Tool is the most common part easy to wear and fail in CNC (computer numerical control) system, machining center and other highly integrated and intelligent equipment. Tool wear condition monitoring is of great significance to ensure the machining quality, machining accuracy and machining efficiency of the parts. At present, the common tool wear condition monitoring signal is vibration signal, cutting force signal, current signal and acoustic emission signal. Compared with other monitoring signals, the frequency of acoustic emission signal can reach 50 kHz - 1 MHz with high sensitivity and strong anti-interference ability, which can effectively filter the low frequency noise in the process. Therefore, this paper analyzed the characteristics state of tool wear’s acoustic emission signal. Metal in the cutting process will produce a wealth of acoustic emission signals, and these signals are easy to be affected by the comprehensive factors such as processing materials, cutting conditions and cutting parameters. A number of cutting experiments showed that the tool wear was uncertain under the same cutting conditions. Cloud model theory is a kind of qualitative concept to quantitative data uncertainty transformation model proposed by academician Li Deyi. The cloud theory has strong adaptability to the uncertainty problem. In view of the tool wear acoustic emission signal’s non-stationarity and uncertainty, signal analysis and feature extraction method were put forward based on cloud theory. This paper aimed at the research topic of tool wear condition monitoring under different cutting condition, and used orthogonal test method to arrange a large number of cutting tests. On the basis of the acquisition of the acoustic emission signal, wavelet packet analysis method was applied to realize the signal filtering processing. A cloud of uncertainty model theory was introduced into the feature extraction of the different cutting tool wear stages. First of all, the different stages of the wear band distribution of acoustic emission signal range were obtained through the spectrum analysis, and served as the wavelet packet decomposition levels of qualitative reference; second, the Shannon entropy in the information entropy theory was applied to characterize the size of the noise in order to determine the best wavelet packet decomposition tree; finally, we used statistical analysis method to determine the best wavelet packet decomposition tree of the optimal branch, and after signal threshold processing, signal was reconstructed, and the denoising effect was verified through the ratio of signal to noise. In view of the tool wear condition monitoring and wear prediction research, feature extraction is a key technology. Therefore, this paper put forward the signal feature extraction method based on cloud theory. First of all, according to the statistical distribution characteristics of reconstructed signal, backward cloud algorithm was utilized to extract the cloud characteristic parameters of signal sensitive band, i.e. expected value, entropy and hyper entropy; the change rule of the 3 types of cloud characteristics parameters of cutting tool with the increase of wear was quantitatively analyzed in different cutting conditions; second, the effectiveness of 3 kinds of parameters characterizing tool wear acoustic emission signal during feature extraction was analyzed through a scatter diagram; finally, the effectiveness of the cloud model to represent knowledge was verified through data histogram and cloud image contrast. The research results show that the tool wear acoustic emission signal has obvious characteristics of cloud, and 3 cloud characteristic parameters and tool wear status have obvious corresponding relation, which can be used for characteristic parameters of tool wear condition monitoring and wear prediction. Cloud theory is applied in the field of tool wear monitoring, expanding the scope of the representation of knowledge.%针对刀具磨损状态监测和磨损量预测研究中特征提取这一关键技术，该文提出了基于云理论的信号特征提取方法。首先，采用小波包分析对声发射信号进行信号分解和重构，滤除噪声对提取特征参数的影响；其次根据重构信号的统计分布特性，利用逆向云算法提取信号敏感频带的期望值、熵及超熵云特征参数，定量分析刀具在不同切削条件下3种云特征参数随磨损量增大所呈现的变化规律；最后，通过散点图分析3种特征参数表征刀具磨损声发射信号的有效性。结果表明：刀具磨损声发射信号具有明显的云特性，3种云特征参数与刀具磨损状态具有明显的对应关系，可作为刀具磨损状态监测、磨损量预测的特征参数；云理论在刀具磨损监测领域的应用，扩大了知识的表示范围。
展开▼
