Physical experiments in industrial product development are often very expensive but also necessary sometimes even for safety reasons (think about the crash tests in the automotive industry conducted with new and completely equipped vehicles). Therefore there is a pressing need to reduce this cost as much as possible. Computer experiments may be a viable solution provided that their results are reliable as much as the physical ones. In most of the cases, computer experiments are opportunely combined with physical experiments and the recipe of such mix is interesting itself.In this presentation we will show two cases of combination of physical and computer experiments in product development. The first case concerns a problem of “robust calibration” of On-Board Diagnostic system for a new car model developed at FIAT research center. In this case thanks to an algorithm of the diagnostic system working offline (i.e. not necessarily installed in the car) it was possible to drastically reduce the amount of physical experiments aimed at making robust the diagnostic system, that is as much insensitive as possible to the action of noise factors.The second case concerns a problem of assessment and optimization of ergonomic performances of a three-wheel city vehicle in the very early development phases (concept design). This was also a problem of robust design where a peculiarity was the definition of anthropometrical features of a potential driver as a noise factor. In this case the computer experiments were made thanks to a CAD prototype of the vehicle and virtual environment software. The experiments were themselves a special mix of computer and physical ones. In both cases the experimentation was successful in terms of amount of new knowledge acquired compared to the experimental effort. Furthermore, in scientific terms, the cases showed many novel interesting problems that were partially solved at the time and are still today matter of research.
|Numero di pagine||0|
|Stato di pubblicazione||Published - 2012|