Improved bayesian network model for supplier selection of complicated product development

Complicated product development (CPD) is a collaborative development process participated in by enterprise and its many suppliers together. Supplier evaluation and selection, and the coordination among suppliers and enterprise are directly related to the success or failure of CPD and the performance of complicated product. A supplier selection indicator system of CPD is established, and then construct a Bayesian network model which is comprised of criteria layer, factors layer, cost items layer and the total cost. Moreover, the nodes belonged to the criteria layer and factors layer be of three states in the model. The group analytic hierarchy process and linear weighting method are adopted to acquire the model parameters in order to decrease the workload and the difficulty of expert and the requirements for expert knowledge and experience during the parameters setting. Total cost for each alternative is computed by the model inference process in which quantitative indicators are transformed into evaluation distribution by introducing the conception of belief degree. The effectiveness of the proposed method is verified by a computational example of supplier selection for a car air conditioning which is on the base of the whole car development of a motor company. Furthermore, we make sensitivity analysis of the proposed Bayesian network model by means of the model parameters reflecting enterprise preference. © 2016 Journal of Mechanical Engineering.