Study On The Evaluation Of Contractor's Engineering Risks Under EPC Model

Recently, the mode of EPC has not been extensively used in every industry except chemical and power industries in china. However, there are many problems in the extensive application. In this mode, the general contractors have to face much more risk so that they can gain more profit and contractual rights. Therefore, the risk management is very important for the general contractor, while the identification and the evaluation of risk is the basis of risk management. So the study of the evaluation of contractor's risk has great realistic meaning.The theoretical study of construction project risk management has already obtained great achievement recently, and there have been many evaluation methods. But the commonly used methods have a shortcoming, which the weight of every factor is given by artificial set so that the decision making is not accurate. The neural network has good adaptability, self-organizability, strong learning-ability, associating-ability, fault-tolerance and anti-jamming ability, which can imitate human brain's information processing function at different level, and execute modeling for polygenous complex unknown parameters easily and flexibility. The neural network gets the weight of each factor according to the result of training by counter calculation, and the relative importance of each factor. Then this method can effectively avoid determining the weight subjectively.Based on the theory of the engineering project risk evaluation, this paper focuses on the contractor's risk in EPC mode, and introduces the development of EPC mode in recent years, the features of EPC mode, the basic theory of engineering project risk. Then this research studies the process and methods of identification and evaluation of the contractor's risk in EPC mode, summarizes the risk factors, and constructs the contractor's risk BP neural network evaluation model in EPC mode. A complete example was given to illustrate this model based on MATLAB. The result of this paper shows that this neural network model is feasible and has important application value.