| With the introduction of Made in China 2025 and Industrialization 4.0,the national manufacturing capacity demand is gradually increasing,the process requirements are getting higher and higher,and the demand for advanced CNC machine tools is increasing day by day.The research and development,manufacturing and maintenance of high-precision and highreliability equipment are essential for ensuring manufacturing.The stable development of industry and economy plays an extremely important role.The machine tool equipment is composed of many complex subsystems.Among various subsystems,the electric spindle plays a key role.As a rotating part,the electric spindle needs to ensure the stability of the rotation process on the one hand,but also the reliability of the machining process.The machining efficiency,machining accuracy and life of the machine tool are affected by the operating status of the spindle,and the status of the spindle also affects the production efficiency of the production line.Periodic or continuous observation of the operating status of the spindle can detect early signs of abnormalities in time,eliminate hidden dangers in time,and effectively realize predictive maintenance of equipment and reduce maintenance costs.Therefore,in order to deal with the diagnosis of abnormal vibration of the electric spindle system,it is necessary to propose corresponding analysis methods and evaluation methods.This article starts with the electric spindle of the horizontal machining center of the actual powertrain production line,mainly by collecting the vibration signal at the front bearing of the electric spindle,combining the vibration signal analysis method and the deep transfer learning analysis method to evaluate the vibration condition of the electric spindle.The main relevant contents are as follows:(1)The vibration signal acquisition process of the electric spindle and the comparison of the results of the vibration signal analysis methods.It mainly describes the sensor arrangement and vibration signal collection scheme at the front bearing of the electric spindle of the horizontal machining center,and conducts preliminary analysis and comparison by collecting actual vibration signals.First,use wavelet denoising to filter out the signal noise,and then verify the feasibility of different time-frequency processing methods such as wavelet analysis,fourier transformation,and empirical mode decomposition on the actual signal analysis results.(2)Research on experiments and related analysis methods based on abnormal vibration signals.Aiming at the two types of abnormalities that often appear in the actual working conditions of the electric spindle: the chip clamping between the spindle and the tool,the self-excited vibration of the machining process,two types of abnormal vibration signal acquisition experiments are designed,and the vibration data of the related processes are analyzed separately.Vibration signal features generate corresponding feature images: spectral kurtosis diagram and wavelet time-frequency diagram,and distinguish different vibration states of the main shaft with feature images.(3)Analysis of abnormal vibration signal of spindle based on deep transfer learning.Using the method of deep transfer learning,the deep transfer learning model uses the VGG16 model for fine-tuning,classifies and recognizes the vibration signal feature map under abnormal conditions of the spindle and the vibration signal feature map under normal conditions,analyzes the results and accuracy of training model,compares two situations separately.The abnormal conditions are analyzed and judged to prove the actual effectiveness of the method.(4)For the judgment of spindle health and abnormality,based on the current mainstream signal analysis method: support vector machine,through the research and comparison of deep transfer learning and support vector machine methods in this article,combined with actual machine tool data,it verifies the practicality,reliability and high accuracy rate of deep transfer learning. |
PDF Full Text Request