A weighted hidden Markov model approach for continuous-state tool wear monitoring and tool life prediction

Tool wear is one of the important indicators to reflect the health status of a machining system. In order to obtain tool’s wear status, tool condition monitoring (TCM) utilizes advanced sensor techniques, hoping to find out the wear status through those sensor signals. In this paper, a novel weighted hidden Markov model (HMM)-based approach is proposed for tool wear monitoring and tool life prediction, using the signals provided by TCM techniques. To describe the dynamic nature of wear evolution, a weighted HMM is first developed, which takes wear rate as the hidden state and formulates multiple HMMs in a weighted manner to include sufficient historical information. Explicit formulas to estimate the model parameters are also provided. Then, a particular probabilistic approach using the weighted HMM is proposed to estimate tool wear and predict tool’s remaining useful life during tool operation. The proposed weighted HMM-based approach is tested on a real dataset of a high-speed CNC milling machine cutters. The experimental results show that this approach is effective in estimating tool wear and predicting tool life, and it outperforms the conventional HMM approach.     

基于小波包能量谱的HMM钻头磨损监测

从工程应用的角度论述了小波包分解原理及其能量谱监测理论，并将该理论应用于钻削力信号特征提取中，针对钻削过程特征矢量与钻头磨损之间具有较强的随机性和不确定性的特点，提出一种基于隐马尔可夫模型（HMM）的钻头磨损监测方法。实验结果表明，通过对钻削力信号进行多层小波包分解，提取各频段能量谱作为特征矢量可准确刻画工艺系统随钻头磨损的演化规律，利用HMM建立的各钻头磨损状态小波包能量谱的统计模型可有效跟踪钻头磨损的发展趋势，实现钻头磨损状态和寿命的监测。     

GRINDING WHEEL CONDITION MONITORING WITH HIDDEN MARKOV MODEL-BASED CLUSTERING METHODS

Hidden Markov model (HMM) is well known for sequence modeling and has been used for condition monitoring. However, HMM-based clustering methods are developed only recently. This article proposes a HMM-based clustering method for monitoring the condition of grinding wheel used in grinding operations. The proposed method first extract features from signals based on discrete wavelet decomposition using a moving window approach. It then generates a distance (dissimilarity) matrix using HMM. Based on this distance matrix several hierarchical and partitioning-based clustering algorithms are applied to obtain clustering results. The proposed methodology was tested with feature sequences extracted from acoustic emission signals. The results show that clustering accuracy is dependent upon cutting condition. Higher material removal rate seems to produce more discriminatory signals/features than lower material removal rate. The effect of window size, wavelet decomposition level, wavelet basis, clustering algorithm, and data normalization were also studied.     

GRINDING WHEEL CONDITION MONITORING WITH HIDDEN MARKOV MODEL-BASED CLUSTERING METHODS

Hidden Markov model (HMM) is well known for sequence modeling and has been used for condition monitoring. However, HMM-based clustering methods are developed only recently. This article proposes a HMM-based clustering method for monitoring the condition of grinding wheel used in grinding operations. The proposed method first extract features from signals based on discrete wavelet decomposition using a moving window approach. It then generates a distance (dissimilarity) matrix using HMM. Based on this distance matrix several hierarchical and partitioning-based clustering algorithms are applied to obtain clustering results. The proposed methodology was tested with feature sequences extracted from acoustic emission signals. The results show that clustering accuracy is dependent upon cutting condition. Higher material removal rate seems to produce more discriminatory signals/features than lower material removal rate. The effect of window size, wavelet decomposition level, wavelet basis, clustering algorithm, and data normalization were also studied.     

Application of backpropagation neural network for spindle vibration-based tool wear monitoring in micro-milling

This study develops a micro-tool condition monitoring system consisting of accelerometers on the spindle, a data acquisition and signal transformation module, and a backpropagation neural network. This study also discusses the effect of the sensor installations, selected features, and the bandwidth size of the features on the classification rate. To collect the vibration signals necessary for training the system model and verifying the system, an experiment was implemented on a micro-milling research platform along with a 700?μm diameter micro-end mill and a SK2 workpiece. A three-axis accelerometer was installed on a sensor plate attached to the spindle housing to collect vibration signals in three directions during cutting. The frequency domain features representing changes in tool wear were selected based on the class mean scatter criteria after transforming signals from the time domain to the frequency domain by fast Fourier transform. Using the appropriate vibration features, this study develops and tests a backpropagation neural network classifier. Results show that proper feature extraction for classification provides a better solution than applying all spectral features into the classifier. Selecting five features for classification provides a better classification rate than the case with four and three features along with the 30?Hz bandwidth size of the spectral feature. Moreover, combining the signals for tool condition from both direction signals provides a better classification rate than determining the tool condition using a one-direction single sensor.     

基于主轴电流信号多特征融合的刀具磨损状态监测

为实现在正常生产条件下进行刀具磨损的长期在线监测,提出了基于主轴电流信号和粒子群优化支持向量机模型(PSO-SVM)的刀具磨损状态间接监测方法.首先对数控机床主轴电机电流信号进行分析,将与刀具磨损相关的主轴电流信号多个特征参数和EMD能量熵进行特征融合作为输入特征向量;其次,通过粒子群寻优算法(PSO)对支持向量机模型...     

A COMPARATIVE STUDY OF FEATURE SELECTION FOR HIDDEN MARKOV MODEL-BASED MICRO-MILLING TOOL WEAR MONITORING

This paper presents a discriminant feature selection approach for hidden Markov model (HMM) modeling of micro-milling tool conditions. The approach is compared with other popular feature selection methods such as principal component analysis (PCA) and automatic relevance determination (ARD) according to their HMM classification rate. In tool condition monitoring (TCM), there are a lot of features that contain redundant information or less sensitive to tool state discrimination. These features are expected to be deleted for less computation and more robust modeling of tool conditions. Fisher's linear discriminant analysis (FDA) is modified for this purpose. The FDA is generally used for classification, and the features are mapped to another space and lose their physical meanings. In the modified discriminant feature selection, the features are selected in the original feature space by maximizing tool state separation and ranked by their separation ability between different tool states. Experimental results from both micro-milling of copper and steel under different working conditions indicate that the FDA is superior to both PCA and ARD for feature selection in HMM's classification. The reasons behind these differences are also discussed.     

Condition monitoring and classification of rotating machinery using wavelets and hidden Markov models

Condition monitoring and classification of machinery state is of great practical significance in manufacturing industry, because it provides updated information regarding machine status on-line, thus avoiding the production loss and minimising the chances of catastrophic machine failure. In this paper, the condition classification is based on hidden Markov models (HMMs) processing information obtained from vibration signals. We present an on-line fault classification system with an adaptive model re-estimation algorithm. The machinery condition is identified by selecting the HMM which maximises the probability of a given observation sequence. The proper selection of the observation sequence is a key step in the development of an HMM-based classification system. In this paper, the classification system is validated using observation sequences based on the wavelet modulus maxima distribution obtained from real vibration signals, which has been proved to be effective in fault detection in previous research.     

Tool condition classification in turning process using hidden Markov model based on texture analysis of machined surface images

Tool condition monitoring has found its importance to meet the requirement of production quality in industries. Machined surface texture is directly affected by the extent of tool wear. Hence, by analyzing the machined surface images, the information about the cutting tool condition can be obtained. This paper presents a novel technique for tool wear classification using hidden Markov model (HMM) technique applied on the features extracted from the gray level co-occurrence matrix (GLCM) of machined surface images. The tool conditions are classified into sharp, semi-dull and dull tool states. The proposed method is found to be cost effective and reliable for on-machine tool classification of cutting tool wear with an average of 95% accuracy.     

采用改进CNN-BiLSTM模型的刀具磨损状态监测

自动化切削加工过程中,准确可靠地监测刀具磨损状态是保证加工质量和加工效率的关键。针对刀具磨损状态相关特征提取繁琐、准确率低及传统的深度学习网络不能全面提取数据隐含信息等问题,提出了一种以卷积神经网络（CNN）和双向长短时记忆（BiLSTM）网络集成模型为基础并通过在卷积神经网络中添加批量标准化层和采用两个双向长短时记忆网络层的改进模型,该模型通过自动提取小波阈值降噪等预处理和降采样后的切削力、振动和声音信号的空间和时序特征来实现刀具磨损状态监测。将改进模型与CNN-BiLSTM模型及传统的深度学习模型进行对比,发现改进模型在精度和稳定性方面有较大提升。所提方法为准确监测自动化加工过程中刀具磨损状态、提高生产效率和加工质量提供了技术支持。     

基于切削声信号与优化SVM的刀具磨损状态监测

提出了一种利用切削声实现刀具磨损状态多特征监测的方法。根据经验模态分解与Hilbert变换理论,提取切削声信号的内禀模态能量与不同频段的Hilbert谱能量作为监测信号的备选特征。采用支持向量机作为分类器,针对备选特征的有效筛选问题,利用多种群遗传算法对分类器的输入特征进行了优化,剔除备选特征中的干扰特征,利用多种群遗传算法对分类器的模型参数进行了优化。利用优化后的分类器对测试样本进行分类,并与优化前的分类结果进行了对比。结果表明,优化后分类器的分类性能得到了明显提升,该方法可以对刀具磨损状态进行有效识别。     

滚动轴承故障程度诊断的HMM方法研究

为提高滚动轴承故障诊断率,提出基于时频域指标的HMM轴承故障程度诊断方法。利用轴承故障模拟试验台,采集不同剥落程度滚动体的振动信号,分别提取均方值、有效值、方差、修正样本方差、标准差、频域中心及带宽共7个时域和频域指标作为特征向量训练HMM,得到基于HMM的诊断分类器。利用该分类器对330组待检滚动轴承振动信号进行分析,诊断正确率达90%以上,说明该方法能有效提取故障特征。     

Identification of location and size of a defect in a structural system employing active external excitation and hybrid feature vector components in HMM

For the fault diagnosis of a mechanical system, various kinds of methods have been developed so far. For a structural system having a defect, pattern recognition methods such as Hidden Markov model (HMM) and Artificial neural network (ANN) are widely used in engineering fields. A statistical model can be constructed with one of the methods using various signals that are extracted from the structural system of interest. In the present study, a HMM employing hybrid feature vector measures is proposed for the fault diagnosis of a structural system having a defect. To obtain the hybrid feature vector components, five frequency response peaks obtained with FFT and two additional components obtained with ANN are employed. For the proposed method, an active external excitation having some specific frequency components is also applied to the structure to overcome the noise effect. To verify the effectiveness of the proposed method, a numerical model of a rotating blade having a crack is employed. Acceleration signals extracted from the structural system are employed to develop the proposed model so that the location and size of the crack can be identified. Using the proposed method, the diagnostic accuracy of the identification is significantly improved even with high level of noise in the system.     

多传感器融合下多工况刀具磨损状态预测的深度森林方法研究

准确监测加工过程刀具磨损状态有助于避免因刀具失效导致的产品质量问题。 建立不同工况的刀具磨损监测模型,往 往需要对每组工况调参以保证精度。 为减少调参并保证预测精度,结合深度森林的超参数少、参数对模型不敏感和训练过程自 适应等优点,利用深度森林建立了多传感器信号及多工况下自主特征选择的刀具磨损状态预测模型。 基于 3 组不同工艺参数 下 TC18 铣削过程的多传感器及磨损数据,以及预测与健康管理(PHM)学会 2010 年高速数控机床刀具健康预测竞赛的开放数 据,深度森林在 3 组工况的预测精度分别为 95. 35% 、96. 63% 和 97. 06% ,在 PHM 数据上为 98. 95% ,验证了深度森林对多工况 下刀具磨损预测的高精度和适用性,为在线监测技术提供了有力的指导。     

主分量分析和因子隐Markov模型在机械故障诊断中的应用

主分量分析(principal component analysis,PCA)是统计学中分析数据的一种有效方法,可以将高维数据空间变换到低维特征空间,因而可用于多通道冗余消除和特征提取.因子隐Markov模型是隐Markov模型的扩展,它比隐Markov模型更有优势,适用于动态过程时间序列的建模,并具有强大的时序模型分类能力,特别适合非平稳、信号特征重复再现性不佳的信号分析.文中结合主分量分析与因子隐Markov模型,提出一种新的故障识别方法,即以主分量分析方法进行冗余消除和故障特征提取,因子隐Markov模型作为分类器.并应用到机械故障诊断中,同时与基于主分量分析的隐Markov模型的识别方法相比较,实验结果表明基于PCA的因子隐Markov模型识别法和基于PCA的隐Markov模型识别法在故障识别上都是有效的,但对于相同的状态空间,前者的训练速度快于后者,尤其是状态空间越大,这种优势越明显.     

基于自适应神经-模糊推理系统模糊信息融合的采煤机截齿磨损在线监测

为实现对截齿截割过程中磨损程度的实时精确在线监测,分别测试和提取不同磨损程度的截齿在截割过程中的振动信号、声发射信号和温度信号,建立不同磨损程度截齿截割信号的多特征样本数据库,根据最小模糊度优化模型计算求解各特征信号的最优模糊隶属度函数,采用自适应神经-模糊推理系统多维模糊神经网络方法实现多传感特征信息的决策融合,输出置信度和权重较高的截齿磨损量融合结果。通过随机测试实验对融合系统进行验证,结果表明,基于ANFIS模糊信息融合的截齿磨损监测系统辨识度较高,测试结果最大误差在6.5%以内,系统具有良好的融合效果以及较高的测试精度。     

Cutting force-based real-time estimation of tool wear in face milling using a combination of signal processing techniques

In this paper, combinations of signal processing techniques for real-time estimation of tool wear in face milling using cutting force signals are presented. Three different strategies based on linear filtering, time-domain averaging and wavelet transformation techniques are adopted for extracting relevant features from the measured signals. Sensor fusion at feature level is used in search of an improved and robust tool wear model. Isotonic regression and exponential smoothing techniques are introduced to enforce monotonicity and smoothness of the extracted features. At the first stage, multiple linear regression models are developed for specific cutting conditions using the extracted features. The best features are identified on the basis of a statistical model selection criterion. At the second stage, the first-stage models are combined, in accordance with proven theory, into a single tool wear model, including the effect of cutting parameters. The three chosen strategies show improvements over those reported in the literature, in the case of training data as well as test data used for validation—for both laboratory and industrial experiments. A method for calculating the probabilistic worst-case prediction of tool wear is also developed for the final tool wear model.     

ESPRIT- and HMM-based real-time monitoring and suppression of machining chatter in smart CNC milling system

Suppression of machining chatter during milling processes is of great significance for surface finish and tool life. In this paper, a smart CNC milling system integrating the function of signal processing, monitoring, and intelligent control is presented with the aim of real-time chatter monitoring and suppression. The algorithm of estimation of signal parameters via rotational invariance techniques (ESPRIT) is adopted to extract the frequency characteristics of acceleration signals, and then, cutting state is categorized as stable state, chatter germination state, and chatter state based on amplitude-frequency characteristics of identified acceleration signals. The model of chatter identification is acquired by training a hidden Markov model (HMM), which combines acceleration signals and labeled cutting state. To implement real-time chatter suppression, the algorithm of fuzzy control is integrated into a smart CNC kernel to determine the relationship between cutting force and spindle speed. Furthermore, spindle speed of machine tool could be adjusted timely in the presented system once the chatter is identified. Finally, the effectiveness of the proposed real-time chatter monitoring and suppression system is experimentally validated.     

Multi-category micro-milling tool wear monitoring with continuous hidden Markov models

In-process monitoring of tool conditions is important in micro-machining due to the high precision requirement and high tool wear rate. Tool condition monitoring in micro-machining poses new challenges compared to conventional machining. In this paper, a multi-category classification approach is proposed for tool flank wear state identification in micro-milling. Continuous Hidden Markov models (HMMs) are adapted for modeling of the tool wear process in micro-milling, and estimation of the tool wear state given the cutting force features. For a noise-robust approach, the HMM outputs are connected via a medium filter to minimize the tool state before entry into the next state due to high noise level. A detailed study on the selection of HMM structures for tool condition monitoring (TCM) is presented. Case studies on the tool state estimation in the micro-milling of pure copper and steel demonstrate the effectiveness and potential of these methods.     

基于无限隐Markov模型的旋转机械故障诊断方法研究

针对传统隐Markov模型(HMM)在机械故障诊断中存在的不足,即HMM过学习或溢出问题以及隐状态数需要事先假定,提出了基于无限隐马尔可夫模型(i HMM)的机械故障诊断方法。在提出的方法中,以谱峭度为特征提取,i HMM为识别器,并以最大似然估计来确定设备运转中出现的故障类型。同时,将提出的方法与传统的HMM故障识别方法进行了对比分析。实验结果表明,提出的方法是有效的,得到了非常满意的识别效果。提出的方法能够有效避免了HMM在建模初期遗留下的不足,可以自适应确定模型中隐藏状态数和模型数学结构,因此,提出的方法明显优于HMM故障识别方法。