Estimate First: Engineering Intuition That Prevents Mistakes

The Missing Skill: Estimating Before You Calculate

Our engineer ran a simple FEA simulation on a bracket and reported that the part would deflect 327 mm under a 200 N load. It took four hours of refining the solid model mesh, updating boundary conditions, and troubleshooting for slow and unreliable numerical solution convergence. Just a few minutes later, a colleague walked by and, glancing at the simulation results on the computer screen, innocently asked, "Does 327 millimeters seem right for that bracket?" The silence that followed was deafening. Never once did it occur to the engineer to wonder whether the answer was correct. It turns out that he had accidentally entered 200,000 N, not 200 N load. All of that extra work for nothing.

While I was in school I thought I was learning how to be an "engineering" skillful engineer. I found out after I graduated that the school really taught me how not to make silly math mistakes. These errors don't occur because of mathematical flaws in the calculations. More often, errors are caused by such simple mistakes as transposing a decimal point, misinterpreting units, entering an incorrect value, selecting the wrong input or load case. Under ideal circumstances, the calculator can then crank out an accurate but incorrect answer. To detect such errors, you have to have some sense for estimating answers before you calculate them.

Estimation is NOT guessing. A 1-meter steel beam for light duty construction is on the order of millimeters not centimeters or microns of deflection under load. A small electric motor has power on the order of kWs not MWs or even 100s of kW. A seasoned professional has a strong sense of what is an order of magnitude out of whack. This is the kind of intuition they develop by first estimating and then calculating to compare. Our students do the opposite. They calculate away fervently accepting whatever the computer gives them as the "right answer".

Order-of-Magnitude Thinking

What weight will a bracket hold when it is used to hold heavy equipment of 50 kg? What does an already made design look like where all the holes are threaded for M6 bolts? Should we calculate for its capacity or is it obviously under capacity? I've never actually calculated the maximum load for anything, I just go by my general impression of what something can hold based on past experiences. This is what happens when you don't build things to fail occasionally — you start to build up a subconcious library of "things that are roughly this strong".

So, for any engineering design or analysis, you don't need to get the numbers precisely right. Any rough estimate that is in the right ball park is good enough. I mean, is this thing going to deflect by something on the order of 1 mm or 10 mm? Is that motor going to need to be 1 kW or 10 kW? Is that part going to weigh 5 kg or 50 kg? You get a factor of two or three wrong, and it's probably still okay for making go/no-go decisions to avoid catastrophe.

Learn to fill your mind with useful numbers stored in your memory for later use. The torque holding a well made M8 bolt in steel is on the order of 20 kN. A person has a weight on the order of 700 N, 100 kg say. The power required to lift 100 kg to 1 m/s is approximately 1 kW neglecting losses. The ultimate tensile stress for steel is on the order of 250 MPa, for aluminum on the order of 200 MPa. Knowing the power of my car engine (100 kW) and my household appliances (1 kW) is one of the things that saves me from making very stupid mistakes.

Engineering Intuition Is Pattern Recognition

Most Junior engineers think that Experienced engineers have some sort of "sixth sense" when it comes to design. What they don't realise is that Experienced engineers have seen enough failures, run enough simulations, and reverse engineered enough products that they have built a vast database of "things that work" and "things that don't". As they review your design they are trying to match it against known patterns in that database.

You build intuition the same way: estimate, compare to reality, and when you were off, explain why. Was it because you forgot that bending effects vastly overpower axial effects? Because you underestimated the impact of length scale? Because you kept confusing stress and force? Each mistake provides another opportunity to refine your mental models.

Repetition is the key. Estimate everything: how much torque will I need to turn that valve, how thick will that plate need to be, how much will that assembly weigh. And yes, you will be wrong, very wrong, but that's how you learn what really matters and what can be safely ignored.

When to Question a Calculation or Simulation

One of the most dangerous moments in the life of an engineering solution is when a simulation is run, or an equation is solved and a number produced. There is a momentary pause while the result is accepted, but no pause while the result is assessed for realism. The answer is so precious that it is accepted without much scrutiny and soon finds its way into a report or manufacturing documentation. Should the answer be 10× too large or too small it may not be discovered until something fails in use or doesn't fit together properly.

Having a smell for things not being right is a valuable thing for anyone to have, and especially for senior engineers, who should spot walls that are too thin in drawing and suspect 400 Nm torques instead of 500 Nm. Typically they don't do it by calculation but by pattern matching with other examples they have encountered previously.

Red flags that should make you stop and recheck:

→ Your FEA results show a stress level of 245 MPa with first yield occurring at 250 MPa. You are running at the limit! Room for optimism in the load estimates or room for a bigger safety factor?

→ Deflection is way to big. A 100 mm part comes out with a deflection of 200 mm. Which means the beam is bending double its size. Either your model is broken or you need a completely different design.

→ Your specified tolerances of 0.01 mm are more appropriate for parts that will be machined on a CNC milling machine, such as our CNC mill. Typically, regular machining has a tolerance of ±0.1 mm. Grinding can tighten this up to ±0.01 mm, but if you don't need grinding, you don't need the tighter tolerance.

→ Numbers in the book seem too rounded off. 1000. 50 mm. 100°C. Real-world numbers don't come out so neatly unless the author forgot a crucial conversion factor or made an unrealistic assumption for the sake of simplification.

→ Hand calculation doesn't agree with FEA result by 100×. Probably one of them is wrong. Don't just rely on software blindely. Know where you are going wrong.

Intuition is important for recognizing good results, and if you don't have it you'll ship lousy designs. Making it a habit to check if your results make sense as you get them will make it a habit to help you recognize good results.

These chapters were written in the hope that someone reading them would say: "That is just the kind of practices I need to start doing in order to become better at estimation!"

How to Build Estimation Skill in Practice

Estimation skills are not learned in a book, you need to actually go out there and estimate, be wrong, figure out why and adjust your mental models accordingly. It is somewhat painful at first, since you will be wrong most of the time. But that is the process of learning.

A rule that I try to follow at the beginning of most design tasks is to make an initial estimate of key parameters before I begin to compute anything using any tool - this could be a calculator, CAD package, FEA tool - whatever fits the task. Then, I calculate, and then I compare my calculated answer to my pre-computation estimate, and when I am far off from my initial guess, I attempt to understand what I can do better in the future.

Learn to reverse-engineer everything around you. Consider the bracket of a bookshelf. How thick is the steel rod? What load is the bracket holding up? What stress is in the bar? Make an initial guess, look up the real dimensions of the bracket, and then calculate the load, stress, and required power to reverse-engineer the design. The same goes for a crane lifting something. How much power does the motor need to lift the load? Make an initial guess, and then look up the real specifications of the crane to calculate the answer. Real objects are practice problems waiting to be solved.

I strongly encourage anyone to build up a reference library of notes and numbers that have been calculated in the past for repeated use. Things such as typical bolt preloads, motor power, prices for various materials per kilogram, and typical tolerances for various manufacturing processes get used over time, and it's much easier to reference something you have previously done, rather than having to look something up from scratch.

Do a lot of "what if" analysis. What if the load increased by 2x? What if you swapped from a steel rod to an aluminum rod? What if you doubled the length of the rod? Get really good really fast at making these simple calculations and figuring out what actually happens as opposed to what you expect to happen.