Research on the Evaluation Model of Engineering Quality Based on GA-BPNN

Artificial neural network (ANN) has been successfully applied into the engineering quality evaluation. With high robustness and fault-tolerant ability, this method works much better than multiple discriminant analysis (MDA) and logistic regression. In order to settle the traditional BP neural network´s problem of slow convergence speed and running into the local least value easily while estimating the engineering cost, the genetic neural network (GNN) is proposed as the estimation method in this paper. By combining the advantages of both genetic algorithm (GA) and NN, the new algorithm has not only the global random searching ability, but also the learning ability and robustness. In view of the defects of traditional error BP algorithm, chromosome coding is used to optimize the weight, the threshold and other main parameters of neural network. In addition, simulation test is used to test the stableness as well as the effectiveness of the network. The results show that this algorithm has high practicability and can be extensively applied in the estimation of engineering cost.