If you haven’t learned to fear adhesive bonds, you haven’t lived a complete medical device life. Adhesives are truly marvels of transmutation: liquids stay liquid until they magically become solid, and a drop or two of base substance can hold dissimilar materials together with superhuman strength.
Yet control of adhesive processes is always a nightmare. UV fluence or position changes from lamp-to-lamp, and oven temperature varies seasonally. The environment is always too damp or too dry. Dispenser accuracy varies. Somehow the location of your adhesive on today’s device has shifted slightly from last year’s location. With adhesives, you just never know which variable is going to cross the line from in-control to out-of-control. You don’t need a masters in statistics to see that a large number of low-probability process failures adds up to a higher-than-desirable probability of bond failure.
I routinely bore people with my assertion that everyone should be required to study and master statistics in high school. We all need statistics to better understand the world we live in and the news we read. Without statistics literacy, we can easily be misled. In our personal lives, we make financial investments, buy insurance, and make decisions with risks. At work, engineers and scientists need statistics to understand designs, processes and experiments. Sales and marketing people need statistics to understand market attractiveness and sales probabilities. Supply chain and operations experts need statistics to understand forecasts, materials plans, and manufacturing processes. Even accountants and finance types need statistics to understand currency risks, stock options, and financial instruments.
Star medical device engineers master statistics to make better designs in less time. How?
To paraphrase my friend Jim, there are at least three ways.
How will tolerances add up in the real world?
In CAD, everything fits perfectly. In the real world, medical device components are manufactured with tolerances. Dimensions of machined, molded and otherwise manufactured parts always vary slightly from their nominal specifications. These real-world variances add statistically. It’s unlikely that every machined part in a product will be received on the high end of tolerance. Star medical device engineers stack up tolerances statistically to design products with margin. I know a company that launched a product based on a laser that couldn’t be made – the tolerance stack-up was just too big. Surely it would have been better to find the issue in the design stage, rather than the pilot manufacturing run.
What does the sample say about the population?
Most engineers feel satisfied with a design when they take a handful of prototypes to the lab and all units meet some minimum specification. Star engineers, on the other hand, worry about the variability of results, even when all prototypes pass muster. Star engineers see their prototypes as a statistical sample of a larger population. They want to know if the results indicate that a sufficient percentage of that larger population of devices will meet specifications. Star engineers design quantitative tests to provide more insight than attribute tests. They assess data for normality, and apply t-tests to means and f-tests to variations.
Viewing prototypes as ‘population samples’ drives good design practices. To reduce unwanted experimental variance, star engineers create detailed assembly instructions even for the earliest builds. Knowing exactly what has changed in each design iteration then enables pooling of data across design iterations. enabling further insight into design performance. Later, star engineers confidently use historical data to power verification studies.
How can we explore multi-factor design spaces?
Every medical device engineer, more or less, knows how to design and interpret an experiment where a single factor is varied, while all other parameters are controlled. These experimental designs are drilled into our heads in high school and college. Yet variables interact – machining a material might produce different results than molding the same material. Imagine a first experiment to select machining versus molding, and a second experiment to select an optimal material. If machining is selected, you might never identify the molded material that works better than the best machined material. Most medical device engineers put a lot of thought into prioritizing a series of single-factor experiments to avoid interaction errors.
Star medical device engineers use Design of Experiments (DOE) to produce better designs with reduced experimental sizes and timelines. With DOE, a pre-determined set of factors is varied systematically and simultaneously in a single experiment, and analysis of variance (ANOVA) is used to understand the influence of each factor.
In principle, you don’t have to be an expert in statistical methods to think statistically. Star medical device engineers tend to become experts in statistical methods because the tools and techniques are so useful. Take a look at an engineering team you know, and chances are that the star engineer is the one with the Minitab license.
- Star Medical Device Engineer – Attitude (jaycaplan.com)
- Star Medical Device Engineer – Specification and Test Development (jaycaplan.com)