Metrics for Improving Engineering Design Education

Rudy Eggert
Boise State University

Can Engineering Design Education Be Improved?

Of course it can. We have read dozens of articles on a variety of tactics taken by different engineering educators telling us so. We have read about the senior design projects with industry, the professional society design competitions, and freshmen teams building straw-towers. Right? Haven't engineering educators proven how revisions to their courses and curricula improved graduating engineers' design knowledge, skills and attitudes? In a few rare, and well-documented cases...yes. Usually, the end of term course evaluations are the only "proof" offered. While anecdotal evidence for improving engineering design may be useful, its lack of structured metrics limits its usefulness. Metrics, or in other words....data, would help us develop a solid body of knowledge necessary to effectively plan and execute appropriate pedagogical tactics & strategies. Also, metrics could help us improve engineering design education while making optimum use of available resources.

Decision-Based Design Metrics?

What is the goal of engineering design education? Let us try to answer this by backward chaining from an assumed goal of graduating engineers that have demonstrable design knowledge, design process skills and appropriate design attitudes. The controllable design variables we have as educators, to achieve our learning objectives, include lectures, homework problems, projects, papers, plant tours and lab assignments. But, how much of each and when should each be used to achieve our goal? Many of us have orchestrated many educational symphonies, in which we distribute design throughout our curricula in various rhythms and melodies. Sometime harmoniously, often discordantly. Still, can't we be more efficient and effective in re-engineering design across the curriculum? Yes, but we need data to do it! How does decision making fit into the design process? Let us assume that the design process largely consists of decision making processes used to decide the "form" of a product (or process) from a desired "function." In mechanical engineering, for example, we would decide shape, configuration, size, materials and manufacturing methods to meet a desired purpose or function. How are these decisions made? Occasionally trial and error can be used as the decision-making process. Sometimes they employ heuristics, or rules of thumb, and in others, more formal methods can be used such as: numerical optimization techniques,

Design for Manufacture, Quality Function Deployment, Synectics, Taguchi, Pugh, Pahl & Beitz, and DOMINIC.

Characterizing Design Content?

Can design content be measured or characterized? Can we look at a homework problem or other type of design assignment and say yes it has design content, and specifically what aspects of design it does have? I believe we can. Permit me to propose the following Design Content Vector DCV, of measurable attributes to characterize design problems as: DCV = {F, A, E, R, C} Design content, as proposed above, is the multidimensional set of attributes represented by the number of decisions made in each of the design process categories including: design problem formulation (F), alternative generation (A), evaluation of alternatives (E), refinement of alternatives (R), and communication of design (C). Formulation decisions include: determining the "real" problem, finding relevant information, specifying functional requirements / functional decomposition, decomposing function(s) into sub-function(s), defining design evaluation criteria, defining design variables and problem definition parameters, determining physical, safety, & economic constraints, planning the solution strategy, and preparing engineering design specifications. Alternative generation decisions include: examining alternative, usable physical principles, synthesizing alternative shapes, configurations, materials, estimating/guessing starting values for design variables, employing creative techniques such as brainstorming & synectics, and employing information resources (corporate files, libraries, experts). Evaluation decisions include: modeling (mathematical, computer, experimental), using effective analysis methods / tools, testing of product / prototype, obtaining feasible solutions, evaluating analysis results, and selecting optimal solutions. Refinement decisions include: detailing design, tolerancing, and automating iterative redesign. Communication decisions include: design specifications, technical reports, and memos, oral reports, conversations, meetings, sketching, drawing, diagrams, and charts. Also, a design content matrix could be defined by expanding each of the design content vector elements, {F, A, E, R, C}, into 3-dimensional vectors for {knowledge, skills, attitudes}. Another variation could follow Dixon's design problem taxonomy: functional, phenomenological, embodiment, configuration, and parametric or detailed. Of course, the design content vector (or matrix) will need to be developed, tested and refined. Professors Tennyson, Ahmed-Zaid, Haws, also at BSU, and I began research into this topic this past summer.

Pedagogical Uses for the Design Content Vector?

How could the design content vector be used? Design knowledge, design process skills and attitudes toward design can and should be developed over the four-year curriculum, not just in the capstone design course, or the freshman introduction to engineering course. The DCV should be able to measure the number and type of design decisions considered by each student for any given assignment. As such, we could establish specific design-related learning objectives and then assess whether we attained our objectives. For example, in a freshman activity oriented to the design process, we would hope to see a balanced DCV, i.e., all (F, A, E, R, and C) elements of roughly equal magnitude. If we aimed a sophomore activity at creativity enhancement, we would hope to see the A element be larger than the others. The DCV would also be used to establish control groups for testing various learning tactics & strategies and to standardize anecdotal evidence. Also, by comparing the magnitudes, of each of the DCV elements, we could begin to evaluate breadth versus depth issues based on objective data.

Summary

The development of a design content vector to measure the type and quantity of design-related decisions will not be a simple task. Nevertheless, we must start efforts. With metrics we will have data. With data will have evidence. With evidence we will be able to develop effective and efficient pedagogical tactics and strategies in an era that is demanding that we do more with less.