OPTIMIZATION IN THE AUTOMOTIVE INDUSTRY

How is optimization currently used in the automotive industry? What are the outstanding challenges? These are the questions that will be addressed in this talk. My goal is to provide the audience with a big-picture perspective and, hopefully, with some ideas they can take back to the office. Accordingly, the talk will not dwell on the details of any one application, though references will occasionally be given so that interested individuals can dig deeper.

Today, most successful applications of optimization in the automotive industry are at the detailed-design level. Examples include structural optimization of metal gauges and beam cross sections, topology optimization of cast parts, piston shape optimization, cutting stock problems, engine calibration, optimization of shock-absorber rates, assembly-line resequencing, and line balancing. Several of these examples will be reviewed. Persons unfamiliar with topology optimization will find this methodology especially fascinating, since it finds good topologies without explicitly defining parameters, objective functions, etc. Despite these successes, however, it is fair to say that optimization is used less often, and has had a smaller impact, than most researchers in the field of optimization might have expected. The reason is the presence of numerous barriers that need to be overcome.

What are the barriers? In engineering, a major barrier has been the poor interface between computer-aided design (CAD) and computer aided engineering (CAE). For example, much CAD work is still non-parametric and not geometrically clean, making it difficult to export data for CAE analysis and to iterate the process with new design parameters suggested by an optimization algorithm. Crashworthiness presents a challenge because crash is a notoriously non-repeatable phenomenon. To obtain a robust design, one must take into account the likely variation in angle of impact, driver position, material stiffness, etc. Yet data on the natural variation of these factors may be poor, and assessing how this variation will impact performance may require more computer runs than can easily be afforded. Managers may be reluctant to trust the results of optimization runs unless the math models are sufficiently accurate; hence, issues of model validation and calibration have come to the forefront. While the value of multidisciplinary optimization is well recognized, different disciplines (structures, durability, crashworthiness, etc.) often develop models at different points in the design process (because they require different levels of detail). As a result, care must be taken to insure that each discipline is working with the same version of the design. In operations research, applications of optimization have sometimes met resistance due to excessive data requirements.

While these barriers are substantial, much progress has been made. As I review this progress, I will point out implications for University curricula, for research, and for successfully implementing optimization within industry.