Prediction Of The Rate Of Qualified Products Based On Improved BP Neural Network

Since complexity, uncertainty, nonlinear have been the leading characteristic of modernmanufacturing, traditional quality control methods can’t meet the requirements of modernmanufacturing. Most of these methods are hysteretic, that is to follow the idea of ProblemArising, Problem Analyzing, and then Problem Solving, which can’t control quality problemsthat may arise before manufacturing. In that way, quality control method based on predictionneeds to be adopted. Predictive control is a lead control, which can take full advantage of thehistorical and current quality information, then modeling and forecasting, further on givingfeedback to abnormal state, which therefore better meet the requirements of the developmentof modern manufacturing.This study analyzes and predicts the variation of the rate of qualified products based onneural networks, genetic algorithm and particle swarm optimization theory and modeling,combined with the rate of qualified products of an enterprise. First, reviews the study ofprediction of the rate of qualified products, analyzes advantages and disadvantages of variousprediction methods and the applicability and pointed out that this article will predict the rateof qualified products based on improved BP neural network method. Then, carry on thedesign research of BP neural network on prediction of the rate of qualified products, using acombination of theory and experiment, from the sample size, the number of hidden layer, thenumber of hidden layer nodes, and transfer function four aspects. Finally, in view of thelocal minima problem of the BP neural network, builds improved BP neural networkmodels respectively based on genetic algorithm and particle swarm algorithm, and predictsthe rate of qualified products of an enterprise to determine effectiveness of the improved BPneural network models.Experimental results show that two kinds of improved model has better predictionperformance than the BP neural network prediction model，and the accuracy of PSO-BPprediction model is better than GA-BP model.