UNCERTAINTY IN ENGINEERING DESIGN

摘要

Design is creation of new things, an activity that is crucial to life. Engineering Design is a subset of these design activities, which focuses on technical aspects of performance of the target system of the design process. Thus, Engineering Design is concerned with the creation of structures and systems and, in addition, with the analysis, the refinement and testing of those concepts. We use Engineering Design in order to cover both the Engineering Analysis and the Engineering Design Synthesis view of the whole process and we focus on formal, that is computable methods. This paper discusses whether or not uncertainty is involved in the overall engineering design process defined above, and what types of uncertainty theories fit best the engineering design process. Considering causes of uncertainty, it appears that lack of information, abundance of information and ambiguity are the most important causes for uncertainty in Engineering Design. The process of Engineering Design Synthesis starts (early phases) with lack of information. Customer preferences and technical specifications are only partly available. The available information is mainly linguistic and if at all only partly set- or interval-valued or even numerical. The design synthesis might be considered to be a process where incomplete or imprecise information is stepwise replaced by more precise information, in other words, it is a process which reduces uncertainty and terminates with a precise system (target system, design artifact, etc.). Concluding from a much more detailed discussion we will find that those uncertainty theories are most relevant for Engineering Design which deal with a mixture of numerical, set- or interval-valued and linguistic input and output information as well. Therefore, Fuzzy Set Theory, Possibility Theory and combinations of Fuzzy Set Theory with Probability Theories seem to be best adapted to the needs of dealing explicitly with uncertainty in Engineering Design. The use of Probability Theories is in particular useful when taking into account uncertainties caused by imprecise measurement and imperfect manufacturing or when imitating the evolutionary process from the nature in order to generate populations of design alternatives. In order to show the use of uncertainty theories in Engineering Design we consider important methods, e.g., the Method of Imprecision (MoI) and extensions and Genetic Algorithms with a fuzzy fitness evaluation in more detail and illustrate them by real world design examples.