Research On Prediction Method Of Bending Machine Compensation Value Based On Improved Genetic Algorithm To Optimize BP Neural Network

Bending machine is a kind of bending and forming equipment that is widely used in the sheet metal industry and plays a very important role in the sheet metal processing process.In the actual production process,due to the equipment operation,the bending accuracy of the bending machine needs to be improved by the bending compensation value,and the determination of the bending compensation value is too dependent on manual experience.Moreover,it is necessary to perform multiple trial foldings for the plate to obtain an accurate compensation value.The above trial folding not only takes a long time but also causes waste of the plate.Therefore,a BP neural network model based on the improved genetic algorithm is proposedto predict the bending compensation value according to the bending machine parameters accurately in this thesis.Here not only the necessary procedure to determine the compensation value by the manual experience can be avoided,but also the operation time and labor costs can be reduced.And meanwhile the bending accuracy can be improved,which provides new ideas for the research in this field.The main research work in this thesis are as follows.(1)Based on the analysis of the current research status in this field and the complexity of the mechanism model,the BP neural network is choosen as the prediction model in this thesis.Since the BP neural network has powerful nonlinear mapping capabilities,the nonlinear system can be effectively modeled,and hence the complex function relationship between the bending parameters and the bend compensation value can be well fitted.However,the above algorithm also has some shortcomings.For example,the traditional BP neural network has slow training speed and is sensitive to the initial weights,which makes it easy to fall into the local minimums.(2)Aiming at improving the above shortcomings of the traditional BP neural network,the principal component analysis method is firstly used to reduce the dimension of the process parameters of the bending machine to achieve the purpose of removing the correlation between the data and reducing the number of the inputs;(3)Secondly,on the basis of the traditional BP algorithm,the genetic algorithm is combined with the BP neural network.The global search performance of the genetic algorithm is used to determine the initial weight of the BP neural network,which effectively overcomes the local convergence problem of the traditional BP algorithm.However,the traditional genetic algorithms have the disadvantages such as poor local search capabilities and being prone to premature;(4)Finally,in view of the shortcomings of the traditional genetic algorithms,a genetic algorithm based on accelerated search strategy(ASSGA)is proposed in this thesis,which improves the genetic operator of the traditional genetic algorithm and introduces the local search algorithm.Combining the improved genetic algorithm with the BP neural network can better overcome the problem that the BP neural network is easy to fall into the local optimum and can effectively improve the generalization performance of the neural network,which finally improves the prediction accuracy of the bending compensation value.The simulation results of the actual operation data of a type of bending machine show that by using the improved genetic algorithm to optimize the BP neural network model to predict the bending compensation value,the performance indicators of the prediction results are better than the traditional BP neural network model and the BP neural network mode using the traditional genetic algorithm.l Moreover,the prediction accuracy reaches the process requirements,which is of great significance to the realization of the accurate bending processing of the bending machine.