Chatter Detection Research Articles

Milling chatter detection with WPD and power entropy for Ti-6Al-4V thin-walled parts based on multi-source signals fusion

Unstable cutting has a relatively large negative impact on the machining quality and efficiency of thin-walled parts. It can easily cause severe vibration of milling systems, resulting in poor surface roughness of workpieces. Although some promising signal processing methods, such as wavelet packet decomposition (WPD), have been applied to process nonlinear signals, few studies have paid attention to determining a wavelet basis and the number of decomposition layers in WPD. Parameters of WPD play a crucial role and they are often empirically determined. In this paper, an adaptive denoising model based on WPD and recursive least squares with a variable forgetting factor (RLSVFF) is firstly established, and the influence of the forgetting factor in the model on signal denoising are investigated. Then, a novel chatter detection method based on multi-source signals fusion using WPD and power entropy is presented. On this basis, an automated selection method based on a margin indicator and power is proposed for applications to WPD. And the influence of different parameters of WPD on a margin indicator and power are investigated. In order to improve the efficiency of different levels of signal acquisition, a method based on a stability lobe diagram (SLD) is used to design experimental parameters. Compared with traditional denoising models and chatter detection methods based on a single signal, simulation and experimental results show that the adaptive denoising model and chatter detection method based on multi-source signals fusion proposed in this paper can more reliably detect the occurrence of early chatter and different levels of chatter.

Study on Bandwidth Analyzed Adaptive Boosting Machine Tool Chatter Diagnosis System

This paper presents an Adaboost algorithm based cutting data analysis for chatter detection. This offline chatter analysis uses the vibration data collected by accelerometers attached to the spindle housing. A comparison of the accuracy achieved with support vector machine, Random Forest, 1D Convolutional Neural Networks and Multilayer Perceptron algorithm is also made. In this paper, the accelerometer data are transformed into bandwidth. Time-accelerometer and time-spectral bandwidth learning models are built in order to realize chatter detection and automated machine learning. A comparison of the models is made. The results of cross validation indicate that an accuracy of 98% is achieved, which is made possible by using the bandwidth signals that are transformed from accelerometer data. Experimental results show that applying the Adaboost algorithm to analyze the spectral data transformed from vibration signals and using them to detect chatters has higher reliability and accuracy compared to other algorithms and analyzing other transform signals.

In-process chatter detection in micro-milling using acoustic emission via machine learning classifiers

In-process chatter detection in micro-milling using acoustic emission via machine learning classifiers

Online grinding chatter detection based on minimum entropy deconvolution and autocorrelation function

Online grinding chatter detection based on minimum entropy deconvolution and autocorrelation function

Piezoelectric active damper for surface roughness improvement in hard turning processes

Surface roughness and profile accuracy on rolling or sliding surfaces are critical for the wear and fatigue of a component. A high roughness or poor profile accuracy results in higher friction and higher wear rate of the surface. One of the major factors affecting the surface roughness is chatter vibrations. This paper presents a novel design and development of an active damper for chatter suppression of hard turning processes using a piezoelectric actuator and strain signal for chatter detection. The active damper consists of a piezoelectric actuator with an embedded strain gauge for measuring the vibration displacement. In this work, the radial strain signal as a result of radial chatter vibrations from the strain gauge is used as a feedback signal to the actuator using a feedback controller. A fixture was designed using special stiffeners such that it limited the tool displacements in the tangential and feed directions to only 0.2 µm and 0.4 µm, respectively, while allowing for movement in the radial direction in the actuator’s force direction. The experimental results showed a significant suppression in chatter vibrations and improvement of surface roughness using the proposed active damper. A reduction of 83% in surface roughness (Ra) was observed during the cutting process. The details of the tool design, control design, hardware implementation and system validation are given hereinafter.

Online chatter detection in milling process based on fast iterative VMD and energy ratio difference

Chatter is normally detected by time–frequency analysis, among which VMD has a perfect theoretical basis. However, the algorithm directly used for chatter detection may have low accuracy and efficiency due to the initial parameters that are difficult to determine. Therefore, a novel method based on fast iterative variational mode decomposition (FI-VMD) is proposed to solve these issues. The simulation process shows FI-VMD is almost equivalent to the extraction accuracy of VMD, but when the energy of chatter signal is very small, FI-VMD is better than VMD. Moreover, since the initial frequency is the true frequency in the FI-VMD iteration process, its calculation efficiency is much faster than that of VMD. Furthermore, the proposed energy ratio difference has strong robustness in stable cutting condition. When chatter occurs, it is more sensitive to chatter than energy ratio. Compared with an optimized VMD, the proposed method is more suitable for online chatter detection.

Chatter detection in turning using machine learning and similarity measures of time series via dynamic time warping

Chatter detection from sensor signals has been an active field of research. While some success has been reported using several featurization tools and machine learning algorithms, existing methods have several drawbacks including the need for data pre-processing by an expert. In this paper, we present an alternative approach for chatter detection based on K-Nearest Neighbor (KNN) algorithm for classification and the Dynamic Time Warping (DTW) as a time series similarity measure. The used time series are the acceleration signals acquired from the tool holder in a series of turning experiments. Our results show that this approach achieves detection accuracies that can outperform existing methods, and it does not require data pre-processing. We compare our results to the traditional methods based on Wavelet Packet Transform (WPT) and the Ensemble Empirical Mode Decomposition (EEMD), as well as to the more recent Topological Data Analysis (TDA) based approach. We show that in two out of four cutting configurations our DTW-based approach is in the error range of the highest accuracy or attain the highest classification rate reaching in one case as high as 98% accuracy. Moreover, we combine the Approximate and Eliminate Search Algorithm (AESA) and parallel computing with the DTW-based approach to achieve chatter classification in less than 2 s thus making our approach applicable for online chatter detection.

Milling chatter monitoring based on sparse representation and image similarity measurement

It is well known that chatter is one of the main bottlenecks affecting the surface quality of workpieces and production efficiency during machining. In this paper, a new chatter monitoring methodology is proposed on the basis of sparse representation and image similarity measurement to recognise chatter in the manufacturing process. Due to its non-stationary nature with regard to chatter signals, variational mode decomposition (VMD) is utilised and timefrequency entropy (TFE) based on VMD is introduced to measure the complexity. Then, the overcomplete dictionary is pre-trained using the characteristic matrix image extracted from the multi-domain features, thus facilitating the description of chatter from various perspectives. Subsequently, the visualised sparse coefficient matrix is acquired from the trained dictionary and regarded as the reference image, in which detailed information can be obtained from the visualisation of the image. Next, an image similarity measurement method is applied to assess the similarity between the tested sparse coefficient image and the reference image, thereby considering the local and global quality maps such that a comprehensive index for chatter detection can be obtained to fuse the various features. Finally, to validate the proposed methodology, experimental chatter tests are conducted under different machining conditions. The results demonstrate that the chatter can be discriminated at the early stage of chatter development, thus leaving more time to take suppression measures.

In-process impulse response of milling to identify stability properties by signal processing

In this work we present a quantitative measurement-based method to describe the stability behaviour of a periodic dynamical system. In-process response for impulse during milling is analysed to provide operational stability prediction. The proposed method combines chatter detection techniques with the theory of time-periodic delay differential equations applied for general milling operations. Signal processing based on dynamic modal decomposition approach not only presents qualitative properties (like stability/instability) but also quantifies a measure of stability through the determination of the Floquet multipliers. The corresponding monodromy operator of the milling process is approximated from the measured system’s response during machining operation. By changing the technological parameters, the variation of the modulus of the Floquet multipliers can be monitored. The stability limit can precisely be interpolated with the unitary multiplier; furthermore, the stability limit can be extrapolated while the manufacturing parameters remain in the chatter-free region. The presented approach is validated by numerical results and laboratory tests.

Chatter detection in milling process based on the combination of wavelet packet transform and PSO-SVM

Chatter is one of the biggest unfavorable factors during the high speed machining process of a machine tool. It severely affects the surface finish and geometric accuracy of the workpiece. To address this obstacle and improve the quality and efficiency of products, it is significantly essential to detect chatter during machining. Therefore, a multi-feature recognition system for chatter detection on the basis of the fusion technology of wavelet packet transform (WPT) and particle swarm optimization support vector machine (PSO-SVM) was proposed in this paper. Firstly, the original vibration signals collected from the acceleration sensor were processed through wavelet packet transform (WPT). The noise and the irrelevant information were remarkably decreased. In addition, the wavelet packets containing chatter-emerging information were chosen and reconstructed. The fourteen time–frequency domain characteristics of the reconstructed vibration signal were calculated and chosen as the multi-feature vectors of chatter detection. Finally, to obtain the optimal radial basis function parameter g and penalty parameter C of the SVM prediction model, the optimization algorithms of k-fold cross-validation (k-CV), genetic algorithm (GA), and particle swarm optimization (PSO) were employed in optimizing the model parameters of SVM. It was indicated that the PSO-SVM improved obviously the accuracy of chatter recognition than the others. In addition, we applied the optimized SVM prediction model by PSO for detecting chatter state in end milling machining. Chatter recognition results indicated that the model accurately predicted the slight chatter state in advance.

Topological feature vectors for chatter detection in turning processes

Machining processes are most accurately described using complex dynamical systems that include nonlinearities, time delays and stochastic effects. Due to the nature of these models as well as the practical challenges which include time-varying parameters, the transition from numerical/analytical modeling of machining to the analysis of real cutting signals remains challenging. Some studies have focused on studying the signals of cutting processes using machine learning algorithms with the goal of identifying and predicting undesirable vibrations during machining referred to as chatter. These tools typically decompose the signal using Wavelet Packet Transforms (WPT) or Ensemble Empirical Mode Decomposition (EEMD). However, these methods require a significant overhead in identifying the feature vectors before a classifier can be trained. In this study, we present an alternative approach based on featurizing the time series of the cutting process using its topological features. We first embed the time series as a point cloud using Takens embedding. We then utilize Support Vector Machine, Logistic Regression, Random Forest and Gradient Boosting classifier combined with feature vectors derived from persistence diagrams, a tool from persistent homology, to encode chatter’s distinguishing characteristics. We present the results for several choices of the topological feature vectors, and we compare our results to the WPT and EEMD methods using experimental turning data. Our results show that in two out of four cutting configurations the Topological Data Analysis (TDA)-based features yield accuracies as high as 97%. We also show that combining Bézier curve approximation method and parallel computing can reduce runtime for persistence diagram computation of a single time series to less than a second thus making our approach suitable for online chatter detection.

An Evaluation Study of EMD, EEMD, and VMD For Chatter Detection in Milling

In modern machining processes, chatter is an inherent phenomenon that hinders efficiency, productivity, and automation. Numerous methods have been proposed using analytical, computational, and artificial intelligence methods to detect and avoid chatter during milling. The vibration signals generated during machining are of non-stationary and non-linearity nature. Hence solely time or frequency domain analysis are not adequate methods for chatter detection. This study investigates the performance of more advanced mode decomposition methods and compares them. Three decomposition methods, namely, empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), and variational mode decomposition (VMD), are used to decompose and identify chatter frequency bands. After decomposition, Hilbert-Huang transform (HHT) was applied for visualization. The comparative results indicate that EEMD or VMD decomposition methods performed better than EMD for intelligent chatter detection.

Exploring the Potential of Transfer Learning for Chatter Detection

Chatter detection and avoidance are indispensable for many industries that rely on the machining process. The physics-based analytical models and recently successful machine learning methods can provide solutions using data from a unique setting. When the primary conditions of machining alter, new data needs to be collected, and analysis/training should be revised. Unfortunately, data collection is time-consuming and expensive for all machine learning applications. Therefore, broader applications of these methods are usually hindered at high production rate machining shops. Transfer learning aims to attenuate this critical barrier of machine learning implementations by transferring knowledge generated from a source domain to a different but related domain. As the concept has immense potential as an accelerator for machine learning applications, it has many prospects in Industry 4.0 framework. This article provides an introduction to transfer learning and briefly overviews its categorizations. Afterward, its potential for chatter detection is explored, and potential strategies are exemplified. Recent studies in the literature within the strategies are briefly presented as well.

A Chatter Detection Method in Milling Based on Grey Wolf Optimization Vmd and Multi-Entropy Features

A Chatter Detection Method in Milling Based on Grey Wolf Optimization Vmd and Multi-Entropy Features

In-Process Chatter Detection Using Signal Analysis in Frequency and Time-Frequency Domain

All machining processes involve vibrations generated by structural sources such as a machine’s moving parts or by the interaction between cutting tools and work-pieces. Relative vibrations between the work-pieces and the cutting tool are the most relevant from the point of view of the regenerative chatter phenomenon. In fact, these vibrations can lead to a chip yregeneration effect, which results in unwanted consequences, rapidly degenerating towards a very poor quality of surface finishing or, in case of severe chatter conditions, to machine-tool or work-piece damage. In the past decades, two different approaches for chatter avoidance were proposed by the scientific community, and they are commonly referred to as Out-of-Process (OuP) and in-Process (iP) solutions. The OuP solutions are off-line approaches, which allow to properly set the working parameters before machining starts. Ip solutions are on-line techniques, which allow to dynamically change the working parameters during machining by using single or multiple sensors. By monitoring the machining process, iP algorithms try to keep the machining process in stable working conditions while keeping high productivity levels. This study dealt with a novel iP chatter-detection strategy based on the Power Spectral Density (PSD) analysis and on the Wavelet Packet Decomposition (WPD) of different sensor signals. The preliminary results demonstrate the stability and feasibility of proposed indicators for chatter detection in industrial application.

Early chatter detection in thin-walled workpiece milling process based on multi-synchrosqueezing transform and feature selection

The onset of chatter is extremely disadvantageous to machining process, which will induce poor surface quality, low processing efficiency and severe tool wear, especially in the milling of weakly rigid thin-walled workpieces, so it should be detected as early as possible to avoid further damage. To handle this issue, a multi-synchrosqueezing transform (MSST) based early chatter detection method as well as a novel feature selection algorithm is proposed in this paper to monitor the occurrence of chatter in thin-walled parts machining. MSST is firstly conducted on the sampled vibration signal to generate a time–frequency (TF) representation. The periodic components associated with spindle rotation are removed from the TF matrix and the chatter-related signal is recovered by multi-ridge detection. Then, features in time and frequency domains are extracted from the reconstructed chatter signal. A feature selection algorithm combining multi-feature distance evaluation technique (MFDET) and distance correlation analysis is subsequently presented to determine the optimal feature subset. Compared with other chatter indicators given in existing works, the selected compact feature set is more discriminative and has low redundancy. Furthermore, random forest (RF) is utilized to intelligently diagnose the machining stability. Several experiments under constant and variable conditions are performed to evaluate the performance of the proposed method in terms of early chatter detection. The experimental results demonstrate that the presented chatter monitoring method is highly efficient and robust, which can recognize the chatter onset timely and accurately for thin-walled parts milling under different machining conditions.

CHATTER DETECTION BASED ON A CONTACT SENSOR IN MILLING OPERATIONS

Chatter vibrations during machining operations are large-amplitude self-excited vibrations, which result in the loss of physical contact between the workpiece and the tool. This paper presents an electric contact sensor, which is used to detect the cutting and noncutting periods during the operation. The spectrum of the signal of the electric contact sensor is free from environmental noise and can be used to select the chatter frequencies as the boundary of stability is approached. The detection is automatic with the use of an analog phototransistor that helps the time-periodic sectioning of the measured signals. The presented approach makes chatter detection possible before harmful vibrations arise.

Detection of modulated chatter using moving average difference spectrum analysis

Chatter monitoring is an essential requirement for high-precision machining systems. Chatter vibration is characterized by amplitude modulation when the beat-frequency effect or geometric eccentricities exist, which hinders methods based on the transient characteristic index. Here, moving average difference spectrum analysis (MADSA) is proposed for detection of modulated chatter; the method can detect chatter frequency and track its amplitude from low signal-to-noise vibration signals early on. MADSA utilizes a moving average to smooth the amplitude trend of chatter frequency candidates obtained by the difference spectrum analysis, and the chatter frequency is detected using monotonicity analysis. MADSA was validated in both simulations and gear-milling machine experiments, and comparisons between wavelet packet-energy entropy method and MADSA are described, to illustrate the performance of the latter.

SB-LMD based online monitoring of tool chatter detection in milling process

In this present study, an experimental approach for detecting tool chatter in its early stages is proposed. Local mean decomposition based on cubic spline approach is invoked to analyse acquired machining sound data at varied cutting settings. The correlation coefficient and relative energy rule are used to identify the significant PFs. Furthermore, the characteristics of tool chatter are investigated by assessing six dimensional and six dimensionless indexes. Then, for every combination of characteristics, a threshold is defined to detect the aforementioned severity of chatter. At last, tool chatter online monitoring is carried out based on the obtained threshold.

EFFICIENT DETERMINATION OF STABILITY LOBE DIAGRAMS DEPLOYING AN AUTOMATED, DATA-BASED ONLINE NC PROGRAM ADAPTION

Process instabilities due to regenerative chatter pose significant limitations on the achievable material removal rates and thus on the profitability of machining operations. Stability lobe diagrams serve to exploit the maximum yet stable cutting depth and can be determined either analytically or experimentally. While analytical approaches suffer from inaccuracies because of the assumptions made for the specific models, experimental stability lobe diagrams require extensive cutting tests. Therefore, this paper introduces a new automated experimental method for determining stability lobe diagrams in milling with reduced effort regarding time. A closed-loop system is designed, containing a sensor-based online chatter detection along with a strategy to set parameters for subsequent cuts based on the stability boundaries known at each iteration. Both cuts with continuously increasing cutting depth and varied spindle speed are deployed to ensure fast detection of stability limits. The method is tested for a slot milling use case and the results are compared to a conventionally obtained stability lobe diagram yielding a significantly reduction in required time (-90 %) and resources (-67 %) whilst maintaining good accuracy. The reduced effort qualifies the proposed method as a tool to rapidly deliver maximum productive yet stable cutting parameters for optimization of existing or enhanced planning of new manufacturing processes.