario: The builder assumes that you have over-engineered the solution (they know that you, and most other engineers, live your design life far to the right of the cliff edge, as they have seen this on every project that they have worked on before) and they may respectfully request that the design be reduced. Sometimes this comes with the passive-aggressive tagline: “so that we, as a team, can save the client money.”
- Scenario: You, as the engineer, may want space to negotiate in case other aspects of the design are on the wrong side of the cliff. For example, you may think that some part of the design may be under-designed but want to experiment on site with the solution. You may need bargaining power to take from one part of the design and give to another without a variation to the project.
- Scenario: You, as the engineer, are concerned that the loads outlined in the codes may not be appropriate for the structure. There is some uncertainty in the use of the building, or in the machinery that is going to be used for mechanical plant for example. Or that the building may change in use over the lifetime of the building.
- Scenario: You, as the engineer, have chosen a solution that you think is appropriate based on the merits of the project. You also think that the solution can be built by a reasonably able builder. However there is a chance the builder is not experienced in this solution, or the builder may choose a different solution, or that your solution isn’t quite right for some reason, there is some doubt in your mind.
- Scenario: It is usually challenging to get close to the cliff edge. It is often easier to simply settle on a design that works, without critically analysing where it sits on the graph. Spending the time to iterate through design options to get the design closer to the edge often doesn’t make commercial or business sense. Most people are satisfied finding the needle in the haystack, it is a rare person who continues searching for another needle. It is worth noting that we do not believe that this strictly fits the criteria of an uncertainty, as it may be an unintentional shifting to the right, rather than reducing the variance embedded within a design decision. However, we feel it is a driving factor in determining where the design fits on the graph.
As a result of these uncertainties which are often cumulative, the typical solution has a large variance and therefore lives well away from the cliff edge, well away from the optimal solution and well away from a solution that is highly sustainable:
If we want the solution to be more sustainable, and feel that reducing uncertainty is the place to explore a more sustainable design, then we need to explore ways to reduce the uncertainty. Reducing uncertainty will reduce the variance, reducing the extent of the zone to which the solution potentially lands, allowing us to shift the solution to the left.
Managing Uncertainty
As Martin Fröderberg discusses in The Human Factor in Structural Engineering, most of the time, no such thing as a correct answer or solution exists. If a correct answer does not exist, then every solution that an engineer develops has some level of uncertainty. So why is it the case that engineers do not regularly discuss uncertainty? The answer is two-fold: firstly, we would need to overcome our inherent laziness; and secondly, as engineers it is imposed upon us that we consistently display overconfidence.
In Thinking, Fast and Slow, Daniel Kahneman describes System 1 and System 2 thinking. System 2, your working mind, is normally in a comfortable low-effort mode. System 1, our automatic machine, usually runs the show. It is only when System 1 runs into difficulty that System 2 is called into action. System 2 is hard work and we normally spend our time trying to avoid tapping into it. But we will need to fight an engrained characteristics built deep in our nature: laziness. We would much rather find a way to work that simply uses intuition or what is automatic, our System 1 machine.
One of System 2 main characteristics is laziness, a reluctance to invest more effort than is strictly necessary. Kahneman goes on to say,
A general “law of least effort” applies to cognitive...exertion. The law assets that if there are several ways to achieving the same goal, people will eventually gravitate to the least demanding course of action. In economy of action, effort is a cost, and the acquisition of skill is driven by the balance of benefits and costs.
Kahneman refers to the work of Keith Stanovich who call the people that avoid intellectual sloth, those who are “engaged” as rational.
Managing uncertainty requires us to actively engage System 2. System 2 is the only one that can follow rules, compare objects on several attributes, and make deliberate choices between options. These are some of the activities that we need to undertake to manage uncertainty. Or to quote Kahneman: “uncertainty and doubt are the domain of System 2.”
The prominence of causal intuitions is a recurrent theme in this book because people are prone to apply causal thinking inappropriately, to situations that require statistical reasoning. Statistical thinking derives conclusions about individual cases from properties of categories and ensembles. Unfortunately, System 1 does not have the capability for this mode of reasoning; System 2 can learn to think statistically, but few people receive the necessary training.
Therefore engaging with uncertainty requires us to overcome this inherent laziness and tap into our system 2 thinking. This is where uncertainty lives.
Expressing doubt or uncertainty about a solution is not what people want. Again Kahneman addresses this social psychological behaviour,
An unbiased appreciation of uncertainty is a cornerstone of rationality—but it is not what people and organisations want. - Daniel Kahneman - Thinking, Fast and Slow
Engineers are engaged to provide solutions. Solutions that work and solutions that, in the minds of the public, have zero likelihood of failing. If engineers were to say that they actually don’t know whether their solution works, this would cause concern, possibly even chaos. What the public wants is an engineer that says with confidence and certainty that their solution will never fail—despite the fact that is impossible to achieve.
What is important is that we are intentional about uncertainty. If uncertainty is inevitable then we need to engage with it. As an engineer, we need to understand how we deal with uncertainty, how it makes us feel, how comfortable we are with it and, then, how we can use it to your advantage. It is through this engagement with uncertainty that we are able to take advantage of it.
We are now going to explore how we can achieve this.
The Green Pear Room
The first time I went into the chicken shed, the smell of ammonia almost knocked me out. My eyes watered. I had images of collapsing to the floor. Running for the exits and breathing in the fresh air of the outdoors. I wondered how anyone could ever work in there. I was surprised that the smell disappeared the second time I visited the sheds. I was 14 and doing some general labour type work for my neighbour. He owned 3 large, two storey sheds full of chickens and my initial duties was with these chickens. I would feed them and cleaning the sheds. My second surprise on this job was the day I showed up and all the full grow chickens had been replaced with chicks. It turned out that when I went to my favour chicken shop it was likely I was eating one of the chickens I had cared for.
It was well paid, particularly for me, coming from a family that was not well off. I didn’t realise it at the time but he put a lot of trust in me, perhaps because I was reliable. But separate to the chickens, part of my responsibility was looking after small orchard of pears. Interestingly it was adjacent to my house with no fence or any barrier, so it was almost an extension to my backyard, separated sporadically by a few small trees. The orchard was about 40-50 trees perhaps. I would go there every day after school and care for them. They didn’t need my daily attention but I felt responsible to go anyway. Not sure why. I trimmed the branches and use the bosses quad bike to take the branches back to his house. I thinned the pears and ended up picking them. I had been taught how to care for pear trees during the previous summers working for one of the main orchard farmers in the neighbourhood, so I knew what to do.
Every Saturday I would take him my timesheet. I didn’t initially know why, but he would bring me a cheque on Monday rather than pay me on the Saturday.
What did you have in your mind when I told you they were pears? When you think of pears you probably thought of green pears. Well, these ones were brown. It is the aristocrat of pears which are common in Southern Ontario were I lived. We would eat them as a dessert, poached with a dash of cinnamon. Eaten like this they were a delicacy. They are denser, crisper and smoother than the traditional green pear.
There are hundreds of ways to eat a pear. You can grab it and eat it whole. You can cut, peel it, slice it, dice it, chop it. You can blend it or crush it and then drink it. You can core it, you can make a pie. You can even cut it in half and do one thing with one half and another thing with the other half. Or you can poach them and eat them with cinnamon and vanilla ice cream like I did.
Importantly, you make a decision and then you follow through and do it. It is done with intent. It is calculated and intentional. You know what you are looking for in the eating experience and you make a decision, follow a procedure and then follow that through.
When you move into the mode of experimenting with uncertainty you need to do it intentionally. You don’t want to half go there and half not. Just like the pear. Make the decision, follow the procedure and go through with it. Perhaps just as important is that when you have finished eating it make an assessment about whether you liked it. Would you do it again that way, or try something different?
We have found that if you do not go there with intent, if you waffle between the real and the experimental, you will not achieve success. You will dabble in the intuitive, rather than the statistical. You will perpetuate enhance the unknown rather than control and define it. And if you don’t assess the experience at the end, then you are not learning anything. You want to go with intent and record your experience there. You want to write a journal of our experience, because the process and the journey matter.
This is “The Green Pear Room.” Use the same technique with uncertainty. Be intentional. Otherwise it will control you, rather than you controlling it.
In Better by Gawande he writes about the US Military Academy’s attempt to reduce the death rate in war. His research revealed an intriguing methodology they used that we should all use. He concludes, we need “to make a science of performance, to investigate and improve how well they use the knowledge and technologies they already have at hand.” Independently, he also talks in the book about diligence which I also think is relevant for our discussions here. “Diligence stands as one of the most difficult challenges facing any group of people who take on tasks of risk and consequence.” Are we, as a profession, diligent about our work? Do we review our designs to see how sustainable they are, could we do better, or are we just happy that we don’t receive that dreaded phone call?
Maturity of Design
The design typically follows a specific process: starting with feasibility, then moving to preliminary, and then to detailed design. We will refer to these as “stages” and they follow 5 clearly defined stages. Although the stages are clearly defined, the transition point between stages is often ambiguous.
Preliminary design is just that: preliminary. They are based on information that is fuzzy and still to be determined. High levels of uncertainty exist at the preliminary stages of a design because, by definition, the information is immature. This is actually how we want it to be. We want it to be flexible, still in flux and undetermined. This results in a wide spread of the distribution (the arrows in our graph). As the design develops, as more information is obtained and the design becomes less fuzzy, the spread of the distribution narrows.
What is interesting here is the point at which you stop acquiring information that may contribute to the narrowing of the distribution of uncertainty. At some point, gaining more information will not assist with the level of uncertainty, or will assist in such a small way that it is not worth considering. It is the engineer’s job to determine the information that is impactful to uncertainty, and to satisfy themselves that they have enough detailed information to determine when to cease gaining additional information.
Simplify the Design
The more complex the design, the more uncertainty exists. As we become less certain and impose more doubt about a solution, we naturally move away from a sustainable solution. Also, as we developed more advanced research and more complex calculations, we become less confident and the chances of making errors is more likely. Complexity in calculations and codes inevitably imposes deviations from what may be a natural way to think about a structural solution - reducing it to a complex formula, where the final answer has little context to what may seem correct. A classic structural engineering example of this is the unrestrained length of a steel beam. The inflection points may be seen as defining the unrestrained length, but this simple heuristic is no longer the method for calculating this length. We have moved away from an obvious and sensible methodology, and reduced it to a complex formula. As elegantly put in Uncertainty in Structural Engineering , William M. Bulleit says about making codes more complex, “it is possible that we are trading one source of uncertainty for another. There needs to be a balance between perceived code complexity and perceived code accuracy.”
Changing a design to simplify the solution should be a goal. Not only will this reduce uncertainty, and generate a more sustainable solution, but will likely produce a more elegant solution. The maturity of the design should also include a simplification process.
Risk assessment
We cannot talk about structural design or structures without talking about risk. Within prescriptive uncertainty, based on statistical data, the distribution of uncertainty is known. We know, for example, that an office building, with an applied design action load of 3.0kPa, the likelihood of this load being exceeded. This likelihood is based on research. And note the word “design”. The word design is used to indicate that it is the value we should use in our calculations based on an expected value.
Limit State Design, at its core, is based on this risk analysis. LSD does not say that structures will never fail. Its guiding philosophy is to limit failure to a small possibility based on a rigorous risk assessment. If the loads of a structure exceed what is assumed in the design—the design loads, and the materials are inferior to the properties assumed (they must both happen at the same time), then the structure has a risk failure. What LSD does is make this possibility “reasonably” small as perceived by public expectations.
We are plotting the left and right extent of the solutions as solid lines, but as in LSD, it should not be like this. As we move away from the centre point of the solution the risk is reduced. Perhaps we should plot these as “soft edges” to represent the reduced risk as we move away from the centre point. As previously discussed, the normal distribution of risk associated with uncertainty may look something like this.
And if we say that the engineer accepts that some small level of risk of failure is allowable then we could make the distribution of uncertainty less and more like this.
Unlike the prescriptive nature of Limit States Design found in prescriptive uncertainty, the level of reasonable error in embedded uncertainty is subjective; some designers may not be comfortable with any level of error, some with more. Their tolerance to risk will determine their range And in some cases it will be difficult to determine the level of distribution and hence difficult to determine the level of error.
If we revisit the idea of engineering judgement, perception of risk from an engineering judgement perspective makes a difference with this distribution. One person may judge uncertainly differently to another. One person may subjectively associate a higher level of risk and uncertainty to be greater than another. Another person may “strategically” impose or reduce risk on the project such as when they want to be particularly helpful or unhelpful. And on a different day, the same person may rate risk differently, such as the day after they received a promotion.
There are other more complex impediments to making clear judgements on risk. A person may believe that they lack skill in a certain area, or be overly confident with their skill level. They may also not know that the risk exists or may be wilfully blind to the risk.
Daniel Kahneman refers to these different variations as noise. Although he refers to their use as they apply to personal judgements, these can easily be transposed to personal assessment of risks in engineering. We will revisit noise in decision making when we talk about managing uncertainty.
The result is a range of perceived uncertainty for the specific solution. This range is driven by engineering judgement, the designers tolerance to risk and the designers perception of the consequence of the risk. But it is not necessarily prescriptive; it is completely at the discretion of the engineer.
Engineer’s discretion to risk
In the introduction we defined the difference between uncertainty and risk as the known and unknown nature of the the distribution. Are the dice biased in a way that we cannot know? However, we are going to challenge this concept here.
For a given solution, with a set normal distribution to uncertainty, an engineer has the discretion to choose the range within the distribution in which they can tolerate the solution landing outside the range. In other words, an engineer has the discretion to choose a solution where there is a high possibility of the reality landing outside the range of embedded uncertainty. This may seem bizarre, but there is no mechanism for controlling this aspect of the design. For example, we can choose a 200 thick concrete slab, where our analysis shows that, due to embedded uncertainty, there is, say, a 50% chance that this solution is not appropriate.
There are a number of reasons why an engineer might justify this decision which we will explore. The most obvious is that, in our justification, we have safety factors within prescriptive uncertainty that we can consider as cumulative when undertaking a holistic assessment of the solution.
Self assessment of embedded uncertainty
I made pancakes with my 10 year old son using a recipe from Jamie Oliver. I am not a natural cook, so Jamie’s recipes are great for me. But this pancake recipe was dumbfoundingly simple. It goes something like this. Grab any measuring thing, doesn’t matter what. Fill it will self raising flour, fill it with milk, add them to a bowl with a cracked egg. Pinch of salt and a fit of fruit if you want. Mix and cook.
The risk here is devastating low. If we have a rating of 0 to 100%, I would give this a risk rating of 95%. It is hard to go wrong. But there is still some risk. Substitute self raising flour with plain flour, crack the egg and accidentally add some shell, pinch of salt is too big a pinch. There may be some error in the cooking—too much or too little. And there may be some error in the ingredients—the egg might be off, or ants in the flour (it has happened in my house!). So I don’t think we could ever say it is a 100% certainty.
Jamie’s cracking burger is a bit more complicated (not complex), but still substantially risk free. Probably a 90%. I would say Jamie rarely goes lower than 85%; his receipts are always readily achievable, even for the unskilled like me. I can only assume that this was his intent.
Go to the other extreme. Lets find a recipe that is near impossible for a skilled cook. The recipe is beyond simply complicated, it is complex. Complex would mean that a written recipe would not be able to capture the nuance of what actually needs to be done. Even a video showing the recipe being undertaken by a 20 year veteran pastry chef is impossible to replicate. This is very high risk. We will give this a rating close to zero.
A zero rating on our scale would be impossible to achieve. Everything between 0 and 50% would require some high skill and become increasingly difficult as the number gets smaller. We might even say that everything below 10% is near impossible.
We want to draw a line at 50% on our rating scale. This line is where we cross over from the complicated to the start of the complex. Although I could possibly attempt to do a recipe above this 50% mark, there would be near zero probability that I would be successful at anything below 50%.
Ranking system
We have used recipes to develop a ranking system. It sits on a scale between 0 an 100 as a percentage. Zero is an extreme of high risk where it is impossible to be successful, while a 100 is almost no risk where it is impossible to fail. A ranking of 50 is one where the transition in a level of uncertainty.
In a moment we are going to explore how we can use a similar ranking scale to rate uncertainty and use this 50 transition point to determine if a solution should be adopted. Before we explore this ranking system, lets highlight a few concepts that are key to understanding our interpretation of solutions.
Beginners luck and “Mount Stupid”
It is worth mentioning a few additional things here: Beginners luck and “Mount Stupid” are the first stage of the Dunning-Kruger Effect. If I try a recipe with an objective rating of 50% and succeed, I could simply have been lucky. This could be beginners luck. The first time I made rice noodles, from a pack, I did a reasonably good job—but I thought I could do better and had a few more cracks at it. With every progressively attempt, I got worse and, well, sadly, have not tried to do it again in several years. I attribute this to beginners luck. Which might be a good thing because now we go to the local Thai restaurant and I marvel at their expertise.
Adam Grant describes Mount Stupid as the arm chair coach, which is the first stage of the Dunning-Kruger Effect. I won’t try to paraphrase the explanation in depth but will simply say that there is also a point, early in your learning, when you perceive that you know everything about a topic. As you learn more, you begin to realise that you, in fact, know very little. You quickly dive into the valley of despair. The concept worth highlighting here is a person’s inability to make a self-assessment of their knowledge. It would appear that only once you start growing out of the valley of despair, could you begin to start making meaningful self-assessments.
Structural engineer as maverick
How does this relate to structural engineering?
Lets go back to the quote from the introduction. I will repeat it again so it is fresh in your mind.
The process of simplifying and idealising the real structure is referred to as modelling, although this term is also this term is also applied to the full process of creating the idealisation and then analysing it with an appropriate method of analysis. The idealisation is a gross simplication of the real building structure, and is referred to as a structural model. A further important step in structural modelling is to take the results of the analysis of the simplified model, and interpret them in relation to the behaviour of the real structure. This requires insight, and an appreciation and mature understanding of structural behaviour on the part of the structural engineer. It is an essential step in the analysis, and hence design, of a real structure. Design and Analysis of Concrete Structures
What are the authors actually saying here? This is one of my interpretations:
No-one is able to interpret a structural model with certainty. They say that someone with “insight, and an appreciation and mature understanding of structural behaviour” would be able to interpret the results. I am going to argue that no-one like this really exists, even someone who is far along the growth stage of Dunning-Kruger’s curve, ie. someone with deep experience and knowledge. Because what is an experienced structural engineer basing this ‘insight’ on? Few engineers, if any, test or monitor a finished structure to see how well it compares to their model. I therefore believe that it is overly generous to call this ‘insight’ and suggest that it is nothing more than unsubstantiated intuition, therefore by fallible by definition.
We cannot think our way into a perfect model like a true rationalist, nor can we test our way through engineering for things that don’t yet exist like a true empiricist; we must synthesis the two as Kant would do.
They are essentially saying that, even if a model that is reasonable could be created, a model can be interpreted in any number of different ways. I therefore interpret these statements as a call for structural engineers to be mavericks.
Chuck Yager epitomised an entire phase of history of the maverick. Yager spent over three decades test flying new aircraft for the US Military with little instruction. It would appear that he relied exclusively on shear intuition and mechanical sympathy. To suggest that structural engineers should act as mavericks, — exclusively use their intuition, to undertake engineering design— is a ludicrous proposition. There is a better way. And we will show you what it is.
Evaluating uncertainty
Lets turn now to our ranking system.
When we undertake engineering solutions we need to rigourously evaluate embedded uncertainty using statistic analysis. We want to develop a rating system that represents our level of confidence that the solution is correct. Rather than an arbitrary rating system that could be ambiguous in its answers and misinterpreted by individuals, we recommend using an evaluation scale ranked against known outcomes. We want a ranking system similar to our recipe scale.
Lets draw on the recipe analogy and build a ranking scale. We will use a percentage scale between 0 and 100% as a way to measure uncertainty.
What would represent a perfect score; the highest ranking on our scale? If the engineer believes that the solution exactly matches the ideal, then the rating for this solution will be 100%. You might think of this as the World Record Standard for Certainty.
As previously discussed, this is never possible for a single element and definitely never possible for a structure made up of multiple elements. Even in the laboratory, where the primary intent is to mimic the ideal solution, we cannot be sure that the experiment exactly represents the ideal. So if we believe the solution cannot represent reality exactly then the ranking of the solution must always be less than 1.
We will now define the mid-point on our rating system of 0.5. This point will represent a solution that is undertaken by a skilled engineered, has significant amounts of complexity, significant amounts of uncertainty, and, importantly, has very little redundancy. It could be said to represent a minimum standard of uncertainty within a solution. The skilled engineer believes the solution “reasonably” represents a solution that “will just work”.
A solution of 90% represents a solution that the engineer believes has a strong correlation to the actual solution. We are setting this point as a benchmark. It could be said that achieving a ranking greater than 90% is difficult. The solution would need to be simple, close to the ideal, load transfer through the structure is highly deterministic, and the solution could have some redundancy. We will explore the idea of redundancy in depth later.
Any solution less than 50% will represent a solution where the engineer has a belief that the solution may not work. This means it is not a solution that should be used. It represents a point where a different or alternative solution should be explored. One options for an alternative solution could be one where there is additional redundancy.
We will say that acceptable engineered solutions will, therefore, live in the 50% to 90% range.
This scale will allow us to rank our solutions as it relates to embedded uncertainty.
Uses for the ranking scale
The ranking scale allows us to evaluate the uncertainty embedded in our solutions. If an engineer is confident that the solution is a good representation of the solution then they can scale the solution in the range of say, 75% to 85%. Examples of solutions that may strongly correlate to the actual answer might be:
- Strength based analysis of a simple concrete structure supported on columns on a regular grid,
- Strength based analysis of a simple steel framed structure with pin connections throughout with the connections designed as pin type connections,
- Strength and serviceability analysis for a simply supported glulam timber beam in the roof of a residence, supported on a timber framed stud wall.
- Strength based analysis of a mass timber building with multiple levels of redundant structure. Where the redundancy is a result of the inherent nature of mass timber typically being more timber than what would normally be required.
- Serviceability based analysis for a multistorey timber framed building for lateral deflections where it is known that the structure can tolerate high deflections due to its ductile properties.
The above examples give a solution where there is only a small amount of uncertainty. There is often a level of certainty from a number of different perspectives. Such as the ability to model the structure accurately, including the boundary conditions and connection detailing. And there may be redundancy in the solution, not only in the volume of material, but in the way the material is model (pin connections). Or the solution is inherently conservative (moment redistribution in a concrete slab).
At the opposite end of scale, but still within the range of acceptable solutions are solution that live in the range of 50% to 60%. These are solutions that are ranked as ones that are just working. Examples of solutions that may have significant uncertainty but can still be classified as working might be:
- Serviceability analysis of a concrete slab supported on columns of an irregular grid.
- Analysis of a cantilevered glass balustrade supported on two isolated pin connections.
- Serviceability analysis of a timber rigid moment connections using bolts threaded through holes in the timber where the holes are oversized.
- Strength analysis of a timber connection relying screws that are orientated such that they are in shear.
- Strength analysis of a multistorey concrete framed building with columns offset on each level such that there is large areas of structural transfer and complicated load takedown calculations.
- Strength analysis of a complicated and multidimensional steel framed building where lateral loads following complicated load paths through the structure.
What is important here is that it will take some expertise and experience in order to make the assessment of uncertainty. But we argue that being diligent and particular about making these assessments, novice engineers can quickly pick up on the strategies for doing these assessments. They will also consciously consider their work using this criteria. As the solution becomes more complicated they will be more conscious of moving away from the ideal.
The special case of redundancy
Consider for a moment a cable stay bridge and a suspension bridge.
Both bridges have key structural elements, we will call them ‘primary elements’ that, if they were to fail, the entire bridge would fail. In both cases these include the towers. In the suspension bridge there are two additional primary elements: the main cable and the anchors. In both bridges the secondary elements are the cables.
Therefore the primary and secondary element types for the two bridges are:
Bridge | Number of Primary Element Types | Number of Secondary Element Types |
Cable Stay | 2 | 1 |
Suspension Bridge | 4 | 1 |
As we mentioned, if the primary elements were to fail, the entire bridge would fail. It is possible that the bridge could be engineered such that if one of the secondary elements (the cables) were to fail, the bridge could continue to perform.
If we take this one step further we can see that we could engineer the bridge such that multiple secondary elements could fail, provided they are not adjacent cables. Clearly there would be a limit to the number of failed secondary elements.
Types of redundancy
If we are going to introduce redundancy into the bridges, we have to consider the three different types of redundancies we have defined above, and determine which is appropriate for the element:
- Primary Element redundancy,
- Secondary Element redundancy, and
- Multiple Secondary Element redundancy.
If we are to introduce redundancy into these bridges, we would need to approach redundancy using three different methodologies, one for each of the types of redundancies. Firstly, for the primary elements, the redundancy would need to be inherent in the element itself. For example, if the tower was constructed of concrete, there would need to be redundancy within the concrete structure itself.
Secondly, for the secondary elements, the redundancy would not necessarily need to be in the secondary element itself (one element could fail), but in the element that it is supporting. In the case of both of the bridge examples above, the deck of the bridge itself could be engineered such that a single secondary element could be removed.
The third type of redundancy, multiple secondary elements, is a specific case where a risk assessment would need to be undertaken on the possibility of multiple secondary elements removed and the consequence depending on which ones are being removed. In the case above, we could consider if two adjacent secondary elements fail, or alternative secondary elements, etc.
Mixing performance criteria within redundancy
We suggested earlier that when exploring redundancy, it relates to one performance criteria. It would be typical that the performance criteria being considered is the criteria of failure, or strength limit state. Another performance criteria is serviceability limit state such as total deflection.
Selecting redundancy
Introducing redundancy cannot be random. There might be little point introducing redundancy in an element that is not the weak link within the system, or to an element that already has another level of redundancy. The ease to which redundancy can be added to different elements must be considered. A risk analysis should be undertaken also, combining probability and consequences.
How does redundancy affect the level of uncertainty?
How does redundancy change the ranking of the solution on our uncertainty scale?
In 2011 Yoshihiro Kanno and Yakov Ben-Haim in Redundancy and Robustness, Or, When is Redundancy Redundant? offered a definition of redundancy which we will use here: “redundancy of a structure refers to the extent of degradation which the structure can suffer without losing some specified elements of its functionality.” We must first identify the functionality to consider, and then we must consider redundancy for this function. In other words, if a structural system has a functional requirement to remain standing (not collapse), and a single structural element in the structural system is removed, degrades, or fails, and the structural system does not collapse, then there is redundancy. If the structural system does collapse then there is no redundancy. Importantly, the redundancy criteria relates only to this single structural element.
What is interesting here is that we have no idea which element will fail—we have uncertainty. Kanno and Ben-Haim connect this uncertainty to redundancy through the use of the term ‘robust’. They state that “redundancy is related to robustness against uncertainty.” As we have just described, they support their statement by saying, “for a real-world structure we do not know in advance how many and which components will fail.” In other words, we are uncertain how the structure will fail, or what single element would fail. Therefore, for a system to have robustness against uncertainty we want to develop a solution that is agnostic to how and where the structure fails: we want the solution to be robust.
They introduce the term to define a level of uncertainty: “Because we do not know the extent of structural damage which will occur in the future we will be refer to as the horizon of uncertainty.”
So how do we represent redundancy in our ranking system? Does it shift the ranking, or does it reduce the range, or both?
The initial concept to consider is this: each of the individual solutions: the original structure and the element that provides redundancy must pass the test of adequacy (be greater than 50% on the ranking scale) before we start to consider the effects of redundancy on uncertainty. In other words, there is little point considering redundancy if one or both of elements are likely to fail. Another way to look at this is that you cannot use a redundancy to rescue a solution from likely failure. This means that we cannot take a solution that is 40% on our ranking scale, add redundancy, and move the solution to 50%, and then say that it passes. The reason this does not work is because this concept does not pass the definition of redundancy. If the redundant element fails, the structure is back to 40%, and unacceptable. Therefore it is not a redundant element.
We propose that the methodology to use when considering redundancy depends on the type of redundancy from our three types and what the redundancy is. With this determined, the way to account for redundancy is to determine a cumulative ranking on the solution holistically, considering both the original solution and the the redundancy.
The goal of redundancy is to reduce uncertainty. Ideally redundancy should increase the ranking of the solution and decrease , the horizon of uncertainty.
Durability
In the 5th century the first timber poles were driven into the waters around Venice. The island that housed the refugees who fled to the lagoon after the fall of the Roman Empire became overcrowded. They didn’t realise it at the time but when timber is permanently submerged in salt water it petrifies, effectively transforming into a robust material able to last for centuries if undisturbed. This is unique to timber in this environment, and luckily we all get to reap the rewards of this random happenstance.
Unfortunately, not all structures have this level of durability. In a harsh external environment, for example, exposed mild structural steelwork will deteriorate if left without a regular maintenance program. Special weathering steel was developed to combat this deterioration. Weathering steel is pre-rusted, with the rusted surface offering a protective layer to prevent further deterioration. If detailed appropriately concrete can be design to last a very long time—the key is to detail the reinforcement so that it is embedded in the concrete so that is is protected from the outside environment where it could rust. In an ideal world a structure will not deteriorate over time, it will be durable. Most structures will need to be maintained over time such that it will not deteriorate.
Importantly, most structures are designed for a specific life expectancy. The level of durability of the structure is specifically designed to last for this amount of time. At the end of this defined life expectancy, the structure will need to be assessed to determine its structural adequacy.
Assessments of uncertainty must consider durability. And it must consider durability for the prescribed life cycle of the building. From a sustainability perspective, we must take into consideration the procedure required at the end of life.
How to use uncertainty to your advantage
There are many examples of when errors result in new discoveries. In fact, many argue that most major innovations did not come from careful planning but through uncertainty. Or when information is relayed incorrectly but the misinterpretation leads down a different road of discovery. More than this Taleb argues that antifragile systems strengthen when they are stressed.
Uncertainty is not something that is going to go away. So rather than fight it, or see it as the enemy or inhibitor, we want to use it to our advantage.
The first step to taking advantage of uncertainty is to identify it. The second step it to rank it. The third step is to use it to your advantage. We are going to explore some areas where we can find these advantages.
Rethink assumptions
Einstein developed a thought experiment whereby he was riding on a light beam, travelling at the speed of light. Beside him was another light beam. To him, it was clear that the other light beam would travel away from him at the speed of light. But how can this be? If I am travelling at the speed of light, and it is moving away from me at the speed of light, it must be travelling at least twice the speed of light. But this defied the laws of physics. It was only when he removed one of his base assumptions that he was able to resolve the paradox: Time is not constant.
A project that we were working on had a retaining wall incorporated into the building. The building would house some expensive equipment in the location of the retaining wall. But there was uncertainty about the maintenance of the wall in the long term to remain waterproof. This uncertainty triggered a rethink of the assumptions embedded in the project. The key assumption is that when you are building on a sloping site, that the walls retain the ground at the back of the slope. This was a common solution, and the entire team had worked on projects that had this solution. It was through uncertainty of the long term maintenance that we question the assumption. The solution of battered the ground behind the wall so that they were not retaining, removing the water proofing issue. Two more unforeseeable benefits emerged from this design decision. Firstly, the wall could be changed from a core filled blockwork wall to a simple timber framed wall, reducing the cost and embodied carbon of the build. Secondly, we decided to include windows in this wall which increased the cross ventilation, light, and allowed views to the excavated rock face.
Uncertainty can be used to trigger investigations into assumptions. In this case we used uncertainty to question the need for a retaining wall at all.
Trigger innovations
Our team developed a solution that we ranked close to 0.5 on the ranking scale. The solution relied exclusively on the shear capacity of inclined screws for a connection detail for a prefabricated roof panel system. Although we believed that the solution was acceptable, we felt there was enough uncertainty in the solution that we should rethink the connection. After exploring a number of different options, we determined that a way to raise the uncertainty to 0.6, was to include some redundancy in the connection. We added ledger plate, which we discussed with the builder and determined would be a straightforward and inexpensive detail to incorporate into the solution.
Imagine that you have come up with a solution that does not pass, or is close to failing the ranking system. This could be used as a trigger to rethink the solution, perhaps even look for more innovative ways to explore the solution.
We will talk more specifically about innovations later.
Exploration
Both rethinking assumptions and triggering innovations are mechanism to explore. Knowing that uncertainty exists, opens up ideas for exploration more generally. When you know that uncertainty exists, exploring a single option and then reading the results at face value is rarely the best methodology to explore a solution. Not only could exploration be triggered by the need to find a simpler solution, but could also be used to explore an option that has less uncertainty.
For example, if you were to produce a solution that has an uncertainty level of 0.5, which is a high level of uncertainty, it is almost certainly a trigger for you to explore a solution that is simpler or less expensive or more efficient or better suits the brief, in other words one that has less uncertainty. Developing a solution with more certainty would allow you to shift the solution to the left, closer the cliff. Having a higher level of confidence in the solutions allows you to develop a solution that is more sustainable.
Ranking
The process of ranking uncertainty brings you closer to understanding it. Instead of relying on intuition and instinct to guess a solution, ranking a solution specifically can lead to more intelligent decisions.
In the context of structural engineering, there are many variables that affect certainty, and reducing uncertainty can lead to more sustainable solutions. However, it is important to recognize that uncertainty is inherent in the design process, and striving for absolute certainty may not always be practical or even possible.
One helpful strategy for managing uncertainty is to search for redundancy. When we identify areas where multiple systems or processes overlap, we can begin to build in redundancies that help reduce the impact of failures or unexpected events. This approach can be particularly effective in situations where the stakes are high or the consequences of failure could be severe. By investing in redundancy, we create a safety net that increases our overall resilience and gives us greater confidence in our ability to weather unexpected challenges.
Gaining insight
We talked early about insight cannot be obtained by simply being successful. When a 200 thick slab will work and you randomly choose 250 or 300 because it is safer, and this solution works, there is no insight gained.
But when you rank the uncertainty, it gives you a framework by which to gain insight. Reflecting on the solution, analysing the success and provide a method for repeating the design with a more refined design. The key is to know the elements of the design to retrospectively rate and measure.
Gaining insight by reflecting on the outcomes of the design is rarely undertaken in engineering unless there is a failure. There are several challenges with failure is that it is rarely done within the realms of uncertainty. Also, failure is often more than one element that contributes to the failure and isolating different elements is challenging. And there is often a motivation to find someone to blame. And more often than not, expert advice is not delivered in a way that embraces uncertainty, but one that is overly deterministic. As if there is one right answer, and the solution chosen was clearly wrong.
Reduce noise
Making decisions based on rational thinking gives designers the opportunity to reduce noise and ensure that decisions are clear, clean, and based on rational thought.
There are several different types of noise, as presented by Daniel Kahneman and discussed here.
The paper discusses two key types of noise that affect embedded uncertainty: perception and noise. Perception refers to the subjective interpretation of information, which can vary from person to person. This can affect the structural modeling and interpretation of results. Noise refers to the various factors that can affect uncertainty differently for each person, such as cultural background, individual experience, and bias. The paper also touches on the concept of stable pattern noise, which refers to the inherent idiosyncrasies and differences in how people approach and solve problems.
The concept of "level noise" refers to the inherent variability in engineering judgement and decision-making. This noise can arise from a variety of factors, including personal knowledge and experience, differing interpretations of information, and uncertainty about future outcomes. It is important to recognize and account for this noise in the design process, in order to arrive at the most effective and sustainable solutions.
Reconciling Uncertainty
One of my very first projects as a junior structural engineer was the Maritime Museum in Sydney, Australia. I attended a couple of meetings with my boss, several of the meetings included the client. During one of the first meetings my boss asked the client an unusual question. “What is your tolerance to movement in the building?” he asked them. “We are doing a maritime museum,” he continued, “boats move and sway. Wouldn’t it be interesting if we allowed the floors to move and vibrate under our feet, much like a boat?” They loved the idea.
I remember walking through the building when it was finished and experiencing the movement in the floors. The movement was obvious without being worrying. We had nailed the brief. If I had not thought my boss was a genius before that day, my decision was confirmed on that day, and stays with me today. I just wish I could have remembered all the other things he told me, which would have washed over me at the time.
At the time of these early meetings, I had no idea why he asked the question and what difference it would make. It was only about six months later, when I was actually engineering the floor members, that I realised how important a question this was. The size of every single floor element, without exception, was dictated by “normal” deflection requirements. By letting go of this criteria, we were able to reduce the size of the members, saving tonnes of steel and thousands of tonnes of embodied carbon.
My boss’s question had three interesting concepts. It had insight, intelligence, and understanding. He had insight into the clients, he was empathetic to their brief. He had intelligence in that he knew deflection was going to be a governing criteria and he had identified how to remove it. He had understanding that the client wanted an economical structure. In hindsight it almost seems obvious, but watching hundreds of these early conversations, I am consistently surprised at how obvious questions like this are never asked. We are going to explore why this happens.
Lets go back to my old boss’s question about the floor deflections for the maritime museum. Why are these questions not asked? Structural engineers typically play a passive role. “Give me the final drawings and I will give you the answer,” Might be a common call cry. But why don’t they play a more active role? A key component is the want to avoid conflict, avoid embarrasment. But the ecosystem may not allow this type of question. Is the structural engineer invited to ask such a question. Don’t agree too early, dont be eagerly compliant.
These are the areas where we should go to find solutions to reducing the uncertainty. It is through the design environment that we can look to reduce uncertainty.
The Design Environment
In the bustling city of Sydney, a team of architects, engineers, and builders have come together to design a new skyscraper that will revolutionize the way we think about sustainable buildings. They believe that the key to building a sustainable future lies in four principles: understanding, questioning, trusted ecosystems, and invitations.
The team starts by focusing on the principle of understanding. They believe that in order to create a truly sustainable building, they must first understand the needs of the people who will be using it. They spend months conducting interviews and surveys to determine what features are most important to the building's users. They learn that people want a building that is energy-efficient, has plenty of natural light, and is made from sustainable materials.
Next, they move on to the principle of questioning. They challenge assumptions and biases by asking questions rooted in curiosity. They ask themselves why certain materials have been used in the past, and whether there are better alternatives available. They question the idea that sustainable buildings are more expensive, and explore ways to make them more affordable.
The team then turns to the principle of trusted ecosystems. They believe that building a sustainable future requires collaboration, co-evolution, co-creation, integration, and synthesis. They work closely with builders, engineers, and other consultants to develop a holistic approach to sustainable building design. They collaborate on every aspect of the project, from the initial design to the final construction.
Finally, the team focuses on the principle of invitations. They believe that in order to build a truly sustainable future, they must authentically invite the client, builder, structural engineer, and other key consultants to be a part of the design process. They listen to their ideas and incorporate them into the final design. They believe that by working together, they can create a building that is truly sustainable and meets the needs of everyone involved.
As the project progresses, the team faces many challenges. They encounter resistance from traditional builders who are skeptical of their sustainable design approach. They face budget constraints and tight timelines that threaten to derail the project. But through their commitment to the four principles, they are able to overcome these challenges and create a building that exceeds everyone's expectations.
The finished skyscraper is a marvel of sustainable design. It is made from locally sourced materials, utilizes renewable energy sources, and has an efficient water management system. The building is flooded with natural light, creating a welcoming and energizing environment for its occupants. And perhaps most importantly, it was built collaboratively, with input from everyone involved in the project.
As the team reflects on their work, they realize that the principles of understanding, questioning, trusted ecosystems, and invitations are not just important for building sustainable buildings, but for building a sustainable future in general. They believe that by embracing these principles, we can create a world that is more equitable, just, and sustainable for everyone.
Buy-in
I received a call from Walter, the builder, regarding our collaboration on a child care and community centre project in Taree. We had been involved in the project from its early stages and were key contributors to the design of the centre.
"I am going to send you a photo," Walter said.
Shortly after, a photo of a tree arrived on my phone. I squinted at the tree, thinking I saw an owl or maybe a koala.
"Is there a bird in the tree, Walter? I think I see something, but I can't tell," I asked.
"No," Walter replied. "I am clearing the site and cutting down this tree."
"OK," I said slowly, not fully understanding what was going on.
"It's a Bloodwood," Walter added. There was a pause. "Do you know what I am thinking?" he asked.
Suddenly, it dawned on me. "Ah, I got it. We can use it!"
"Yeah, what do you think?" Walter asked.
"Well, Walter," I said calmly. "There is no way that I am going to stop us from doing it."
Six trees were cut down from the site and used to support the roof of the centre. They were situated outside the building line, polished, and meticulously detailed to support the large timber glulam beams that held up the roof.
In our experience, a factor that is often underestimated is buy-in. We have repeatedly demonstrated that buy-in on a project has a significant impact on our level of uncertainty. If an engineer is given the opportunity to co-create a project, their desire to make the project successful is high. This desire translates directly to an increase in a tolerance to uncertainty. It not only narrows the normal distribution curve, but also narrows the range of risk that they will tolerate. They are more willing to take on risk with the potential solution falling outside the zone defined by the arrows. These environments are typically safer to work in. Ecosystems that revolve around letting everyone into the design are typically psychologically safe. In fact, we have never been in an ecosystem where this is not the case.
Skill and Experience
Another limitation is the reliance on expert opinion, and while this is [often] treated in a transparent way, it is necessarily subjective and therefore potentially contentious. Pon and Kirk???
The more skilled an engineer is, the better they are at managing uncertainty. When an engineer successfully completes a project that meets the design criteria, the experience gained from that project can be beneficial for managing uncertainty in future work. Through this experience, engineers develop customs or heuristics that they know to be effective, and these heuristics are reinforced as they are applied repeatedly.
However, the challenge with experience is that structural engineering is not a singular, verifiable event, and the results are not explicitly analyzed to determine how closely the model matches the predictions from the analysis. Without this statistical data, it's possible, and even likely, that an engineer could produce unsustainable solutions throughout their entire career.
Moreover, we argue that experience can sometimes be a hindrance. If a structural engineer has had a negative experience with a specific type of problem, they may associate more uncertainty with that problem than someone who is encountering it for the first time.
In addition to the challenges related to experience, we must consider the impact of noise on embedded uncertainty, which can affect each person differently. Noise comes in various forms, and it's important to understand how it affects uncertainty to reduce its impact.
There is ongoing research to refine the methodology for analyzing most common materials like timber, steel, and concrete. However, these materials have different methodologies, and some regional codes offer options or do not directly address the problem, leaving it to the engineer to choose a methodology that they believe to be reasonable.
Similarly, there are various analysis packages on the market today, which allow one to choose the specific methodology to use in the design. However, with so many variables affecting certainty, we must question the need to spend time increasing it. For example, concrete is a material that is created on-site, influenced by a multitude of factors that are site-specific. Even with an enormous amount of effort, energy, and money, it's unlikely that we will ever fully understand concrete to an enormous amount of certainty.
Time
The level of uncertainty can be affected by the amount of time spent on a solution. One might assume that more time leads to less uncertainty, but research has shown the opposite is often true. In some cases, the more time spent, the less certain one feels about the outcome, making too much time detrimental. Instead, a sweet spot exists where there is a reasonable time constraint that allows for contemplation without causing undue stress. Teresa Amabile has conducted detailed research into this critical balance of time.
Premortem
There is a common strategy for testing a structural engineering solution. Imagine that the solution was to fail. How would it fail? Well let’s stop that failure mechanism. This is similar to the procedure of the premortem introduced by Gary Klein. The idea is that when you develop a solution and commit to it, doubt is suppressed. This leads to overconfidence. This is often the enemy of engaging uncertainty. The premortem legitimises doubt and encourages supporters to engage with uncertainty.
Complexity
One year my son played on a football team with a player who was a national 100m sprinter. He could outrun anyone on the field. He played in the back line, in defence. Three or four times in a game he made these spectacular runs to shut down anyone who made an attempt to get near our goal. Seemingly coming out of nowhere he swoop down, outrun them and usually with a spectacular sliding tackle, stop them in their tracks.
These events would be followed up by the crowd going wild.
After a few games, I noticed a pattern. The player would pick up the ball in front of his own goal as they played out from the back, and dribble the full length of the field. However, by doing this, he would get out of position by the time he reached the other end of the field. He would then get tackled and lose the ball, allowing the other team to quickly counter-attack. It was only at this point that he would make his lightning-quick dash back the full length of the field and make his trademark extraordinary tackles to the delight of the parents. They would exclaim, "He's a hero! He saved a certain goal!" He was undoubtedly a skillful player, but he created complexity and then solved it himself.
Engineers have a similar propensity. They get great satisfaction out of solving very complex problems. Many of the most spectacular solutions are required for spectacularly complex problems.
It is actually not in their interest to simplify a problem. They would never get the cheers from the crowd. There is no fun in changing complexity into something more simple. The level of satisfaction from removing the complexity would not be as spectacular as struggling your way through complexity.
It is also self-serving. Keeping it complex brings perceived value to a project. You would be much more likely to value someone who solves complex problems, than value someone who has the skill to change a complex problem into something manageable and solving these more simple problems.
But I argue that this simplification of complex tasks is what skilled engineers do, and is in fact how they solve a complex problem, even if they are not aware that that is how they do it.
There is another benefit to simplifying a complex problem. It makes uncertainty more clear. Rather than view uncertainty within the entire complex problem, where uncertainty may be seemingly impossible to clearly define, we can analyse uncertainty for each separate part of the problem. It changes uncertainty from some unwealdly, to something manageable.
A complex system - TalebDevelopment of Customs
Customs, or heiristics, are rules that individuals or groups follow to complete their work. Customs are interpretations of the standards that are specific. They can be individual or group or company based. They are a methodology for dealing with uncertainty. They take a range of possible options and narrow the range so that there is less ambiguity in interpretation of the choices.
Customs can be developed for specific elements of a material, such as the yield strength of steel, or they can be developed around entire concepts that can be ignored in a design, such as ignoring the earthquake requirements for a large, light weight building when it has been shown that wind is a governing factor for previous similar buildings.
An example of customs used by structural engineers includes customs on deflection criteria. The codes have a series guidelines that are not requirements that can be freely interpreted by the individual engineer. Companies typically have customs on how to interprete these guidelines. Customs are typically based on experiences. Solutions are trialed and are demonstrated to have challenges, so they revise and update their customs.
What we need to specifically be aware of (Koen 2003) is that, by definition, heiristics are fallible. This was true with the Tacoma Narrows bridge (Petrosky 1994).
Prefabrication and DfMA
It is worthwhile discussing prefabrication and DfMA. Much uncertainty arises from the variability of the workmanship that occurs on site, and the challenge with requesting on-the-spot changes with the pressures that exist on site. The skill level that exists on site often is not a function of the reputation of the contracting company, or the team that is doing the work, but the individual person that is actually hammering in the nails. When the designer turns their mind to this specific person, which is often chosen as a completely lottery, their perception of the risk of the specific task will increase.
When construction work occurs in a factory, where there is opportunity to explore, make changes, refine and play with the construction in a comfortable factory that is not exposed to the external weather, not only can the uncertainty be reduced, but the designers perception to the risk can be reduced.
The work undertaken within a factory is not undertaken by a specific person. It is supervised, reviewed, scrutinised and refined. It is not at the mercy of an individual person, the work is undertaken by a highly skilled team.
There are lots of comments made about the advantages of DfMA. These revolve around savings in construction time, more consistent quality and just-in-time construction. But there are none that focus on the significant contribution it can make to the reduction in the perceived risk that the designer feels about the solution. In our experience, when the designer knows that the construction will occur in a controlled environment with high levels of supervisions, and where they are more free to offer revisions, suggestions and improvement, their tolerance to risk increases. This increased tolerance for risk results in a decrease in uncertainty.
In the construction world, where the level of building is often dictated by the lowest standards allowable by the local governing codes, DfMA allows for a standard that the designer, working in collaboration with the fabricator, is able to control. DfMA has the potential to allow for the level of uncertainty to be eliminated completely.
DfMA is significant: we suggest that DfMA has the potential to completely remove uncertainty. It is highly predictable. It has the potential to be the most significant contributor to producing the ideal solution. A solution that potentially sits at the cliff face, in the location of the ideal solution, the most sustainable solution that can exist. In fact, we will claim that the DfMA is the only solution that can sit at this point.
This paper is not about DfMA, it is about how uncertainty influences the level of sustainability of a solution. Within the current environment, designers are forced away from this ideal by inherent uncertainty. If the construction community is to innovate for the substantial benefit of society, we should make the connection between DfMA and its potential to produce ideal solutions. We would identify the potential of its ability to produce highly predictable solutions and see how this would change designers view of uncertainty.
The Boundaries of Certainty
There will never be a time in the future when we will be able to completely reduce uncertainty. There are boundaries to certainty. So where are they? And are these boundaries moving?
Certainty boundaries in prescriptive uncertainty
This is well understood when it comes to prescriptive uncertainty and it is built in to the design in a highly conceived methodology. The rare events when the loads exceed those that are prescribed, or the material has a flaw are well catered for in structural design using limit state design.
Further to this, it is unreasonable to allow for Black Swan events in the design: events that are not only outside the predicted for the structures prescribed use, but also outside a reasonable safety buffer defined by probability statistics and risk assessment.
At any point in time, the boundaries to prescriptive uncertainty are defined based on a risk assement from statistical data. The statistical data is collected from past events. The hope is that past events is a fair predictor of the future, particularly if the data is collected over a relatively reasonable amount of time.
Examples of boundary changes in prescriptive uncertainty
Here are two examples of where the boundaries in prescriptive uncertainty have changed. Both are from Australia. In 2005, the Australian codes changed the factor used in limit states design for permanent loads. The factor, which can be viewed as a safety factor, changed from 1.25 to 1.2. This change can be interpreted as a reduction in uncertainty. The committee’s justification for this change was that, with the advancement in construction techniques, and the increase in the precision of materials, permanent actions can be predicted at a higher level. In other words, there was more certainty in the physical material that was placed on site. This was based on actual measured materials on real construction sites.
Another example from Australia are the wind loads in Sydney. Weather gauges placed on the Sydney Harbour Bridge have been collecting wind data since 1979. When the codes were first formulated, the tables were based on a smaller sample period. With the increase in the sample period, a better prediction of the future wind trends can be predicted. It is inevitable that the winds are likely to change again in the future.
Certainty boundaries in embedded uncertainty
But what about embedded uncertainty? Will there ever be a time in the future, when a number of engineers, using different analysis packages to model their solution come up with the same result for the same problem? The answer is clearly ‘no.’
Human factors affecting embedded uncertainty
Two key human factors will consistently affect uncertainty: perception and noise. The reality is, when a situation is described to a person, whether with drawings, verbally, within a collaborative environment, or a combination of methods, each person will have a different interpretation of the information. Their own personal subjective perception will affect the solution. How does this play out in the design process?
This perception will affect the structural modelling that they undertake to determine a solution and will also affect how they interpret the results of the model.
We have spent some time in previous chapters describing noise. Noise, in its various forms will affect embedded uncertaintly differently for each person.
Research into materials
There is continuous research being undertaken to refine the methodology for undertaking analysis. Most common materials, such as timber, steel and concrete, have accepted methodologies for assessing the structural capacities of materials. Interesting, they all have a variety of methodologies. Many methodologies are written into the regional codes. However some regional codes offer options or do not directly address the problem and therefore, leave it to the engineer to choose a methodology that they believe to be reasonable.
Software analysis packages
There are an unthinkable number of analysis packages on the market today. Despite enormous research being undertaken with different materials, there are, and always will be, a variety of ways to analyse a structure. And a variety of ways to analyse a material. Many analysis packages allow you to choose the specific methodology to use in the design.
But Why?
The challenge we have is simple. If there are so many variables affecting certainty, why do we need to spend time increasing it? Take concrete for example. We could spend huge effort, energy and money refining the analysis so that we understand concrete to an enormous amount of certainty, but the reality is that concrete is a material that is created on site, which is influenced by a multitude of factors that are site specific. We may know concrete it an incredible level of precision at some point in the future, but when it is influenced by these unpredictable on site variables it is clear that there will always exist some level of uncertainty regardless of our precision about the material.
Beyond Uncertainty
The word “uncertainty” carries negative connotations. It is even more disconcerting to think that uncertainty is embedded in maths, science, and engineering where, at first appearances, certainty is the ultimate goal. Centuries of research and millions of human hours, all at extraordinary expense, has been dedicated to providing structural engineers with more and more levels of certainty. Lets for a moment imagine that we are perhaps 70% of the way to being able to provide absolute certainty in embedded uncertainty (the unknown unknowns) in terms of knowledge, I will argue that even getting to 71% will take an unimaginable about of effort and time. This is the theory of diminishing returns.
So the question we must ask ourselves, as an industry, is “What is the ultimate goal? What level of embedded uncertainty will be be comfortable with?”
I have presented arguments for using uncertainty for advantage. But now I want to explore powerful factors that are outside of the scope of uncertainty that we can use. I am not saying that these factors don’t have uncertainty within them, but they are concepts that we can explore without uncertainty being a driving factor. These concepts are “beyond uncertainty.”
Innovation
We were working on a new multi-unit development on the lower north shore of Sydney. Through a series of “innovations” we ended up saving the client over $1,000,000 in excavation and structure.
The development was a townhouse development with each townhouse consisting of three levels. Below the units was two levels of below ground carparking. The house were at the rear of a battle axe shaped block that sloped up from the street. It was proposed that the driveway come in level to the street and enter straight into the upper level of the carpark. So it was excavated into the ground.
What was interesting about the layout was that the below ground car park had no obvious correlation to the townhouses above. And when you looked at the architectural drawings the lack of alignment was not obvious. The plans for the ground floor did not show the car park under and the plans for the car park did not show the townhouses over.
We diligently went about our work and engineered a ground floor transfer slab. At its maximum the thickness of the transfer slab was 900mm deep. The architect had shown this slab to be 300mm deep and we quickly realised that 900mm would require some architectural changes, perhaps even make the solution impossible to achieve. It would have been simple for us to stop there present the solution and throw our arms up in the air when the solution was not feasible. But that is not how we think engineers should work. We went about exploring other options.
While maintaining the number of car spaces, we shifted the entire car park so that it aligned with the building over, we aligned the car spaces with the walls of the townhouses and shuffled the ramps at each end outside the lines of the buildings. In this way we were able to completely remove the transfer structure and reduce the slab to 280mm at its maximum.
Not only had we come up with a more economic solution, we had simplified the solution and reduced the embedded uncertainty.
The discussions so far have been about taking a single solution and exploring directions or processes that may reduce the uncertainty. Our example of the concrete slab spanning between four columns on a 6m x 6m grid was presented as a given for the use of our discussion. But how did we come to this decision; what happened in the feasibility stages and preliminary stages to land on this geometry? We spent our time discussing and optimising this solution, a solution that was given to us and was, perhaps, guided by an architectural design decision. How did we get to the point where this arrangement was deemed the most appropriate?
We have said little so far about the fact that there may be a completely different solution that could be a more sustainable option. We are going to use the word “innovation” here to very broadly mean a different, whether it is obvious or not, solution that varies from what we may refer to as the obvious solution.
It is worthwhile categorizing innovation into two separate categories. Those that “play” within the same general boundaries or resemblant innovations, and those that are completely different solutions altogether or disparate innovations.
Resemblant innovations
The different solution could use different variations of the same material, for example, a different concrete strength. A different solution could use a completely different variety of the same material, for example post-tensioned concrete, instead of conventionally reinforced concrete. These are examples of resemblant solutions. They are subtle differences of the same solution just minor shifts in materials or similar versions of the same material.
We could summarize this types of solutions by saying that the architectural design does not need to be altered to achieve the variations within the solution.
The key to resemblant solutions is that it is almost certainly feasible that these options can be plotted on the same graph in direct comparison. They are close enough in fundamental arrangements that they are comparable.
We hinted previously that the solution could be represented by two lines, forming a vertical zone within which the solution lies. In our example we kept everything else constant and changed the concrete slab thickness. This is a base assumption that is rarely questioned. We did this presumably because there is a larger range in concrete slab thickness than in other variables such as reinforcement grade of concrete strength. In this way changing the thickness move the solution left to right along the X-axis and the other material changes moved the solution in the Y-axis. For example, a post-tensioned concrete slab using a high concrete strength may move the solution to top of the zone resulting in a highly efficient structure using less materials. To add some complexity to this option, it may be the post-tensioned concrete may introduce higher uncertainty. This will shift the shape of the zone. We need to take this into consideration.
This highlights the fact that we should proceed with caution as we could easily over simplify the complexity, even for minor resemblant solutions.
Disparate innovations
A different solution could use a smaller or larger column grid system. A different solution could use complete different materials, for example, a steel framed structure instead of a concrete slab. A highrise building constructed as a concrete frame instead of a steel diagrid. As you can see, it would difficult to include an exhaustive list of all the possibilities. These are examples of disparate solutions, they are fundamental changes to the design beyond just the harmonisation of architecture and structural engineering.
We could summarize this types of solution by saying that the architectural design likely needs to be altered to achieve the variations.
The key difference with disparate innovations, is that it would be challenging to place disparate innovations on the same graph. And, in fact, we argue that we should not place them on the same graph as it might be adventageous to think of disparate innovations completely independently. A comparison could then be made between this disparate innovations by a more holistic approach to the project, using a more broad success matrix.
We further argue that disparate innovations are more closely related to what we would traditionaly refer to as truly innovative solutions, while resemblant solutions may not typically be seen as innovative as they are likely more solutions that a structural engineer explores on a day-to-day basis, and can typically be explore within a standard engineering design software package.
We recommend that disparate innovations be explored exclusively in the early feasibility stage of a project, while resemblant innovations can be explored at any stage in the design stages to refine the design. Suffice to say that the decision on this final arrangement needs to be diligently investigated. The matrix on which this decision is made on the choice of disparate innovations is multifaceted and needs to be made on the merits of the project, without the biases and motivations of individual team members.
We should not underestimate the importance of both of these methods of innovation. Both are important.
Regenerative Design
The methodology described here is essentially a drive to optimize the design. A highly sustainable solution, one that uses the least amount of material for example, is one that is optimized. Or if the solution is explored with options of different materials, then a sustainable solution is one with the least embedded carbon.
Regenerative Design does not necessarily support a singularly focused approach of a highly optimized solution. It supports a solution which also considers regenerative aspects. These include aspects such as adapability, durability, beauty, transportation, disassembly and reuse to name a few. Inherent in some of these ideas is the notion of including some redundancy, beyond the factors of safety embedded in Limit States Design.
We could argue that, because it is near impossible to get an ideal solution, and therefore it seems pointless to try to achieve it, regenerative design may offer a superior methodology for choosing a solution. What is important to take into consideration is that regenerative design should be explored holistically. As a very simplistic example, there is little point engineering a super-structure that could support additional floors, or an increase in load allowances, if the footings are not engineered to also support these loads. In fact, we argue that if anything should be engineered to support additional loads it should be the footings because these are the most difficult to retrofit.
Conclusion
We have explored how uncertainty can determine the level of sustainability of a structural engineering solution. Reducing uncertainty will lead to a more sustainable solution because we can then get the solution closer to the ideal.
The key reason for uncertainty being a driving factor for a sustainable solution is due to the asymmetry inherent in adequacy considerations in the structural engineering design. A cliff exists that defines this asymmetry. A small move to the left will put the design over the edge of the cliff and into the inadequate. A small move to the right will put us in the ideal solution with the highest levels of sustainability. If need to stay away from the cliff edge, we don’t want to take the risk that we could fall off.
Uncertainty is created on many levels and we explored where uncertainty can come from and how we might look to reduce it.
We also explored the concepts of innovation and defined it with two separate classifications: resemblant and disparate and how each of these can be used to guide a more sustainable design. We also reviewed the idea of regenerative design and how it can guide an entirely different methodology for design.
Notes
- Limit States Design, the philosophy followed by structural engineers globally, has two mechanisms by which to check if the structure is adequate: strength—structurally strong enough to support the loads without collapse or serviceability—structurally stiff and durable enough such that the structure meets the appearance, aesthetics and comfort requirements of the building.
- We are going to define sustainability in its broadest sense: less material, less embedded carbon, and highly efficient design.
- We will explore the idea that this level of failure plotted below zero may be a moot point. To fail in a structural design, is to fail, there may not be a “level” of failure. But as with all structural design there is no black and white.
Damian Hadley (c) 2022
damian@cantileverstudio.com.au
Damian is Managing Director of Cantilever Studio, a structural engineering firm focusing on projects embracing integrated design for high quality architectural projects.
Notes
The graph started as a simple concept. But it soon revealed complexity that we, at first glance, did not anticipate. We hope you come to appreciate not only its simplicity but its multiple depths.
> But bringing a statistical thinking as a way to interpere with our intuition is not comfortable. We would much prefer the idea that it feels right, or as Danny says our objective ignorance.
> Or we might talk about safety within our solution and give them the label of factors associated with the codes or LSD, when in fact they are associated with our uncertainty or doubt in our thinking or our limits in knowledge around this subject.
> These difference are enjoyable and interesting. We should not strive for consistency. We need to embrace these idiocyncrocies. Stable pattern noise.
> Some of the undying principles or values that drive risk aversion of a particular uncertainty may be quite complex, and the engineer may be completely unaware of them. We need to be more calculated when we come up against these uncertainties. We need to make sure we are using the right scales. You may have been stung once in a freakish and unusual context and now you are overly cautious for no good reason.
Regenerative design would not favour lowest standards. Lowest standards often result in a structure that does not meet the clients expectations. Because by definition it is the lowest standards, it is the level that just works.
> LSD and engineering judgement uncertainty are independent and additive. LSD is clear and prescriptive., what is less intuitive is that uncertainty found in engineering judgement uncertainty is varied, riddled with bias and results in a less sustainable build.
> What about lowest standards in accordance with the codes. Regenerative design could encompass this.
>I do this on every project. It is my signature architectural piece. This is to say that the client has no say and the design is not driven by the client.
> I would say without reservation that there is no common success parameters that exist in every project. And that we are continually surprise (and sometimes dumbfounded) by what success means to some people. Costs and value for money are common but are often irrelevant. Livability is often pushed aside for amazement and awe. Satisfaction can mean a myriad of different things. It takes intense probing to know what people are looking for. And I think the answer is as diverse as there are people.
How does buyin affect the level of uncertainty and how it is perceived. Reduce risk, want to make it work, complete control and
What about people that need to stick to something they previously agreed to but now know is wrong?
This would be great if this value perfectly correlated to the answer,that is, if there were a correlation of 1 between the two factors. That is obviously not the case.
Uncertainty and Risk Reduction in Engineering Design Embodiment ProcessesDesign with uncertain qualitative variables under imperfect knowledge - Pons, RaineUncertainty in Structural Engineering - William M. Bulleit, M.ASCEThe human factor in structural engineeringResourcesIf you get bogged down in being perfect you will never succeed.