Research On Real-time Chatter Detection And On-line Suppression For Turning

Turning has been widely used in manufacturing systems such as aviation,aerospace,shipbuilding,chemical engineering,automotive and so on.Turning is one of the most basic machining processes.Machining chatter occurs in almost all kinds of cutting process.Chatter not only imposes negative effects on machining accuracy and surface quality,but also causes damage to machine tool components,resulting in high economic cost.The fast development of modern industrial technology has put forward higher requirements for part quality,productivity and yield.Therefore,it is essential to investigate in depth the mechanism,characteristics and suppression of chatter.This paper aims to avoid the occurrence of chatter in real machining processes,without reducing productivity.Based on theoretical analysis and machining experiments,this paper analyzes chatter vibrations from both theoretical and experimental perspective.We propose a real-time chatter detection and suppression system,which include chatter identification,chatter frequency estimation and chatter suppression.This paper is carried out in terms of following aspects:Propose an optimal weighted wavelet packet entropy(OWWPE)to detect the weak feature of the fault in the premature stage.By modelling WWPE in both stable and chatter states respectively,verify the existence of the optimal weight,which is obtained through machining experiments.The WP matrix is derived to replace the MAllAT algorithm,reducing the computation burden dramatically.The extreme value statistics is employed to achieve statistically rigorous threshold determination.In this way,the threshold can be obtained in a less empirical way than the traditional method of threshold determination.Experimental results reveal that the OWWPE is able to capture the weak characteristic of premature chatter,and therefore chatter can be detected before it has fully developed.Challenges of chatter frequency estimation are analyzed.Firstly,the chatter feature in its premature stage is very weak,leading to a low signal noise ratio.Secondly,the requirement of high timeliness leads to a short length of sliding time window,resulting in a small number of acquired sine wave cycle(NASC).We are the first in the world to derive the analytical expression of the interpolated DFT.We establish statistical properties of the approach contaminated with white noise,including upper and lower bounds of the theoretical variance.Simulation results indicate that the proposed algorithm is more accurate and 100 times faster than the optimizationbased algorithm.Moreover,it is illustrated that the simulation results agree well with upper and lower bounds of the theoretical variance.Furthermore,a switch-based IpDFT is proposed to provide high frequency estimation accuracy for NASC < 1.With the algorithm,chatter frequencies can be accurately estimation in an on-line way.Two high-order full-discretization methods using Lagrange interpolation are proposed for stability analysis of time periodic DDEs when the time-period/time-delay ratio is arbitrary,especially when the time period and the time delay are not integer multiples.The computational efficiency and accuracy of proposed algorithms are shown to be superior to those of existing discretization-based methods.The proposed algorithm of the second-order version is utilized to analyze the efficiency of chatter suppression strategies based on process parameter variation,i.e.spindle speed variation and stiffness variation.The effects of control parameters,i.e.the variation amplitude and variation frequency,on turning stability are investigated.This provides theoretical support for the selection and optimization of control parameters.This paper presents a real-time chatter detection and on-line suppression system(RCDOSS)to avoid chatter in turning processes.The weighted wavelet packet entropy is employed to monitor the turning process and it detects chatter in the premature stage.Once chatter is identified,the dominant chatter frequency is immediately estimated by the scale factor-based interpolated DFT.The spindle speed variation which periodically modulates the spindle speed around a nominal value,is instantly activated to suppress chatter.The variation parameters,i.e.amplitude and frequency,are computed based on dominant chatter frequency.In this way,chatter can be automatically identified and suppressed in an online fashion before it has fully developed.Secondary development of Simense 840 D is performed to timely update R parameters by external analog inputs from dSPACE.To experimentally verify the proposed method,the end-face turning of a flexible disc is carried out on a numerical control lathe with Siemens840 D.Two turning experiments with and without the RCDOSS are conducted using the same cutting parameters,cutting tool and workpiece.Experimental results demonstrate that chatter is automatically detected and suppressed by the RCDOSS before chatter arrives at fully developed stage.