| Rare earth is a non-renewable and important strategic resource,known as "industrial gold".The quality of rare-earth alloys determines the quality of products in the downstream rare-earth industry chain,and the carbon content of praseodymium(Pr-Nd)and other rare-earth alloys can reflect the quality of the alloys most directly and quickly.Therefore,in order to ensure the quality of rare-earth metal products,the carbon content of praseodymium and other rare-earth metals must be detected in real time.However,in actual production,at present,it mainly relies on off-line chemical analysis method to get.And the manual laboratory analysis of carbon content has problems such as long testing period,high testing cost and can not achieve full inspection,which is far from meeting the requirements of modern large-scale praseodymium and other rare earth alloy production process real-time control.Soft Sensing technology is one of the best choices to solve this problem with its advantages of strong real-time and software instead of partial hardware.In this paper,with the support of Jiangxi Province Key R&D Program(20192BBE50010),we take the rare earth alloy production workshop of an enterprise in Ganzhou as the implementation area to study the Soft Sensing detection system of praseodymium alloy carbon content based on cutting force characteristics to realize the online real-time and intelligent detection of praseodymium and other rare earth alloys carbon content.The main research contents are as follows.(1)Deep analysis of the production process and characteristics of praseodymium alloy prepared by molten salt electrolysis of fluorine salt system,for the defects of current chemical analysis method,this paper proposes a physical carbon measurement method based on cutting force characteristics,establishes the Soft Sensing mechanism model between cutting force of alloy and carbon content of praseodymium alloy,and studies the correlation law between them.For the problem that a single Soft Sensing model can hardly meet the complex working conditions,three data processing algorithms based on machine learning are proposed to find the optimal input features of the original signal of cutting force to improve the detection accuracy.(2)To verify the Soft Sensing model of cutting force to predict the carbon content of praseodymium alloy,an online automatic detection system based on virtual instrument technology(Lab VIEW)is designed.This paper describes in detail the overall design scheme,testing principle and overall structure,and analyzes the important components of the automatic testing system,and uses finite element analysis(Ansys)technology to simulate and verify the strength,stiffness and other institutional functions of the testing device.Based on the above design ideas,the functions of each module in the software part of the system are determined to ensure a stable connection between the two parts of software and hardware.(3)The experimental results show that the Soft Sensing detection system of praseodymium alloy carbon content based on cutting force characteristics can achieve online accurate detection,which is better than the traditional chemical detection method,and compare and analyze the recognition rate of three algorithm models of machine learning,among which the SVM algorithm has the highest detection accuracy,and finally the Soft Sensing model is integrated into the practical application of detection. |
PDF Full Text Request