Research On The Application Of Ant Colony Algorithm In NdFeB Hydrogen Pulverization Control System

As a milling method for preparing in emerging in the rare earth permanent magneticmaterials, hydrogen crushing of NdFeB is widely used all around the world. The state ofalloy crushing,the hydrogen content in alloy etc. can not be detected online by sensors,andthis situation limit the optimal control in the process of NdFeB hydrogen pulverization.How to improve the quality of NdFeB hydrogen crushed products,shorten the productioncycle and reduce the manufacturing costs in the process of control is urgent problem to besolved for the manufacturer.In this paper,we make thorough the exploration and research on the determinants ofalloy smash quality and relevant factors causing prolonged cycle of smash in the hydrogendecrepitation process.Against the features such as nonlinear and multi-variable thatproduced in the hydrogen pulverization process,on the basis of the preliminary work whichcompleted by the research team,then we can solve the problem of detection and controlwhich based on the thought of data-driven.The main contents of this article which havebeen completed include:Firstly,analyze and study the related factors which determine the quality of thecrushing alloy and cause the crushing cycle prolonged in the angle of cross-disciplinarybetween the preparation of magnetic materials and control area.Secondly,on the basis of the preliminary work which completed by the researchteam,against the problem that the hydrogen content in alloy can not be detected online,theimproved ant colony algorithm and soft-measurement technology are integrated tooptimize the parameters of the regression prediction model which based on the supportvector machine,then we can achieve the purpose of improving the prediction accuracy.Thirdly,we complete the online test,optimization and validation of the model above onthe basis of the NdFeB hydrogen decrepitation equipment in the laboratory,then we cangive the contrast of prediction performance of the regression model between the hydrogencontent regression model built by the improved ant colony algorithm which used to optimizethe parameters of support vector machine and the hydrogen content regression model builtby PSO-SVR.Fourth,object to the NdFeB hydrogen decrepitation equipment in the laboratory,wehave completed the design,installation and debugging of NdFeB hydrogen pulverizationprocess control system(undertake the design of function module such as the OPCcommunication between configuration monitoring interface and MATLAB predictionmodel,the data acquisition of hardware devices and the real-time communication withconfiguration monitoring interface etc.).