Design Robustly

Robust design is one of the most creative and rewarding design endeavors for a DFM engineer. Here, we challenge ourselves in how creatively and elegantly we can design out a manufacturability problem.

The Six Sigma elephant in the room

Six sigma is a loaded subject matter, thus in true DFM fashion, let's simplify this subject matter by focusing on the basics. Most engineering problems focus on inputs and outputs, where inputs drive predictable outputs. Whether by theoretical means or through empirical and statistical observation, a predictable correlation or equation can be made between inputs and outputs. Six sigma efforts involve studying how inputs affect outputs through experimentation and statistical tools.

Any number of engineering inputs can be identified for an engineering system. These inputs are commonly organized into 6 categories known as the 6Ms: Method, Material, Machine, Measurement, Man, Mother Nature. All of these categories are legitimate engineering inputs, however design seemingly makes for a small portion of these inputs.

💡

Now let's talk about the elephant in the room... Design!!!

Design arguably fits into a sub-category of "Material." Moreover, when industry utilizes 6M brainstorming, all to often the design is never even questioned. Instead, the design seems to become an input that is off topic for discussion. Why? Because it may be difficult, impractical or impossible to change given where the product is in it's lifecycle phase. As a result, bad design becomes an almost taboo topic. People instead, all too often turn to the scape goat input that is "Man," 🫥 and solve problems by "retraining the operator." More ambitious engineers will charge down the rabbit hole of processing parameters. True, there is unquestionable value in coming out the other end of that rabbit hole with a robust set of processing parameters, but this is not DFM!

What is Robust Design?

Robust design follows the same six sigma methodologies, exploring inputs as a relationship to outputs, however robust design puts full emphasis and priority on design first. The more you can eliminate defects and widen acceptance limits by design, the higher the chance of a rock solid product. Products that are designed robustly do not need tightly controlled process parameters nor skilled operators.

💡

Robust design is like 6 sigma that only focuses on design inputs.

Why focus on design first?

Here are 2 notable reasons: Because a low quality design results in more time wasted and dollars spent for everyone. Also, because it's extremely harder to change the design later in the product lifecycle phase.

How to implement Robust Design?

Implementing robust design is more challenging than most six sigma projects. These projects often find themselves against the flow of corporate inertia. A typical six sigma problem will often focus on a volume production issue tied to a high cost savings opportunity. A new design has yet to produce any real production numbers to justify cost savings, let alone any production maturity to even identify any quality issues. So how do we approach robust design?

New product teams must take extra effort to tackle robust design at the onset of a project. First, we must know where the quality issues can arise. Finding a close predicate design with production data can be very valuable. Drivers of DFM must study this data thoroughly to identify quality issues that will likely remain on the new design. FMEAs and CTQ exercises may slightly help, but are an extremely poor substitute for real data and are better served as communication frameworks.

Ultimately, vetting a new design entails building product early and often. Prototype builds are arguably the most important element of Robust Design, and the only true way to vet a new design. These build efforts must be pushed by passionate and driven engineering teams, eager to test all that can go wrong. After building, teams are all to often surprised to find what can go wrong. The steps for iterating to a rock solid design are quite simple: Prototype, Analyze, Repeat. We explore more on each as follows:

1. Prototype

First, thou shall prototype! Don't overthink this step as its okay to bootstrap. Prototype builds do not need to be in large quantities, nor do they have to be made from finished parts. 3D printed prototype parts or other hacked together manufacturing methods are better than nothing. Capturing qualitive and quantitative data is sacred for each prototype build. For every build, teams must at least track both yield data and cross functional team feedback. Yield data does not even need to be based on a specification (if one does not exist) and can instead be based on good engineering judgment. Feedback absolutely must include anything from operators and technicians, whose feedback is invaluable. These notes do not need to be formal, and in fact are often better done quick and dirty. Time is crucial at this stage, so capture data efficiently and move on to iterating better designs!

2. Analyze

Next, analyze where top yield losses are occurring and then target focused brainstorming power on how these issues can be eliminated by design. Can we simplify the design to eliminate the problem? Can we fine tune design properties to robustify? There are some very deep six sigma statistical tools that can help, however usage of these tools are honestly often overkill. Much of this brainstorming relies on the infinite creativity of the human brain to design something conceptually better.

💡

Definition of Robusify (verb): To make something robust by the shear awesomeness of engineering brainpower. Example: Can we please robustify this crappy design? Merriam-Webster, please add.

3. Repeat

Finally, we put those ideas into action by prototyping, building again, and seeing if we improved! Voilà, the iteration loop that is DFM. Want to really maximize your learnings? Try creating several design concepts in parallel to multiply learnings with each iteration. Repeat these steps to learn as much as possible before time runs out and the design is frozen.

💡

DFM challenge: How fast can we iterate? How many iterations can we squeeze in before the design is frozen?