基于改进的BP神经网络的产品生产成本估算

基于BP神经网络的成本估算模型,利用全局搜索能力较强的遗传算法优化BP神经网络连接权,克服传统的BP算法易陷入最小值的缺点,使模型预测性能、预测精度和泛化能力得到有效改进.以列车转向架为例,建立产品生产成本GA-BP估算模型,通过8组检测样本检验训练好的遗传人工神经网络.计算结果表明:预测值与期望值的误差小于4%,说明利用遗传神经网络模型对产品成本进行估算切实可行.

Production cost prediction of product based on improved BP artificial neural network

In order to predict the production cost of product in the design process, a model of production cost prediction of product was set up based on BP artificial neural network. During the process, to avoid local minimum points of training traditional neural network, the neural network connection weights were improved by genetic algorithm, which had better search ability. The prediction property was optimized effectively, prediction precision and generalization ability of the model were improved. In the last, taking the bogie as an example, the GA-BP prediction model of production cost was set up, and the model was tested by eight samples. The results showed that the prediction error was less than 4%, which indicated that it was applicable to predict the production cost of bogie by the model.