|
In last month's (February 2013) newsletter, I explained what Ideal Functions are and their importance to developing a high-integrity foundation as a design concept is converted into a base-line design that works under nominal conditions. Part 2, Design for Additivity, is the subject of this month's newsletter.
I once did some consulting work in Germany where a development team was trying unsuccessfully to convert a subsystem concept into a working prototype. The problem was clear - the prototype did not have the ability to work. They tried and tried to get it to function properly using the finest materials, machining (and the Germans are very good at machining things!) and analysis. They just could not get it to work. The problem was they had only one concept from a very strong-willed technical leader and it was his way or the highway! Furthermore, this single concept was fatal - no matter how much modeling, simulation and build-test-fix iteration they cycled through, this architecture was doomed. The only thing they could do in this case was go back to the concept development phase and come up with new ways to physically execute the required functions. The Pugh Process was designed to look at many concepts concurrently and hybridize one from many. The Pugh Process will significantly increase the probability of emerging with a truly superior concept that will be "develop-able". One problem I often see is that concepts have poorly defined Ideal Functions. Teams bypass this essential characterization that will show the mechanisms that underwrite the functional integrity of the design options they are proposing. The lesson here: start with many concepts with well-defined Ideal Functions - hybridize the best elements into a superior architecture.
Once you are in possession of a superior concept underwritten by a well-defined Ideal Function, you need to look at the building blocks that are available to control the function(s) as you consider the design's bulk materials, dimensions, surface finishes and factors of shape. I call these "basic factors", and they are the first step in developing engineered control parameters. A basic factor can be different than an engineered control parameter.
By way of example, let's say we want to design a set of rollers whose Ideal Functions are to "Transfer Heat", "Apply Pressure", "Melt Material", and "Adhere Material" such that a material is adhered to fabric without generating any wrinkles. In layman's terms this is like ironing - heat up two materials and push down on them at the same time for a controlled amount of time across an interface area where energy and mass can be transformed. There are numerous basic factors to use during the development of these Ideal Functions: length, inner and outer diameters of the rollers, Young's modulus and Poisson's ratio of the rollers' materials, the heat transfer coefficients of the rollers, surface finishes and profilometry of the rollers, on and on the list goes. The question is, how do these basic factors go together to enable the stability and adjustability of the Ideal Functions? Should we put them all into a Designed Experiment, run vast numbers of experiments, and see what happens by counting the number of unwanted wrinkles in the base fabric? Obviously, the answer is no, and there is a better way to approach this by observing the following rules for Design for Additivity.
The Rules for Design for Additivity:
- Assess the sub-level functions that must occur for all the Ideal Functions to happen correctly - this is like the "domino-effect". It's tedious, but absolutely necessary to define all the physics-based mechanisms of energy and mass dynamics that have to happen in serial-parallel flows as events build up to fulfill the high level Ideal Functions.
- Generate a list of basic factors that are available during the design of a functional sub-level design.
- Identify the way Conservative Laws of Mass and Energy apply to the specific transformations your sub-level basic factors must perform. This is where Additivity is identified because all the mechanisms that make things happen by design follow additive physical laws. As engineers, you know these as equations that balance across a boundary such as Bernoulli's Equation, Lavoisier's Equation, Fourier's Equation, and so on. These are engineered relationships that are integrated factors that show us how to group design elements into "intelligent" parameters that truly control stability and adjustability within the design. This is called "parameterization".
- Identify Dimensionless Units and other forms of "engineered" or lumped factors that form an engineered control parameter (think Reynolds Number, etc.).
- List the mix of engineered control parameters and basic factors that can be selected to explore how the Ideal Function is truly controlled. You will always have a mix of both.
- Make sure you are measuring and controlling Xs (the engineered control parameters and basic factors) and measured physics-based, continuous variable output responses (Ys). This means measuring the physics - do not count quality defects!
In our example, two things illustrate the point:
- Design basic factors in the rollers to control and change the structural stiffness of the rollers. That is an engineered control parameter needed to control several of the Ideal Functions. The basic factors available to define the rollers must be grouped into stiffness levels. "Changing stiffness" is the real engineered control parameter - not changing the basic factors independent of one another! Stiffness has an additive affect in relation to other basic factors and engineered control parameters that govern the physics behind stable and adjustable Ideal Functions.
- You must measure the physics of impending wrinkles; do not count inches of wrinkles in the fabric! This means measuring the change in the position or internal energy of the fabric as it is pulled and pushed, sheared and bent as the mechanisms of rolling, pressing, heat transfer, boiling and material melting dynamics are accomplished. Taguchi expressed this as, "Measure impending failure! Do not count failures; they have low information content and will deceive you into thinking you know what you're doing when the failures happen to reduce in number or go away for the time being."
These are some insightful tips to help you get back to designing engineered control parameters. Yes, sometimes you have to use basic factors along with the engineered control parameters to learn about the Ideal Function of a design. That is your job as engineers, and it is why you studied all the physics and engineering principles in engineering school. Taguchi asked me one day, "Why don't you use physics? Why do you count defects and focus on quality attributes and time-based reliability?" I replied that I was thoroughly trained in quality metrics as a policy at Kodak upon entering the company. I was taught to focus on quality. He quickly put an end to that way of thinking and soon after, the quality of my designs went up dramatically!
Additivity is all around us. Everything follows the conservative laws of mass and energy - everything. Yes, there are very real interactions between basic factors and true physics-based responses, especially in chemical and material systems. They must be identified and included in our models of Ideal Functions.
Is there such a thing as an artificially-induced interaction between basic factors and engineered control parameters? Indeed there is! You can easily make them happen by running a designed experiment using basic factors only, with no thought to designing engineered control parameters, and then count defects as your Y variable - you will see artificial, non-physics-based interactivity between your Xs and, no, you will not be able to explain them. That is because you mixed apples with oranges; you changed physics-based Xs that were not directly related to a physics-based continuous variable Y that is the true expression of the Ideal Function.
Additivity is an engineering tool we must use to help define our Ideal Functions. These concepts are rooted in the Laws of Conservation of Mass and Energy. As Taguchi advised me all those years ago, "Follow the energy, follow the mass, and you will understand the Ideal Function - then you are prepared to move on to robust design."
|
