Research On Intelligent Optimization Method Of Complex Grinding Process For Multi-Indicator Creation

Roll is the "teeth" of the rolling line and must be repaired by grinding after the rolls are worn out,therefore,roll grinding occupies an important position in steel rolling production.Roll grinding is a typical complex multi-step grinding process,which is difficult to establish the process parameters.At present,the process parameters of roll grinding mainly rely on the experience of grinding workers,and the stability and intelligence level are poor,while part of the process can guarantee the grinding quality but the efficiency is insufficient to meet the demand of efficient development of the steel industry.In view of the above problems,this paper carries out in-depth research on the process intelligent optimization method oriented to the creation of multiple indicators with roll grinding as the object,and the main contents are as follows.A study on the relationship between surface quality(roughness and gloss)of roll grinding and process parameters was carried out.Based on the single-factor test method and Box-Behnken response surface method,the single-factor effects and interactive effects of four grinding process parameters,including grinding depth,grinding wheel speed,workpiece speed and pallet speed,on surface quality were studied and discussed.Further full-factor tests and Box-Behnken tests were conducted to obtain the evolution model of surface quality under multiple passes of grinding,and the analysis of variance,significance test,adequacy test and misfit test were performed on the evolution model to achieve effective prediction of surface quality of roll grinding.A study on the relationship between roll grinding form accuracy(roll shape and roundness)and process parameters was carried out,and a strong non-linear relationship between form accuracy and process parameters was found through experiments.Since linear regression is difficult to fit the variation of profile accuracy,a roll shape error and roundness error prediction model based on the improved Elman neural network was established,and the Sine chaos mapping was used in the training process to make the population distribution more uniform.Based on the aforementioned prediction models of surface quality and form accuracy,an improved particle swarm algorithm-based multi-indicator process parameter optimization method for roll grinding is proposed,with material removal and processing time as the optimization objectives,grinding process parameters as the decision variables,surface quality as the constraint,and form accuracy as the test condition,while using dynamic weight learning strategy and adaptive grid strategy to improve the convergence speed.The validation results show that compared with the existing optimal empirical process parameters,the surface roughness is improved by 24.26%,the surface gloss reaches 91.4 GU,the roll shape error is reduced by 17.5%,and the roundness error is reduced by 8.5%,while the total machining time is reduced by 33.46%,which fully proves the effectiveness of the proposed optimization method.The above research contents and results can provide process technology support for improving the quality and efficiency of roll grinding.