Formalizing negotiation in engineering design

Citation

Scott, Michael Joseph (1999) Formalizing negotiation in engineering design. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/thvr-5v61. https://resolver.caltech.edu/CaltechETD:etd-02262008-145407

Abstract

Negotiations are common in engineering design, especially on large projects, and are typically conducted informally. Often, negotiation is used to handle the imprecision or uncertainty that is inherent in the design process. Performance targets, initially specified as hard numerical constraints, are adjusted throughout the design process in negotiations between engineers and managers. Crucial unmeasured or unmeasurable aspects of performances, such as aesthetic concerns, are commonly negotiated. Negotiations settle conflicts between engineering groups over values of shared design variables and distribution of limited design resources. In this thesis, a formal description of negotiation in engineering design is presented. This formal model builds on earlier work at Caltech in the modelling of imprecision in engineering design. Negotiation is modelled mathematically as the aggregation of preferences. A complete characterization of the aggregation problem and of the aggregation operators suitable for engineering design is given. This class of operators spans a range of rational decisions and allows for different possible levels of compensation among goals. Furthermore, the entire range of aggregation operators is necessary to capture all possible engineering design decisions. Techniques are presented for determining which aggregation operators are appropriate for particular problems. As the aggregation of preference is also a component of other fields, notably decision theory applied to economics and social choice, various concerns raised in those fields about the legitimacy of preference aggregation are treated. A more comprehensive justification is presented here of the approach to modelling imprecision known as the Method of Imprecision, or MoI, than has previously been offered. Although the decision model presented here is statically one of choice among given alternatives, refinement and redesign are crucial in the engineering design process. The consideration of information about entire sets of designs not only accrues computational benefits, but is a more natural model for how designers reason, and can be of significantly more use to designers in refinement and redesign, than information about individual point designs. Conditions under which the negotiation model can support set-based information and still yield consistent answers are here explored and presented. Finally, an example of the application of these ideas to a preliminary vehicle structure design is presented. This example was undertaken as a demonstration of the method for research engineers at Volkswagen Wolfsburg, and serves to help introduce many of the ideas in a more concrete manner.

Thesis Files