In this interview, I talk with Ray Rilling about applying mathematics to manufacturing medical plastics.
JC: Ray, could you start by saying a little about yourself?
RR: Sure. My name is Ray Rilling, and I am the Director of Technology at Putnam Plastics.
My initial training was cellular biology with an emphasis on epidemiology, but I decided to move to a more applied (and lucrative) field. I’ve been in medical plastics for the last 23 years.
When I saw that epidemiology was a crowded field, I pursued substantial amounts of additional coursework that allowed me to get into the medical engineering plastics space. In lots of ways I think this pick-and-choose approach was much better than a traditional bio-medical engineering degree would have been. It allowed me to focus on exactly what I needed to learn.
JC: Thanks. And tell me a bit about where you work.
RR: Putnam Plastics is a 30 year old medical plastics manufacturer, specializing in polymer-based medical components (i.e. plastic tubing for medical usage).
JC: I did some consulting work for a company that manufactures stents, so I imagine your companies have a few things in common. What is your day-to-day mathematical toolkit?
RR: We cross a lot of boundaries and there are many layers of computational needs. The majority of the computations are basic algebraic formulas to calculate simple things like annular tooling draw down ratios or the ultimate tensile of a tube. From there, you may find that calculating the burst pressure of a composite or may require the solving of a linear system.
The most challenging computations that we perform are for our complex polymer extrusion tools. These are based on Computational Fluid Dynamics (CFD) and Finite or Boundary Element Methods (FEM / BEM). Many complexities emerge because polymer flow is compressible, non-isothermal, and viscoelastic in nature. It does not allow us to use conventional Navier-Stokes or other common equations. In all reality, we rely on proprietary equations to point us in the right direction. No model has ever provided a perfect solution. We combine our experience and expertise with the applied calculations.
The most complex of the computations are performed with third party software packages and often require the use of complex meshing, materials testing for inputs, and can even require us to create modified code to suit the specific application. We work with some very creative minds and there is never a shortage of new challenges.
Aside from the design engineering mathematics, we think statistically about our results. This can be to analyze the process capability, equivalency, and even product risk analysis. This statistical analysis can be intertwined with the design engineering aspects and deal with a range of biological and medical constants (i.e. devices might need to handle liquid nitrogen or remain in bodies for years or decades).
One of my mentors used to say, “The best equation is teamwork.” Teamwork is a sort of expectation, a crowd-sourced form of expert opinion. It allows you bring a focus on what the real variables are.
Some calculations can take weeks, so having a good appreciation of how to approach the challenge is important. You can get away with not knowing what’s under the hood. But that’s dangerous. It is much better to get things done more quickly, and with better understanding. Especially when a client is waiting on results.
JC: What is some math you wish you learned more of?
Vector and tensor-based physics. As the bar in the medical device industry increases, so do the expectations of the products that we make. Being able take large number of variables or inputs and transform them into a workable model is extremely valuable. My dream is to be able to reliably calculate extrusion tooling, product failure modes, and performance properties before I start cutting steel. But in our time and age, we still rely on some development iterations and calculate what we can.
I also wish I had more applied materials science. When I am developing something in the lab, sometimes the product does not perform the way you want it to. Everytime I start learning something new, like applied surface chemistry or the effects of radiation on polymers, I think of 100 things that I could have done in previous projects.
JC: Did you learn any math you haven’t had much use for yet? I’ve used most of the math I learned, though sometimes there’s been a gap of twenty years between learning something and applying it.
RR: I actually use pretty much all of the math I’ve learned. But that’s likely because I was able to pick and choose because I went back and did supplemental studies. Even the epidemiology is useful.
JC: I wanted to pick up on what you said earlier about physical models. Could you say more about the interplay of physical and mathematical models at Putnan?
RR: The models are never completely right. We use them as a tool. They often fail in three key ways:
- We find additional variables we need to account for or constrain. In many cases our first attempt failed because the product did not perform properly in an unaccounted for way. At the end of the day it might take many iterations before we have all of the performance properties that are needed.
- We are translating soft or subjective numbers from physicians or engineers. A model will provide concrete numbers. It is often important to create a baseline product that can take the customers subjective description and correlate it to a model.
- We need to additionally constrain for cost effectiveness (i.e. minimize the amount of expensive wire in a catheter).
Those three trade offs mean that we are never able to just take a model print out and run with it. There always has to be some back and forth.
JC: It’s interesting that you consider shortcomings besides inaccuracy as failures. Someone new to applying math in the real world might think of your first point but the other points. The latter two aren’t failures from an academic perspective, but they certainly are from a business perspective.
* * *