Research On Thermal Error Characterization For Z-axis Feed Axis Of CNC Machine Tools And Parameter Identification Algorithm

CNC machine tools are industrial mother machines,basic equipment for manufacturing equipment,and important strategic materials of the country.Its high precision and high stability are the goals that are continuously pursued.The thermal error caused by the thermal deformation caused by the influence of internal and external multiple heat sources during the operation of CNC machine tools is the main error source affecting the high-precision machining of CNC machine tools.The error avoidance method of reducing the thermal error of CNC machine tools by physical means will affect the processing efficiency of CNC machine tools and bring additional processing costs.The error compensation method of compensating the thermal error of CNC machine tools by establishing a thermal error model with software compensation has wide applicability and significant cost advantages.It is an important research hotspot in the field of precision improvement of CNC machine tools and its general modeling method is still the current research difficulty.At present,the research of deep learning artificial intelligence algorithm in thermal error compensation modeling has received extensive attention.Usually,such modeling methods are extremely dependent on temperature data,screening of temperature sensitive points and large sample temperature data collection.This prolongs the model building cycle and makes it difficult to realize engineering applications.At the same time,affected by the time-delay characteristics of structural thermal deformation,the prediction accuracy and stability of the thermal error model cannot meet the needs of real-time applications.This paper takes the Z feed axis of a three-axis vertical CNC machine tool as the research object,and conducts relevant research on the problems existing in the above thermal error modeling research.The research content is as follows:(1)Aiming at the problem that the thermal error model is highly dependent on temperature data,the thermal error mechanism of the Z-direction feed axis of the CNC machine tool is analyzed,based on the actual working conditions of the Gaoke three-axis CNC machine tool,the Boosting method is improved,and the thermal error fusion model is proposed.The data acquisition experiment based on the arrangement of temperature sensors at equal intervals inside the CNC machine tool is designed to lay the foundation for subsequent modeling experiments.(2)Aiming at the low prediction accuracy and poor stability of the thermal error model,the genetic algorithm and the efficiency-first differential evolution algorithm were used to identify the internal parameters of the thermal error fusion model.The performance evaluation index of the thermal error model is proposed,the performance of the mainstream thermal error models is compared and analyzed,and the modeling experiment verification is carried out.The experimental results show that the thermal error fusion model has the highest prediction accuracy(the average prediction accuracy is as low as 6.0 μm)and the best stability.But when the genetic algorithm and differential evolution algorithm identify the internal parameters of the thermal error fusion model,they fall into the local extremum problem and over-fitting problem.(3)Aiming at the problem that genetic algorithm and differential evolution algorithm are easy to fall into the local extremum problem and overfitting problem,a DEGA parameter identification optimization algorithm combining genetic algorithm and differential evolution is proposed.The three parameter identification optimization algorithms were compared and analyzed,and modeling experiments were carried out to verify them.The experimental results show that the prediction accuracy of the thermal error fusion model identified based on the DEGA algorithm is significantly improved.Compared with the thermal error fusion model identified based on the genetic algorithm and the thermal error fusion model identified based on the differential evolution algorithm,the prediction accuracy is increased by 16% and 13%.