Research On The Setting Method Of Standard Working Hour For Aerospace Fasteners Based On Neural Network

Working hours quota is an important basis for enterprise planning,resource scheduling and management optimization.In terms of the previous formulation methods,in order to ensure the accuracy of the calculation results and the interpretability of the principle,the managers focus on the analysis of the specific processing details,the study of production environment factors,and the establishment of regression equation of working hours.As a result,the construction process of working hours quota calculation model is cumbersome and the adaptability of flexibility is poor.And the contradiction between calculation accuracy and calculation efficiency is prominent,which can not adapt to the discrete manufacturing mode of small batch and multi variety of aerospace fastener manufacturing enterprises.Therefore,it has become the focus of many enterprises in this field to find a fast,flexible and accurate working hours quota calculation method.Aiming at this problem,in this paper,with the help of neural network’s computational advantage in the regression problem,the working hours quota corresponding to the conventional products of aerospace fastener manufacturing enterprises is taken as the research object,based on the real-time data collected by the business system,and based on the industry product spectrum characteristics and classification rules,the following conclusions are obtained: in this industry,the core task of working hours quota calculation is to calculate the average working hour of a single product.Then,this paper selects and analyzes the higher-order and more general feature factors that affect the working hours quota of single product,extracts and combines the data,preprocesses the characteristic factors,and finally forms the research data set.Furthermore,on this basis,this paper takes the calculation of the average working hours of a single product as the prediction object of the neural network model.According to the characteristics of the calculation tasks,BP neural network,Wide and Deep neural network and convolution neural network models are selected and built.The k-order verification mechanism is introduced to train,verify,test and compare the models.The data show that the prediction effect of convolution neural network is the best,the mean square error is 0.019,and the goodness of fit of the model is 0.858.Finally,using the new data to verify the calculation effect of the model.The experimental results are as follows: in the small batch and multi variety manufacturing mode,it is feasible to use the high-order features such as product classification attributes and neural network modeling to realize the prediction of single product working hours quota;compared with BP neural network and Wide and Deep network,CNN has better prediction effect.Also,the research results can provide a more rapid and accurate working hours quota calculation method for aerospace fastener manufacturing industry.