Wear State Recognition Research Articles

Tool wear state recognition under imbalanced data based on WGAN-GP and lightweight neural network ShuffleNet

Tool wear state recognition under imbalanced data based on WGAN-GP and lightweight neural network ShuffleNet

Tool wear state recognition based on feature selection method with whitening variational mode decomposition

• A signal decomposition method based on WVMD is proposed, which makes the intrinsic mode functions independent of each other and can effectively avoid mode mixing phenomenon. • The strong correlation force signal features from sensitive feature set are selected by proposed JIE method . Redundant features unrelated to tool wear are removed to achieve data dimension reduction. • The classification of tool wear state is realized by OPF algorithm, which can effectively reduce calculation time while ensuring classification accuracy . Tool states affect the surface quality and equipment stop time. It is essential to seek a method with high accuracy and efficiency to model and predict tool states. Cutting process signal is usually used to in industry monitoring, which contains lots of fault information. In this paper, the spectrum analysis of milling force is carried out to obtain the signal frequency that can reflect tool wear degree. A force signal decomposition model based on whitening variational mode decomposition (WVMD) is established to screen out sensitive signals, which effectively avoids mode mixing. Joint information entropy (JIE) is adopted to select the force signal features. Compared with other dimensionality reduction algorithms, the optimal feature subset obtained by JIE method can reflect strong relativity between features and the wear state. The classification is realized by the optimal-path forest (OPF) algorithm. According to the experimental results, compared with other recognition models, WVMD-JIE-OPF method has a classification accuracy as high as 98.41%. The training speed of OPF is increased by 58.43% compared with LSSVM, which shows excellent tool condition monitoring performance.

Control Force Compensation and Wear Monitoring of Variable Stiffness Joints in Drive Machining Process

Aiming at the joint flexibility and wear state existing in the process of driving mechanical parts, this paper first proposes a stiffness and position decoupling control method for variable stiffness joints, which realizes the joint position control and the unity of joint compliance. The joint stiffness model was obtained by using the static relationship between the Jacobian matrix and the model, and the nonlinear equations composed of the mechanical model and the stiffness model of the variable stiffness device were solved by the optimization method to realize the nonlinear decoupling of the stiffness and position of the variable stiffness joint. Secondly, this paper proposes an online monitoring method of wear state in the machining process based on machine tool information. In this method, OPC-UA communication technology was used to collect and store the information of CNC machine tools online, and the internal process information related to the wear of the machine tools was obtained. Based on such information and the corresponding wear information, a wear state recognition model is established by using a convolutional neural network. The feasibility and effectiveness of the proposed compliance control scheme and the performance of online monitoring of wear condition are analysed and verified by simulation experiments.

Enhancing Clustering Algorithm with Initial Centroids in Tool Wear Region Recognition

Autonomous manufacturing allows the system to distinguish between a mild, normal and total failure in tool condition. K-means clustering has become the most applied algorithm in discovering classes in an unsupervised scenario. Nevertheless, the algorithm is sensitive to the initial centroids giving various solution every time the system updating. Regular unsupervised K-means is refocused as semi-supervised Fixum K-means. It is embedded with a new tactic to recapture the K value and new initial seedings computation to kick off the system until it converges. Force components of cutting force $${F}_{c}$$ , thrust force $${F}_{t}$$ and perpendicular cutting force $${F}_{cn}$$ were extracted from Neo-MoMac cutting force measurement device. The analysis threshold represents a natural-sorted input vector as Z-rot coefficient (RZ) corresponds to the number of cutting accomplish a strong correlation (R2 = 0.8511) over wear evolution. The clustering system adopted a new calculation of initial centroids has successfully determined the three regions for only a single assignment and achieving the optimal distance squared through eight given data sets. It is conflicting with the standard K-means that return different clustering structure in each run, while K-means + + replicates several times to achieve minimum objective function. During the course, F-Km delivered robust and consistence clustering results of 85% accuracy over standard K-means and four times converges faster than K-means + + . The silhouette value average score is 0.8504 (highest score is 0.9207) of how well-distributed the resulting clusters. The clustering system has identified the tool to stop cutting at approximate VB = 0.213 mm before the tool condition enters the failure region of abnormal phase (VB < 0.250 mm).s The proposed system functioned effectively in clustering the data obtained from cutting tests performed within a reasonable range of wear stages. Precision and robustness analysis have proved F-km to score 100% attainment for clustering assignment output and replicability. In contrast, K-means scored 76.3% for precision and ranging from 5 to 33% for robustness. Whereas, K-means + + scored 33% for robustness and a higher chance of time complexity compared to F-km. F-Km is found to be more accurate, time savvy and robust than standard K-means and K-means + + . Therefore, the method can be reliably used for observing tool wear state recognition without training and equivocate traditional direct tool wear.

Tool Wear Online Monitoring Method Based on DT and SSAE-PHMM

Abstract The real-time requirements of tool wear states monitoring are getting higher and higher, at the same time, tool wear monitoring lacks a modeling data comprehensive carrier, which hinders its application in the actual machining process. In order to solve this problem, combining the high fidelity real-time behavior simulation characteristics of digital twin (DT) and the powerful data mining capabilities of artificial intelligence, an online tool wear monitoring method based on DT and Stack Sparse Auto-Encoder-parallel hidden Markov model (SSAE-PHMM) was proposed. First, a DT which can reflect the real state of the tool was established, and the tool wear state was predicted by visual display and analysis in the virtual space; Second, a tool wear state recognition model based on SSAE-PHMM was established, which can adaptively complete time domain feature extraction. And for each tool wear state, multiple HMM models were combined into a PHMM model to realize accurate recognition of tool wear state. PHMM overcome the defects of poor convergence and long training time of artificial neural network, and greatly improved the performance of classifier. Through the deep integration of DT and artificial intelligence, real-time data-driven tool wear qualitative and quantitative online monitoring was realized, and the effectiveness of this method was verified by experiments.

Wear state recognition of mechanical equipment based on stacked denoised auto-encoding network

Wear state recognition of mechanical equipment based on stacked denoised auto-encoding network

Tool wear state recognition based on gradient boosting decision tree and hybrid classification RBM

Machined surface quality and dimensional accuracy are significantly affected by tool wear in machining process. Tool wear state (TWS) recognition is highly desirable to realize automated machining process. In order to improve the accuracy of TWS recognition, this research develops a TWS recognition scheme using an indirect measurement method which selects signal features that are strongly correlated with tool wear to recognize TWS. Firstly, three time domain features are proposed, including dynamic time warping feature and two entropy features. The time, frequency, and time-frequency domain features of the vibration and force signals are extracted to form a feature set. Secondly, gradient boosting decision tree (GBDT) is adopted to select the optimal feature subset. Lastly, contrastive divergence (CD) and RMSspectral are used to train hybrid classification RBM (H-ClassRBM). The trained H-ClassRBM is used for TWS recognition. The PHM challenge 2010 data set is used to validate the proposed scheme. Experimental results show that the proposed features have better monotonicity and correlation than the classical features. Compared with CD and Adadelta, CD and Adagrad, and CD and stochastic gradient descent with momentum, the H-ClassRBM trained by CD and RMSspectral improves recognition accuracy by 1%, 2%, and 2%, respectively. Compared with feedforward neural network, probabilistic neural network, Gaussian kernel support vector machine, and H-ClassRBM, the proposed TWS recognition scheme improves recognition accuracy by 37%, 51%, 9%, and 8%, respectively. Therefore, the proposed TWS recognition scheme is beneficial in improving the recognition accuracy of TWS, and provides an effective guide for decision-making in the machining process.

Intelligent recognition of milling cutter wear state with cutting parameter independence based on deep learning of spindle current clutter signal

In the process of milling, tool wear directly affects the quality and accuracy of workpieces. Online recognition of milling cutter wear state has been and remains a growing interest in intelligent manufacturing to increase the machining efficiency and control the unqualified rate of workpieces. The effective value of spindle current can effectively characterize the wear state of milling cutter, but it will change along with machining process parameters, which are not suitable for the wear state recognition of milling cutter (WSRMC) under complex working conditions. We present LeNet-WSRMC network, a novel approach to recognize wear state of milling cutter based on the clutter signal of spindle current. The cutting vibration and tool wear are the main reasons for exciting the dynamic cutting force and the clutter signal of spindle current. In order to fully describe the generation mechanism of the clutter signal, we divide the wear state of milling cutter into four categories (i.e., normal wear, severe wear, abnormal vibration caused by tool wear, and abnormal vibration caused by improper selection of cutting parameter when the tool is sharp). LeNet-WSRMC network uses the deep convolutional neural network (DCNN) model to extract features from the spindle current clutter signal (SCCS) as the wear state of milling cutter classification index. A series of experiments with different cutting parameters and conditions are implemented to validate the effectiveness and generalization of our proposed methodology. The experimental results show that this method can realize the online accurate recognition of the wear state of milling cutter under the condition of complex working condition. This study lays a foundation for the prediction of the remaining life of the milling cutter under complex working conditions and the reasonable formulation of the replacement rules of the milling cutter.

DYNAMIC IDENTIFICATION OF WEAR STATE BASED ON NONLINEAR PARAMETERS

This paper presents a new methodology of wear state recognition by using fractal parameters, multifractal parameters and recurrence parameters. The relationship between these nonlinear parameters is analyzed. A nonlinear state point of worn surface is established by fractal dimension, average diagonal length and spectrum width. Further, a steady state sphere is obtained by the nonlinear state point and [Formula: see text]-means clustering algorithm. Results show that fractal, multifractal and recurrence parameters characterize the worn surface from different perspectives. They should be used simultaneously to comprehensively characterize the integral structures, partial structures and internal structures of worn surface. The proposed nonlinear state point shows a variation process of concentration–stabilization–separation during the wear process. The wear states can be identified effectively by the relationship between nonlinear state points and steady state sphere.

Tool wear state recognition based on GWO–SVM with feature selection of genetic algorithm

Tool wear is an important consideration for Computerized Numerical Control (CNC) machine tools as it directly affects machining precision. To realize the online recognition of tool wear degree, this research develops a tool wear monitoring system using an indirect measurement method which selects signal characteristics that are strongly correlated with tool wear to recognize tool wear status. The system combines support vector machine (SVM) and genetic algorithm (GA) to establish a nonlinear mapping relationship between a sample of cutting force sensor signal and tool wear level. The cutting force signal is extracted using time domain statistics, frequency domain analysis, and wavelet packet decomposition. GA is employed to select the sensitive features which have a high correlation with tool wear states. SVM is also applied to obtain the state recognition results of tool wear. The gray wolf optimization (GWO) algorithm is used to optimize the SVM parameters and to improve prediction accuracy and reduce internal parameters’ adjustment time. A milling experiment on AISI 1045 steel showed that when comparing with SVM optimized by commonly used optimization algorithms (grid search, particle swarm optimization, and GA), the proposed tool wear monitoring system can accurately reflect the degree of tool wear and achieves strong generalizability. A set of vibration signals are adopted to verify the presented research. Results show that the proposed tool wear monitoring system is robust.

Combining translation-invariant wavelet frames and convolutional neural network for intelligent tool wear state identification

On-machine monitoring of tool wear in machining processes has found its importance to reduce equipment downtime and reduce tooling costs. As the tool wears out gradually, the contact state of the cutting edge and the workpiece changes, which has a significant influence on the vibration state of the spindle. The performance of traditional intelligent fault diagnosis methods depend on feature extraction of dynamic signals, which requires expert knowledge and human labor. Recently, deep learning algorithms have been applied widely in machine health monitoring. In this paper, we present a novel intelligent technique for tool wear state recognition using machine spindle vibration signals. The proposed technique combines derived wavelet frames (DWFs) and convolutional neural network (CNN). Constructed based on dual tree wavelets, DWF are equipped with merits of centralized multiresolution and nearly translation-invariance. In this method, DWFs are employed to decompose the original signal into frequency bands of different bandwidths and different center frequencies, which are more pronounced as the tool wears. Further, the reconstructed sub-signals are stacked into a 2-D signal matrix to match the structure of 2-D CNN while retaining more dynamic information. The 2-D convolutional neural network is utilized to automatically recognize features from the multiscale 2-D signal matrix. End-milling experiments were performed on a S45C steel workpiece at different machining parameters. The experiment results of the recognition for tool wear state show the feasibility and effectiveness of the proposed method.

Milling Tool Wear State Recognition by Vibration Signal Using a Stacked Generalization Ensemble Model

Milling tool wear state recognition plays an important role in controlling the quality of milled parts and reducing machine tool downtime. However, the characteristics of milling process limit the accuracy and stability of tool condition monitoring (TCM) employing vibration signals. To improve this problem, this paper explores the use of vibration signals as sensing approach for recognizing tool wear states during milling operation by using the stacked generalization (SG) ensemble model. In this study, vibration signals collected during the milling process are analyzed through the time domain, frequency domain, and time‐frequency domain to extract signal features. The support vector machine recursive feature elimination (SVM‐RFE) algorithm is used to select the main features which are most relevant to tool wear states. The SG ensemble model based on SVM, decision tree (DT), naive Bayes (NB), and SG ensemble strategy is constructed to recognize tool wear states. The proposed method is experimental verified, and the results show that the recognition accuracy of the established SG ensemble model is 98.74% and the overall G‐mean and AUC evaluation value of the model is 0.98 and 0.98, respectively. In addition, compared with other ensemble models and single models, the SG ensemble model based on vibration signals has better recognition accuracy and stability than other models.

Dynamic States Recognition of Friction Noise in the Wear Process Based on Moving Cut Data-Approximate Entropy

Wear experiments are performed to explore dynamic states changes of friction noise signals. A new characteristic parameter, moving cut data-approximate entropy (MC-ApEn), is adopted to quantitatively recognize dynamic states. Additionally, determinism (DET), one key parameter of recurrence quantification analysis, is applied to verify the reliability of recognition results of MC-ApEn. Results illustrate that MC-ApEn of friction noise has distinct changes in different wear processes, and it can accurately detect abrupt change points of dynamic states for friction noise. Furthermore, DET of friction noise rapidly declines first, then fluctuates around a small value, and finally increases sharply, which just corresponds to the evolution process of MC-ApEn. So, the reliability of wear state recognition on the basis of MC-ApEn can be confirmed. It makes it possible to accurately and reliably recognize wear states of friction pairs based on MC-ApEn.

Tool wear state recognition based on LS-SVM with the PSO algorithm

Tool wear state recognition based on LS-SVM with the PSO algorithm

Engine Wear State Recognition Based on Image Analysis and Genetic Algorithm

Engine Wear State Recognition Based on Image Analysis and Genetic Algorithm

Milling tool wear state recognition based on partitioning around medoids (PAM) clustering

This paper presents a partitioning around medoid (PAM)-based novel method to realize the recognition of the tool wear state in milling. In PAM, the representative objects called medoids are used to define clusters and average dissimilarities are applied to assess the medoids, which make PAM robust to outliers and therefore improve the clustering performance. Meanwhile, locality preserving projections (LPP) method is utilized to further increase the clustering accuracy by dimension reduction. To show the effectiveness of the proposed method, end milling experiment of Ti-6Al-4V alloy were carried out and the commonly used k-means and fuzzy c-means (FCM) algorithm are introduced to make a comparison with PAM algorithm by using five clustering evaluation indicators. The results show that PAM performs higher accuracy and robustness than the other two clustering algorithm.

Chaotic Characteristics and the Application of SVM in the Tool Wear State Recognition

Metal cutting process is a nonlinear system to obtain the tool wear state and chaos theory are introduced tool wear and feature extraction of acoustic emission signal analysis and classification of tool wear state and wear prediction based on support vector machine (SVM). First, optimal embedding dimension of the time delay of phase space reconstruction of nonlinear dynamic system, the chaotic attractor; secondly, three characteristics: correlation dimension, the largest Lyapunov exponent and the Kolmogorov is extracted from the AE signal denoising feature vector and construct the different wear conditions. Finally, the feature vector is fed into the support vector machine (SVM), and the tool wear condition is classified. Research shows that: the cutting tool wear acoustic emission signal possesses the characteristics of chaos, chaotic characteristic parameters and tool wear status has intrinsic relationship; combined with chaos theory and support vector machine (SVM), can be very good to achieve the tool wear state recognition and prediction.

Research on Cutting Tool Wear Based on Fractional Brownian Motion

Cutting tool wear is a very complex process. Various factors have a direct or indirect effect on cutting tool wear, resulting in uncertainty, so it is difficult for experimental data and result to have good stability. However, Vibration analysis is a very important means for condition monitoring and fault diagnosis. This paper aims to study the methods of tool vibration signal processing, pattern recognition and trend prediction. Collected on tool vibration signal at different times, wavelet noise reduction is used to pretreat the vibration signals. Then, for the self-similar vibration signals, we propose the fractional Brownian motion (FBM) theory with long-range dependence (LRD). Combined with Wigner-Ville spectrum, characteristic parameter can be extracted, so the cutting tool wear state can be determined according to fractal dimension and average slope of the fitting curve of the logarithm power spectrum. Finally, we use FBM model to predict the trend of tool vibration signals. Experiments show that the methods have a good effect on tool wear state recognition and trend prediction.

Tool wear state recognition based on linear chain conditional random field model

Tool condition monitoring (TCM) system is paramount for guaranteeing the quality of workpiece and improving the efficiency of the machining process. To overcome the shortcomings of Hidden Markov Model (HMM) and improve the accuracy of tool wear recognition, a linear chain conditional random field (CRF) model is presented. As a global conditional probability model, the main characteristic of this method is that the estimation of the model parameters depends not only on the current feature vectors but also on the context information in the training data. Therefore, it can depict the interrelationship between the feature vectors and the tool wear states accurately. To test the effectiveness of the proposed method, acoustic emission data are collected under four kinds of tool wear state and seven statistical features are selected to realize the tool wear classification by using CRF and hidden Markov model (HMM) based pattern recognition method respectively. Moreover, k-fold cross validation method is utilized to estimate the generation error accurately. The analysis and comparison under different folds schemes show that the CRF model is more accurate for the classification of the tool wear state. Moreover, the stability and the training speed of the CRF classifier outperform the HMM model. This method casts some new lights on the tool wear monitoring especially in the real industrial environment.

Wear state recognition of drills based on K-means cluster and radial basis function neural network

Drill wear not only affects the surface smoothness of the hole, but also influences the life of the drill. Drill wear state recognition is important in the manufacturing process, which consists of two steps: first, decomposing cutting torque components from the original signals by wavelet packet decomposition (WPD); second, extracting wavelet coefficients of different wear states (i.e., slight, normal, or severe wear) with signal features adapting to Welch spectrum. Finally, monitoring and recognition of the feature vectors of cutting torque signal are performed by using the K-means cluster and radial basis function neural network (RBFNN). The experiments on different tool wears of the multivariable features reveal that the results of monitoring and recognition are significant and effective.