Research On Optimization Of Temperature Measuring Points In Thermal Error Compensation For CNC Machine

With the rapid development of modern industry, the requirement of machiningprecision is becoming higher and higher in manufacturing industry. Among all thesefactors affecting machining precision, thermal deformation of CNC machine is themost important one. According to relative research, thermal error which is induced bythermal deformation accounts for almost40%to70%of the over machining error. Tosolve this problem, researchers at home and aboard have done a lot of research. Asthe very first step of thermal error compensation, optimization of temperaturemeasuring points plays a critical role for improving the robustness and accuracy ofthermal error compensation model. In thermal error compensation, the number oftemperature measuring points varies from a few to a dozen. Generally, the more thetemperature measuring points are, the more accurate the compensation model is.However, too many temperature measuring points also lead to heavy work load ofarrange sensors and large calculating quantity of compensation model. On thecontrary, too few measuring points can’t provide enough message reflecting thechange of temperature field. So the reasonable and proper methods for selectingtemperature measuring points is the basement of constructing robust compensationmodel.Aimed at optimizing the temperature measuring points, two methods ofoptimizing measuring points are presented in this paper. For testifying these twomethods, thermal error detection experiment is also made. The main study contentsare as follows:Based on FMC CR5116, structure analysis and the analysis of the heat sourceare made. And experiment detecting thermal error is also planned according to theanalysis. Then, according to the experimental results, the relationship betweenthermal error in three directions and the structure is analyzed. In addition, the relationbetween the temperature change of the temperature measuring points and thermal error analyzed. Finally, it is concluded that optimizing measuring points is verynecessaryTwo methods of optimizing temperature measuring points are proposed. Oneoptimization method is a combination of thermal error sensitive stability analysis andfuzzy cluster analysis. Based on the definition of thermal error sensitivity, thedefinition of thermal error sensitive stability is put forward. And thermal errorsensitive stability is taken as the criteria of optimizing measuring points to makeprimary selection of points. Then, fuzzy clustering is used to optimizing temperaturemeasuring point further and to eliminate the coupling between the temperaturevariables. The other method combines the similarity analysis of time series andK-means clustering analysis. Through similarity analysis of temperature time seriesand thermal error time series, candidate temperature measurement points are selectedprimarily. Then k-means clustering is used to select temperature measuring pointwhich has weak coupling between each other.Two kinds of thermal error compensation model, multivariate linear regressionthermal error compensation and BP neural network thermal error compensationmodel are set up. And under the two different model, the verification of the twomethods of optimizing temperature measuring points is made.