This may seem like a blasphemous post. Oh well; the point of this blog is to cause trouble.
At Seven Cycles we sometimes hear the question, “Do you use Finite Element Analysis—or what type of FEA do you use—in your design process.” It’s a very reasonable question. It’s an engineer type question. FEA is an engineering tool. It is a computer program that simulates the way a product might be used; in the case of bicycles: loads, stresses, impacts, etc.
‘Finite’ is the key word. It also stands for finite value, in my view.
I preface this conversation by explaining that Seven Cycles is lucky enough to have six engineers on staff. For a business our size, that’s way overstaffed—it’s about 15% of our team. Granted, they are not working on engineering and design projects fulltime. However, they do all participate in design projects every day. In addition, their primary roles at Seven include engineering on a daily basis. And, lastly, while I don’t have an engineering degree, I am part of the engineering & design team.
We have a hierarchy of design tools for making products for which we offer lifetime warranties. In this case I am only talking about tools that impact safety and durability; I’m not talking about tools that serve all the other elements of product development and design.
I will post about details of each of these tools at a later date. For the moment, I’ll provide only a brief explanation of how and why we use and think about each tool. Of course, nothing at Seven is a true hierarchy, but for the purposes of writing a reasonably short post, here is how our tools fall in order or value.
Experience
This may not seem like a tool, but the reality is that experience is the single most important tool we have. Without experience, no design can be optimized. Fortunately, I’ve been intimately involved in the manufacture of nearly 50,000 bike frames during the past 20-years and I’ve made just about every mistake possible. In addition, six engineers are part of our experience.
Real World Testing
Translation: riding. This is the best way to learn about what products work properly and which don’t. The only challenge here is that, as you can imagine, it takes way too long to collect data—particularly when you are designing products with lifetime warranties. Our race team relationships and our employee fanatics help here—for product testing acceleration—but it’s still not nearly fast enough. Real World Testing tells you if the product is way off base but it won’t tell you if you are 90% there.
Fatigue Testing
This is a great tool, and often times it’s what people imagine when we talk about testing. Seven has invested heavily in some very cool testing equipment—we designed ours in-house and co-manufactured it. I’ve also worked with third parties on fatigue testing over the past 20-years. The challenge with fatigue testing is that it never perfectly matches real world results. However, Seven has gone through a great deal of work to match real world outcomes.
Prototyping
I include this in the list of tools because it is often misunderstood. Making a prototype part on a CNC machine does not yield the same results as a production part that is cold forged. A prototyped molded carbon part will not likely behave the same as a production part. So, thinking about how a design is prototyped has an important impact on the results of testing that part.
Stress & Strain Gauging
Including some other measurement tools. This is more of a tool to confirm expectations. However, if experience is lacking, it can help as a surrogate to experience. Or, at least, it can help accelerate the experience curve if used carefully.
Finite Element
Analysis
Finally, we come to FEA. It’s pretty far down Seven’s hierarchy. For all the interest and press that FEA gets as a tool, you might expect it to show up higher on the list. It doesn’t; here are some reasons why.
Of course, I’m only talking about FEA specific to bicycle frame design, not any other application. Bike frames look really simple—a classic frame is ‘just’ nine tubes welded together—but is deceptively complex: welded thin wall tubes, single sided chain drive, they come in many sizes, have thousands of component variables, and see very diverse terrain use. And that’s just with a classic frame design. Add to that list: carbon in all its forms and 20-plus metal alloys in popular usage, and the FEA challenge is suddenly, and exponentially, more complex. So, the real challenge isn’t FEA; the challenge is the assumptions used to drive the outcomes of FEA.
I’ve done FEA bike projects many times over the past 20-years. Seven Cycles even did a large project last year in which we employed FEA. I can tell you that we didn’t make any critical decisions based entirely or primarily on FEA results. Most times the results of FEA do not match with real world results. So, my experience is that FEA is not a reliable source for critical product design—where safety and durability are concerns.
For example, I was once working on a project for a company—not Seven or Merlin—that requested a design parameter which didn’t make any sense; but they found their FEA required it. We had never seen a field failure even remotely similar to what the FEA showed—all of our experience said otherwise. In the end, the design they went with was very overbuilt in a way that was a waste of frame weight, cost, and complexity. The opposite could have easy occurred; one could be persuaded to under-design a part—in order to save precious grams—and end up with a failure simply to save weight or simplify a design.
At best, my experience with FEA is that it can be generally directionally correct. One of the primary values I’ve found in FEA is as a training tool—used for learning purposes. It can be a good way to accelerate and augment experience.
Until FEA software is able to perfectly model the requirements of the bike industry, it will be directionally helpful at best, and dangerous at worst. The industry’s current requirements that are not being met include:
- Modeling capability of carbon fiber and fabric—including perfect fabric layering, orientation, and resin relationships
- Modeling capability of thin wall welded structures that include the human welder element
- Perfection of single sided chain tension modeling
Current FEA is not there yet, but it will be. I’m guessing we are still about six or more years away from software that meets these requirements. And, of course, it has to be affordable to the bike industry; that might take a few more than six short years.
“It looks cool”
Last 'tool' on the list, just below FEA, is the dangerous tool of, ‘looking right’ or, ‘looking good’. Don’t worry, we don’t use this one. I’m including this one for a number of reasons. I’ll explain why in a future post.
Alright, now you know why this post may be considered blasphemous, let me know in what ways I’m crazy. And, stay tuned for follow-ups.
Comments