The Forecast Of Copper-rare Earth's Performance Based On Artificial Neural Network

A reasonable neural network model was presented in this article. Using the method of bond of artificial neural network and genetic algorithm together. Taking different elements' content in the copper alloy to be the network input layer, and the performances' made the output layer to make a back propagation neural network, and then choose proper parameters of the sample number, learning-rate parameter, transfer function, momentum factor, the number and layer of hide layer. Using genetic algorithm to optimize the threshold value and weight of BP neural network, so as to obtain a reasonable neural network model. Utilizing MATLAB to realize the design for copper-rare earth performance forecasting software. The result indicates that the network works well. The electrical resistivity's relative error is smaller than 0.95%, so as the tensile strength. The hardness' relative error is smaller than 1.75%. This neural network can provides valuable reference for the copper alloy, which has the property of high strength and specific conductance.