Research On The Rotational Error Measurement And Fault Diagnosis Of CNC Machine Tool Spindle

As the "mother machine" of equipment manufacturing industry,CNC machine tool is one of the ten strategic areas of "made in China 2025".The spindle system as the "heart" of CNC machine tool,which has a direct impact on the product quality and processing efficiency of CNC machine tool.At the same time,due to the complex structure of the spindle system and the noise interference of the production environment,it is difficult to guarantee the machining accuracy and frequent failures,which not only causes huge economic losses but also threatens personal safety.Therefore,how to ensure the slewing accuracy and high stability of CNC machine tools is an urgent problem to be solved.In view of the above problems,this paper researchs the method of measuring spindle rotation error and fault diagnosis of CNC machine tools.Firstly,the spindle rotation error measurement model is established based on the three-point method,and the spindle rotation error measurement is realized by the multi-error separation technology.Secondly,an optimization method based on the combination of modal analysis and particle swarm optimization(PSO)is proposed to ensure that the measured signals contain all the features and have no cross redundancy.Finally,the fault diagnosis model based on BP neural network and D-S evidence theory is established to realize the fault diagnosis and state recognition of the spindle of the machine tool,and ensure the machining accuracy and working stability.First,in order to solve the problem of multiple error coupling in the process of measuring spindle rotation error,a measuring model of spindle rotation error of machine tool based on three-point method is established.Through theoretical derivation,the workpiece roundness error,spindle rotation error and installation eccentricity can be separated by this model.Considering the influence mechanism of installation eccentricity,the process of error separation is simplified,and the separation efficiency is improved on the basis of accurate separation of spindle rotation error and workpiece roundness error.In view of the noise amplification and harmonic suppression caused by the laser displacement sensor interval Angle,the placement Angle of the displacement sensor is optimized based on the harmonic error transfer factor and differential evolution algorithm.Secondly,in order to solve the problem that the fault mechanism of machine tool spindle is not clear and the sensor is difficult to be effectively and rationally arranged,the fault mechanism of each component is deeply analyzed and the sensor placement method based on modal analysis and particle swarm optimization algorithm is proposed.Starting with the rolling bearing and transmission gear,which are common components of CNC machine tool spindle,this paper discusses and summarizes the common fault types and failure causes,analyzes the fault frequency and signal characteristics of bearing and gear,and selects vibration signal as the status monitoring signal of machine tool spindle.Based on modal analysis and particle swarm optimization,the layout of the measuring points of the multi-vibration sensors was optimized.In order to make the measured signals of the multi-vibration sensors have no redundancy and crossover,and reflect the status characteristics of the spindle of the machine tool comprehensively.Thirdly,a fault diagnosis method of machine tool spindle based on multi-domain analysis and intelligent algorithm is proposed in order to realize accurate identification and fault diagnosis of machine tool spindle working state.The improved singular value decomposition(SVD)filtering and noise reduction method is used to reduce the influence of noise interference on vibration signals.Simulation results show that the proposed method achieves effective noise filtering without loss of effective features.In order to effectively and quickly obtain the status characteristics of machine tool spindle from a large number of vibration signals,the omni-directional feature index extraction and preliminary screening were carried out based on the time domain,frequency domain and frequency domain of vibration signals respectively.At the same time,Relief F algorithm and principal component analysis algorithm were used to reduce the dimension of the multi-dimensional index.Based on BP neural network and D-S evidence theory,the decision-making and fusion of multiple vibration sensors are completed,and the state recognition and fault diagnosis of machine tool spindle are realized.Finally,in order to verify the validity of the measurement of spindle rotation error,the layout method of multi vibration sensors and the fault diagnosis method.Based on the spindle of CNC machine tool,the spindle rotation error measurement experiment was carried out.The results show that the method can complete the spindle rotation error measurement after separating the roundness error of the table surface,and the measured result was 21.1 μm.In order to verify the multi-vibration sensor layout method proposed in this paper,the numerical control machine tool was taken as the verification object.The results show that the amplitude of characteristic frequency of the sensor layout method proposed in this paper is 0.005 g,and the amplitude of uniform interval layout is 0.001 g.Taking the gearbox as the verification object,the experimental verification of fault diagnosis method of BP neural network and D-S evidence theory is carried out.The results show that the fault diagnosis rate was poor when using a single sensor,and the accuracy rate was increased by 21.7% after using D-S evidence theory for decision level fusion.