Technical Risk Identification And Assessment For Complex Product Development Based On Bayesian Network

Complex product is getting more and more important in modern society, and it plays an important role in the development of a country. But complex product has a complex structure and uses all kinds of high technologies, the success ratio is very low in the process of development and researchers will face various major risks, so it is very important to study different kinds of risks which will be faced in the process of development. And due to the technical risk is as one of the most important and most primary risks in the development, so the special research for it is very urgent. In order to take an effective identification and accurate assessment for the technical risks, this article made some beneficial improvement based on the existing research results.For the technical risk identification problem, the basic idea of identifying technical risk in this paper is to find the technical risks and technical risk factors by going deep into the component level and functional module level of the product. The entire identification process is divided into three steps: firstly, take the PBS and WBS as a secondary identification tool, find the technical risk factors that resulted technical risk in the functional model by using the risk checklist. Then, establish the relationship between the various risk factors by using the analytic structure model(ISM). Finally, transform it into a Bayesian network model and it is a preparatory work for the further study to quantify the probability and consequences of the technical risk. This method can improve the efficiency and accuracy of identification.For the technical risk assessment problem, in order to improve the accuracy of technical risk estimation of complex product development project, this article studies a new method of learning Bayesian network parameters under small samples with missing value data. Specifically, this paper restores the missing data using the machine learning theory, the method of neural network and support vector machine, and then, compares it with the traditional EM algorithm in accuracy, eventually, we can draw out that SVM can repair the missing data more excellently in accuracy. Through this research, this paper can create a more accurate Bayesian network model for the R & D technology risk assessment of complex products. Then, we can estimate the probability of technical risk of complex product development project more accurately, and provide decision support for the risk management of complex product development project.