Research On Error Compensation For Macro-micro Motion Ultra-precision 5-axis CNC Turn-milling Combined Machine Tool

In manufacturing,precision is considered to be a highly significant aspect.Precision has massive impacts on parts quality in machining processes.Demand for ultra-precision products is currently high from aerospace to electronics industry,which subsequently demands ultra-precision machine tools with a high positioning resolution to satisfy the submicron accuracy requirements.However geometric errors impacts the positioning accuracy of the machine axis motion negatively.Due to inevitability of the inherent geometric errors and external disturbance,compensation of these errors is significant.Later on precise detection of these errors for compensation incur time,expertise and cost associated challenges hence error prediction model based real time error compensation is becoming an attractive methods in machine learning to minimize machine tool error fast,robust and at low cost.A nanometer resolution positioning piezoelectric ceramic macro-micro motion platform hardware and control system is designed and tested based on the research on smart materials and precise error detection methods.The platforms nanometer motion micro stage is designed compensates for the micron level macro motion errors.A multibody system error compensation model for a macro micro motion ultra-precision five axis turn – milling complex machine tool running on a multi-axis motion controller PMAC is verified in a x-axis laser interferometry error detection experiment and micro stage compensation of macro errors database as a function of axis position.The average maximum positioning and straightness error before and after compensation reduced from 35?m to 0.006?m and 3.5?m to 0.45?m respectively with high repeatability.A novel support vector machine regression algorithm for response prediction with high prediction and better generalization,compensation accuracy and robustness is trained and tested on the experimental data using a grid search and k-fold cross validation method implemented by a SVR program code in R-studio.After tuning and obtaining optimum parameters with a kernel trick,the average maximum positioning and straightness x-axis motion error before compensation are predicted by 34.72?m and 3.46?m representing 99.2% and 88.5% prediction accuracy level respectively.Based on the prediction accuracy confidence levels a SVM error prediction based real time error compensation model is proposed to be implemented in the PMAC module.