Machine learning method for damage classification based on acoustic emission waveform analysis in composite lap bar

The goal of this study is to classify Acoustic Emission (AE) signals based on their characteristics into several fracture types. AE method allows us to classify the damage process of CFRP with machine learning. However, unsupervised learning was often applied for classification which required manually matching the resulting clusters to specific damage types. In contrast, supervised learning to AE signals can automatically determine the damage types. A supervised learning technique uses many AE signals corresponding to each damage mode (supervised signals) in advance. In this study, supervised signals for Mode I and Mode II delamination were extracted from AE signals generated during Double Cantilever Beam (DCB) and Short Beam (SB) tests, respectively. While these delamination modes are dominant in each test, AE signals from other damage types, such as matrix cracking and fiber breakage, are also present. To account for this, additional AE data were obtained from tensile and three-point bending tests, where such damage is dominant. Finally, principal component analysis (PCA) was applied to visualize the distribution of AE signals. The results showed that AE signals from DCB and SB formed distinct clusters, while some overlapped with clusters from tensile and three-point bending signals. This suggests that by excluding overlapping clusters, it is possible to extract features that are more specific to Mode I and Mode II delamination.

To study the deformation and damage behaviors of three-dimensional (3D) woven composites, uniaxial tensile tests of the composites were conducted and real-time acoustic emission (AE) signals and speckle images were simultaneously obtained. A k-means clustering algorithm combined with principal component analysis (PCA) was used to analyze the AE signals, and the deformation fields on the surface of composite specimens were measured by the digital image correlation (DIC) method. It was found that the corresponding AE signals were divided into three typical clusters. On the basis of the frequency range of each type, the relationships between the resulting clusters and damage mechanisms were established. The results showed that different damage mechanisms in 3D woven composites are well determined by the frequency rather than the peak amplitude. The characteristics of AE signals such as the high frequency, the medium frequency, and most of the low frequency are associated with fiber damage, fiber/matrix debonding, and matrix cracking, respectively. The change in the displacement distribution can effectively enable visual inspection of damage accumulation in composites.

The failure progression characteristics of adhesively bonded Carbon Fiber Reinforced Polymer (CFRP) composites are investigated using Acoustic Emission (AE) technique. Different failure progression modes such as matrix cracking, fiber breakage, delamination and through-thickness crack growth releases AE waveforms in different frequency domains. The characteristic features of these different AE waveforms are studied in Mel Scale, which is a perpetual frequency scale of average human hearing frequency. The recurring noise in the recorded waveforms has been identified more efficiently when the waveforms are analysed in Mel Scale. The recorded AE signals from the adhesively bonded CFRP under static tensile loading are stretched to match the Mel filter banks. The sampling rate of the recorded signal is adjusted from 1 MHz to 20 kHz. Following that, the Mel spectrogram and its cepstral coefficients are used for identifying the different failure modes from which the AE signals are generated. A comprehensive comparison of the AE analysis in Mel scale with conventional waveform processing techniques such as Fast Fourier Transform (FFT), Continuous Wavelet Transform (CWT), Wavelet Packet Transform (WPT) and Hilbert-Huang Transform (HHT) has been made. The advantages and further applications of Mel Scale over traditional waveform processing techniques in defining the failure modes in the composites are also discussed.

A novel machine learning-based model is introduced in this research paper to forecast tool wear using acoustic emission (AE) signals. Adaptive boosting (AdaBoost) and a sophisticated feature engineering strategy are employed by the model to enhance the precision of its predictions. The proposed model, Machine Learning Tool Wear Prediction (MLTWP), analyzes AE signals generated during machining operations to distinguish between healthy and worn-out tool conditions with remarkable accuracy. The crux of our approach consists of meticulously eliminating and enhancing the temporal and spectral characteristics of the AE signals. We employ the Kolmogorov-Smirnov test to identify the most valuable classification features. We implemented AdaBoost with the objective of progressively enhancing a set of weak classifiers’ ability to identify instances that were incorrectly classified in previous iterations. Utilizing this method increases the model’s sensitivity to minute variations in tool wear conditions and its overall classification precision. The MLTWP model underwent extensive testing on a benchmark data set comprising 25,304 AE signal records from cutting mill tools, using a training tool split of 9,989 worn (positive) and 8,990 benign (negative) instances. The results of our experiments, validated through four-fold cross-validation, indicate that the MLTWP model exhibits superior performance compared to the existing Tool Wear Prediction using Acoustic Emission Signals (TWPAE) model. To provide greater specificity, the MLTWP exhibited the following metrics on average: precision (92.2%), specificity (91.38%), sensitivity (90.42%), accuracy (90.9%), and MCC (81.72%). The fact that these metrics exhibit significant improvement over those of the TWPAE model demonstrates that our method of feature engineering and adaptive boosting is effective at precisely predicting tool wear. This research not only advances the existing understanding of tool wear prediction but also establishes a robust framework for the implementation of machine learning in manufacturing predictive maintenance.

Purpose Low-velocity impacts may cause barely visible internal damage in carbon fiber-reinforced polymer (CFRP) laminates, which can significantly reduce the load-carrying capacity of Type IV hydrogen storage cylinders. This study aims to propose an acoustic-emission (AE) sensing framework to quantitatively evaluate low-velocity impact damage in CFRP laminates relevant to Type IV cylinder structures, and to establish severity grading criteria based on an interpretable scalar indicator. Design/methodology/approach Drop-weight impact tests were conducted from 10–80 J while AE signals were continuously recorded. AE features were extracted using empirical mode decomposition (EMD) and principal component analysis (PCA). A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) classifier was trained to identify damage-related AE signals, and wavelet entropy was calculated to quantify impact damage severity. Nonlinear ultrasonic testing was additionally used to validate damage progression. Findings The CNN–LSTM model achieved 93.3% classification accuracy for AE signals associated with matrix cracking, delamination and fiber fracture. Wavelet entropy showed strong correlation with impact damage severity and identified three damage stages. Two critical thresholds were found: wavelet entropy 0.275 (transition to moderate damage) and 0.55 (transition to severe damage), corresponding to impact energies of 40 and 60 J, respectively. Originality/value This study links AE signal characteristics to impact damage severity through an interpretable scalar metric (wavelet entropy) and provides quantitative thresholds (0.275 and 0.55) for three-stage damage grading. It further combines EMD–PCA feature extraction with a CNN–LSTM classifier (93.3% accuracy) to identify damage-related AE signals, and validates damage progression using nonlinear ultrasonic testing. The framework offers a practical sensing strategy for composite pressure-vessel monitoring.

Damage assessment studies in CFRP composite laminate with cut-out subjected to in-plane shear loading

In this work, a combination of an acoustic emission (AE) technique and a machine learning (ML) algorithm (Random Forest (RF) and Gradient Boosting Regressor (GBR)) is developed to characterize the particle size distribution in gas-solid fluidized bed reactors. A theoretical approach to explain the generation of acoustic emission signal in gas-solid flows is presented. An AE signal is generated in gas-solid fluidized beds due to the collision and friction between fluidized particles as well as between particles and the bed inner wall. The generated AE signal is in the form of an elastic wave with frequencies >100 KHz and it propagates through the gas-solid mixture. An inversion algorithm is used to extract the information about the particle size starting from the energy of the AE signal. The advantages of this AE technique are that it is a cheap, sensitive, non-intrusive, radiation-free, suitable for on-line measurements. Combining this AE technique with ML algorithms is beneficial for applications to industrial settings, reducing the cost of signal post-processing. Experiments were conducted in a pseudo-2D flat fluidized bed with four glass bead samples, with sizes ranging from 100 μm to 710 μm. AE signals were recorded with a sampling frequency of 5 MHz. The AE signal post-processing and data preparation for the ML process are explained. For the ML process, the AE frequency, AE energy and particle collision velocity data sets were divided into training (60%), cross-validation (20%) and test sets (20%). Two ensemble ML approaches, namely Random Forest and Gradient Boosting Regressor, are applied to predict particle sizes based on the AE signal features. The combination of these two models results in a coefficient of determination (R2) value greater than 0.9504.

Hand lay-up Fibre Reinforced Polymers (FRP), made of either carbon or flax textiles, were considered in this study and were subjected to water immersion at 40 ℃ over 24 weeks. Tensile tests were carried out periodically to assess the mechanical properties of the composites and Acoustic Emission (AE) monitoring was simultaneously realized to characterize the different microstructural damage mechanisms during loading. Maximum force per unit width, stiffness and ultimate strain were determined based on tensile testing data whereas frequency features and number of events were extracted and cumulated from AE signals. The obtained results showed that the mechanical properties of composites were significantly reduced meanwhile a modification of failure mode over 24 weeks was observed. Moreover AE signals analysis allowed a quantification of the level of damage. It also provided a discrimination of the different damage mechanisms and clarified how each mechanism contributed to failure.KeywordsHygrothermal ageingDurabilityCarbon fiber reinforced polymerCFRPFlax fiber reinforced polymerFFRPAcoustic emissionTensile behaviorDamage clustering

In this study, we experimentally investigate the impact damage and failure mechanisms of foam-core sandwich composites with carbon fiber reinforced polymer (CFRP) and/or glass fiber reinforced polymer (GFRP) face sheets. Specimens are conditioned and impacted over a wide temperature range (from room temperature of 23°C down to -70°C) using Instron CEAST 9350 impact test machine. This work also explores the use of hybrid composite face sheets, with various configurations of layering CFRP and GFRP in the face sheets of foam-core sandwich composites. Results show that exposure to low temperature causes more severe damage, particularly in specimens with CFRP face sheets, due to their extreme brittleness. GFRP face sheet specimens are able to achieve larger deformation and absorb greater impact energy even at low temperature conditions. There is a significant benefit in using hybrid configurations, by striking a balance between brittle (CFRP) and ductile (GFRP) materials and harnessing their mechanical advantages of high strength and high fracture toughness, respectively. Results also demonstrate that hybridization is a solution to reduce temperature effects in impact deformation. Careful placement of CFRP and GFRP in layup order can harness the best impact performance. X-ray micro-computed tomography (μCT) images reveal substantial delamination at the interfaces between CFRP and GFRP layers. It is also observed that CFRP layers fail by brittle fracture, while delamination is the major damage mode within GFRP layers. Other complex failure mechanisms in the composite face sheets (such as matrix crack and fiber breakage) and foam core (core crushing, core shearing and interfacial debonding) are also presented.

Assessment of damage evolution of concrete beams strengthened with BFRP sheets with acoustic emission and unsupervised machine learning

One of the most commonly used techniques to strengthen steel reinforced concrete structures is the application of externally bonded patches in the form of carbon fiber reinforced polymers (CFRP) or recently, textile reinforced cements (TRC). These external patches undertake the tensile stress of bending constraining concrete cracking. Development of full-field inspection methodologies for fracture monitoring are important since the reinforcing layers are not transparent, hindering visual observation of the material condition underneath. In the present study acoustic emission (AE) and digital image correlation (DIC) are applied during four-point bending tests of large beams to follow the damage accumulation. AE helps to determine the onset of fracture as well as the different damage mechanisms through the registered shifts in AE rate, location of active sources and change in waveform parameters. The effect of wave propagation distance, which in large components and in-situ can well mask the original information as emitted by the fracture incidents is also discussed. Simultaneously, crucial information is supplied by DIC concerning the moments of stress release of the patches due to debonding, benchmarking the trends monitored by AE. From the point of view of mechanics, conclusions on the reinforcing contribution of the different repair methodologies are also drawn.

Strain localization peculiarities and distribution of acoustic emission sources in rock samples tested by uniaxial compression and exposed to electric pulses

The aim of this study is to characterize different damage mechanisms in an Aluminum (Al)/TiC particulate composite using acoustic emission (AE) monitoring, microstructure-based peridynamic (PD) modeling, and scanning electron microscope (SEM) observations. The Al/TiC composite with a 20% volume fraction of TiC particles was produced using hot extrusion. Several representative volume elements (RVEs) were extracted from the SEM images by image processing, and the effect of particle morphology on the composite behavior was investigated. Rise time, energy, peak amplitude, and duration of the AE signals were utilized for clustering and relating the AE signals to deformation and damage mechanisms. PD modeling results and SEM pictures revealed that the damage mechanisms were initiated in the narrow TiC particles, Al matrix between close particles, and interface of the particle/matrix. The distribution of the particles had more effect on the damage initiation pattern than elastoplastic deformation before the damage initiation. Four sources of the AE signals were detected, i.e., plastic deformation of the Al, Al/Tic debonding, Al cracking, and fracture of the Tic particle. The AE signals of each mechanism were clustered by properties of waveform and by considering the predicted time interval of occurrence of each mechanism by PD modeling. PD model and AE technique predict the same sequence time of deformation and damage mechanisms, but the AE technique predicts the occurring time of each mechanism slightly earlier than the PD model.

Cluster analysis of acoustic emission signals for bending damage mechanism of flax fiber reinforced composites under hydrothermal aging

A multisignal nanosecond synchronous acquisition system to measure acoustic emission (AE) and electromagnetic radiation (EMR) generated during the process of loading and failure of coal and rock samples is established. The correlation between the energy of the AE and EMR signals and the loading stress of outburst coal-rock samples was studied, and the characteristics of the AE and EMR signals during the process of loading and fracturing the outburst coal and rock samples were analyzed. The results show that (1) before the failure of the outburst coal and rock samples, the fluctuation of the AE and EMR signals is the largest, with the same rising and falling trend, and the intensity is not strictly positively correlated, with the phenomenon of low EMR when the AE intensity is high; (2) the EMR and AE deviation degree and frequency exhibit a good response to coal and rock fracturing. The correlation between EMR and stress drop is stronger than that of AE, and the AE signal is richer than the EMR signal. The results show that it is feasible to develop combined AE and EMR early warning technology to improve the early forecasting accuracy of coal and gas outbursts.