Research On Quality Prediction Of Multi-variety And Small-batch Production Based On Combination Model

At present,the global manufacturing industry is constantly seeking to upgrade,the domestic industry is advancing industrial structural adjustment,and the market demand is also tending to be personalized and differentiated,manufacturing enterprises have shifted from mass production to multi-variety,small-batch flexible production models.It has become an urgent issue to develop quality control technology that is effective for the current production model.With the increasing maturity of intelligent technology,the introduction of artificial intelligence and machine learning technologies to build a quality intelligent control and prediction system has become a research hotspot for multi-variety,small-batch product quality control methods.In this thesis,the formation process and characteristics of the multi-variety and small-batch production model are analyzed,a quality control process suitable for multi-variety and small-batch production is constructed.Taking a certain axle sleeve product of R company as an example,the critical to quality identification and prediction in this process is demonstrated in detail by using complex network,grey theory and ls-svm method.(1)A complex network method based on graph theory is proposed to identify the critical quality of piece parts.According to the technological process and design requirements of internal relationship between the machining features of parts,establish the machining feature network of parts.By calculating the topological parameters of the network to determine the importance of the network nodes,then concluded that the left cylindrical size is the critical to quality,which lays the foundation for the piece parts machining quality prediction.(2)The GM(1,1)quality prediction model is established by using the grey prediction theory which has good adaptability to small sample data and good learning ability and prediction performance.Through the improvement of grey dynamic model group and grey relation degree weighting,an optimized grey relation weighting model group forecasting method is proposed.The prediction results show that the average absolute error and the mean absolute error of GM(1,1)prediction model are reduced by 57.4% and the mean absolute percentage error is reduced by about 7.3%,which are preparation for the establishment of the combined model.(3)Particle Swarm Optimization(PSO)optimizes the LS-SVM prediction model.The advantages and disadvantages of the LS-SVM algorithm for small sample prediction are analyzed.Based on the study of the algorithmic characteristics of swarm intelligence,a PSO algorithm for parameter optimization is proposed.(4)The combination model of quality prediction is constructed.Combining the pso-ls-svm prediction model with the grey relational weighted model group,a combined prediction model with residual compensation was established to effectively predict product quality and improve the prediction accuracy of a single model.Compared with the GM(1,1)model,the mean square error of the predicted value decreased from 0.6604 to 2.58e-04,and the mean absolute percentage error decreased from 12.96% to 0.2144%.The results show that the combined forecasting model is effective and can be applied to the quality control of multi variety and small batch production.