Decision sciences-Bayesian statistics in engineering analysis and modeling

In engineering design, decisions often need to be made when complete knowledge is not available. Getting complete deterministic information may be impossible or the resources (time, money) required to get the information may be unaffordable. The problem then is to select the best information-gathering strategy by assessing the utility of the various information-gathering alternatives available.; Engineering modeling is an information-gathering strategy where predictive information regarding the performance of a design can be acquired virtually by means of mathematical models. The selection of an appropriate model based on the preferences of the designer is critical to the success of a design. Accordingly, this paper reviews previous work on Preference Based Model selection and also develops a Methodology based on Bayesian Statistics to incorporate the uncertainty inherent in engineering analysis models in the context of an iterative decision-based design process.; Specifically, this work enables generation of the required probabilities for the decision tree and the selection of the best analysis model from a maximum expected multi-attribute utility perspective. In this methodology, concepts from Bayesian statistics, utility theory, maximum entropy principle, simulation modeling and statistical design of experiments are integrated towards this purpose.; The choice of an appropriate modeling strategy is made considering the uncertainty of the results and the utility of the design outcomes that the uncertainty might lead to.; This work suggests an additive multi-attribute DECISION TREE . Multiplicative models can also be developed similarly based on user preferences.