A Chatter Prediction Method Based On Control Chart And ARIMA Model

Grinding is an indispensable processing method for the surface of parts in modern machinery manufacturing industry which can achieve high-precision machining.But grinding chatter will produce a series of negative effects on machining quality in machining process.Due to the non-stationary,multidimensional and nonlinear characteristics of grinder's vibration signals,traditional signal analysis methods such as short-time Fourier Transform,singular value decomposition and S transform have great limitations in dealing with such signals.BEMD can effectively extract the characteristics of grinding chatter to identify and predict while BEMD has the characteristics of completeness,orthogonality and adaptability.It is of great engineering significance to identify and predict the grinding chatter to ensure the normal operation of the grinder and improve the precision of the machined parts.Therefore,KD4020X16 type CNC Longmen Guideway Grinder is taken as the research object in this paper.A chatter signal sampling platform is set up and the vibration conditions of the grinder under different grinding parameters are tested.On this basis,several aspects such as vibration signal processing,chatter state recognition and chatter prediction are studied.(1)The development status of grinder vibration signal processing,grinding chatter recognition technology and grinding chatter prediction technology is expounded.And summarizing the current problems.(2)Chatter characteristics of the vibration signal extracted from the grinder.The vibration signals in the direction of X and Z of the main shaft of the grinding wheel have been reconstructed in the test,a two-dimensional complex-valued test signal of the grinder vibration was obtained.BIMF component is obtained by processing the test signal through BEMD.And according to the orthogonal characteristics of each component,an intrinsic modal function based on the correlation coefficient is proposed to extract the effective intrinsic modal functions.The real-time variation is obtained by comparing characteristic variables.The real-time variation is taken as the characteristic of chatter,and the real-time variation of the test signal is extracted.(3)It is difficult to determine the chatter threshold,a control chart model is established to identify the working state of the grinder.Selection the real-time variation of stable phase sequence as feature.On the basis of verifying normality and correlation,the upper and lowercontrol limits of the control chart are obtained by ?3? rule,then,the control chart model is established.The following real-time variation is input into the control chart model,The real-time variance sequence is monitored according to the criterion of the control chart,and the grinding chatter status of the grinding machine is identified.(4)In order to predict the time of flutter,this paper combines the control chart with the ARIMA model.And the stationary time series is obtained by smoothing the sequence.By solving the auto-correlation and partial auto-correlation functions of the stationary time series,the order of the model is identified and the optimal ARIMA model is established.So,the ARIMA model can be used to predict the working state of the grinder.A method combined the control chart and ARIMA model is proposed to predict chatter.