Deep Learning-Based Acoustic Emission Signal Recognition Method for Bolt Damage

Catastrophic Tool Failure Detection Based on Acoustic Emission Signal Analysis

MRI image classification of brain tumors is critical for accurate and early diagnosis. New developments in deep learning have revealed that inserting attention mechanisms into convolutional neural networks can greatly improve classification performance. This work assesses the effectiveness of Squeeze-and-Excitation (SE) and Convolutional Block Attention Module (CBAM) integrated with the ResNet50 model for brain tumor classification. The SE module recalibrates channel-wise feature responses, while CBAM enhances feature representation by combining channel and spatial attention. The SE and CBAM are combined with ResNet50, greatly increasing classification accuracy, precision, and recall when compared to the basic model, according to experimental results on MRI datasets. The suggested model employs attention mechanisms to focus valuable information selectively and suppress irrelevant information. The experiments on publicly available brain tumor datasets display great improvements over basic CNN models, with precision, recall, accuracy, as well as F1 score at 0.9914, 0.9903, 0.9945, and 0.9908, respectively. From these results, the importance of attention mechanisms in deep learning models for medical imaging is highlighted, which suggests that SE and CBAM modules can be available as more dependable and effective instruments for brain tumor classification in clinical settings. Future studies should investigate transformer-based and hybrid attention techniques to enhance automated brain tumor categorization.

There is considerable interest in techniques for ensuring the quality of the laser joining process. One approach to quality monitoring is through analysis of acoustic and optical emission signals acquired during welding. Preliminary analyses confirm that these signals fluctuate chaotically, suggesting that there may be benefits to a nonlinear signal processing approach. Lyapunov exponents of cleaned signals were calculated to quantify the nonlinear nature of the process. Inspection of Lyapunov exponents calculated from optical and acoustic emission signals corresponding to a variety of weld penetration conditions revealed that no sharp distinction could be found. The Lyapunov exponents were found to depend strongly on the power and travel speed settings of the process. This variation masked any variations due solely to weld penetration. There was some distinction in Lyapunov exponents noted between non-gap and gap butt joint welds. The Lyapunov exponent for butt welds with significant joint gap was higher than for non-gapped welds. It was concluded that monitoring of weld gap might be possible using the Lyapunov exponent of the optical and/or acoustic signals.There is considerable interest in techniques for ensuring the quality of the laser joining process. One approach to quality monitoring is through analysis of acoustic and optical emission signals acquired during welding. Preliminary analyses confirm that these signals fluctuate chaotically, suggesting that there may be benefits to a nonlinear signal processing approach. Lyapunov exponents of cleaned signals were calculated to quantify the nonlinear nature of the process. Inspection of Lyapunov exponents calculated from optical and acoustic emission signals corresponding to a variety of weld penetration conditions revealed that no sharp distinction could be found. The Lyapunov exponents were found to depend strongly on the power and travel speed settings of the process. This variation masked any variations due solely to weld penetration. There was some distinction in Lyapunov exponents noted between non-gap and gap butt joint welds. The Lyapunov exponent for butt welds with significant joint gap was higher...

Acoustic emission signal analysis during chip formation process in high speed machining of 7050-T7451 aluminum alloy and Inconel 718 superalloy

AE (acoustic emission) signals generated in laser welding were studied using wavelet time-frequency signal analysis methods. The AE signals were decomposed into a series of discrete sequences distributed over different frequency bands through wavelet transform. Further data processing was then carried out after selecting several decomposition sequences consistent with specific research goals. More detailed information of the welding process can be revealed through analyzing the time-frequency properties of the AE signals than can be obtained through statistical analysis or simple magnitude measurements. There is a significant difference between the AE signal from the desired deep-penetration weld and that which is acquired when there are problems with misalignment and excessive joint gap. The intensity of low-frequency (<781 Hz) components of the AE signals is dramatically reduced when welding defects occur. The same phenomena are found in the oscillation amplitude of frequency components between 781 and 1562 Hz. Based on the detailed information obtained from AE signals via wavelet analysis, a signal intensity moving average (IMA) curve is defined in this article. The curve is an effective tool for recognizing transitions of welding states and identifying defects quickly and reliably. The potential application of the IMA curve is also described in this article.

An improved mathematical model describing acoustic emission (AE) signals generated by different types of partial discharges (PD) that occur in electric power transformer insulation system is presented in the paper. AE signals are analyzed within the AE method as applied for power transformer failure detection due to occurrence of PD. There are several types of basic defects, which are characterized by different types of PD. The mathematical model presented here is crucial for numerical analyses and simulations, where it acts as the function describing the acoustic source in an acoustic model of power transformer insulation system. The regression procedure was performed based on empirical AE signals, registered in a laboratory experiment. The AE signals are described by a mathematical model being a multi-parameter function, which involve both the time domain and the frequency domain. Goodness of the model was evaluated based on analysis of 480 data samples in the time, frequency and time-frequency domains. Also coherence between the registered and modeled signals was calculated. It was stated that the improved model fits very well to the real data, although, due to high level of noise embodied in signals registered in experiments, the coherence values remain low. Moreover, analyses of the estimated data were performed and some example results are presented in this paper. Based on the achieved outcomes a collection of parameter values was prepared for each of the eight considered PD basic types. One can simple use it now in a numerical model for simulation of AE signal source generated by specified type of PD, what corresponds to a particular power transformer insulation system failure. Furthermore, the regression procedure presented in this paper can be easily transferred to any other types of AE sources including processes of compression, tension and cracking.

Acoustic emission (AE) is a non-destructive testing technique, and establishing correlations between AE signals and material damage modes is one of its primary challenges. However, it is difficult to identify damage modes in ceramic matrix composites (CMCs) due to AE signal attenuation occurring after propagation and complex damage modes. In this study, AE signals generated by the breakage of C and SiC fibers were monitored at different distances and angles on the C/SiC plate. The attenuation of energy and the frequency spectra were analyzed. The Mel-frequency cepstral coefficient (MFCC) method was used to analyze the waveform data of AE signals and extract MFCC features. To identify the damage caused by C and SiC fiber breakage, AE parameter features and MFCC features were selected as inputs, and a fully connected neural network was constructed to train a supervised pattern recognition model. The results show that the MFCC feature has higher recognition accuracy than the traditional feature when AE is used for damage identification.

Conducting tensile experiment with high pressure gas cylinder materials to analyze its acoustic emission (AE) signals, can realize the stretch features and AE signals regularity, which is very important to distinguish high pressure gas cylinder detection. The AE signals are divided into four different stages through the tensile experiment, which show that AE signals and theory analysis are fitted well and thus be used to distinguish materials defects.

The primary focus of this research article is to advance convolutional neural networks (CNNs) through the incorporation of Squeeze and Excitation (SE) and Convolutional Block Attention Module (CBAM) blocks, with an emphasis on recalibrating channel wise features. This approach highlights channel relationships, with the SE block seamlessly integrated into a pre-trained DenseNet CNN model to recalibrate channel wise feature responses by explicitly modeling channel interdependencies. The study demonstrates the construction of SENet and CBAM architectures with SE and CBAM blocks on pretrained DenseNet models using transfer learning, achieving exceptional generalization on a cataract image dataset. The proposed architectures utilize layers of compressed network blocks that can establish direct connections with preceding layers through adjustable weight modifications. Comprehensive evaluations indicate that the DenseSE and DenseCBAM models both achieve 99% testing accuracy and exhibit strong performance across various metrics. Notably, DenseNet169 with SE and CBAM blocks attained 99% precision, surpassing the baseline model’s 95% precision. Furthermore, the Grad CAM technique enhances cataract disease identification, with the DenseCBAM and DenseSE models generating detailed heatmaps that support healthcare professionals in making swift and accurate retinal diagnoses.

Frequency as a key parameter in discriminating the failure types of thermal barrier coatings: Cluster analysis of acoustic emission signals

The aim of the paper is focused on the analysis of the mechanical properties of the concrete specimens with plasticizer at three point bending test by the signal analysis of the acoustic emission signal. The evaluations were compared the measurement and the results obtained with theoretical presumptions. The Joint Time Frequency Analysis applied on measurement data and its evaluation is described. It is well known that the Acoustic Emission Method is a very sensitive method to determine active cracks into structure. However, evaluation of acoustic emission signals is very difficult. A non-traditional method was used to signal analysis of burst acoustic emission signals recorded during three point bending test.

Parameter analysis is a traditional and widely used method for the analysis of acoustic emission (AE) signals due to its simplicity and easy realisation. However, this approach has limitations and is not suitable for detailed analysis of the signals, and waveform-based methods are becoming attractive for the analysis of AE signals. This paper uses waveform-based analysis methods to investigate the time-frequency characteristics of AE signals arising from defects in rolling element bearings. This includes the composition of the signals, the generation mechanism of the defects and the influence of the operational condition of the bearing on these characteristics. The wavelet scalogram of the continuous wavelet transform is used to perform the qualitative analysis and the discrete wavelet transform is used for the quantitative analysis. The results demonstrate that the operational condition of the bearing will influence the time domain, the frequency domain and the time-frequency domain characteristics of the AE signals in various ways. The results also prove that the wavelet transform, and especially the wavelet scalogram, has great potential for the waveform-based analysis of AE signals from defects.

Monitoring the condition of the engine compression ring in an engine operation is very important since it affects the engine performance. One of the most promising ring wear monitoring methods is based on the analysis of acoustic emission (AE) signals, which is an extremely powerful technique that can be deployed in a wide range of applications of non-destructive testing [Vallen Systeme (2000)]. This technique is already used for monitoring tool wear almost in all machining operations, but in this study, the AE signals were applied for monitoring ring wear in internal combustion (IC) engines. The AE signals generated in the ring sliding zone are very sensitive for correlation with ring wear, which in turn affects ring performance. This study was carried out with a single compression ring mounted on the piston. The AE signals were analyzed by considering signal parameters such as ring down count and RMS voltage [Krzysztof Jemielniak (2000) J Mater Process Technol 4752:1–6]. Analysis showed that the AE signal technique is applicable for ring wear monitoring in IC engines.

Background: The accurate classification between malignant and benign breast lesions detected on mammograms is a crucial but difficult challenge for reducing false-positive recall rates and improving the efficacy of breast cancer screening. Objective: This study aims to optimize a new deep transfer learning model by implementing a novel attention mechanism in order to improve the accuracy of breast lesion classification. Methods: ResNet50 is selected as the base model to develop a new deep transfer learning model. To enhance the accuracy of breast lesion classification, we propose adding a convolutional block attention module (CBAM) to the standard ResNet50 model and optimizing a new model for this task. We assembled a large dataset with 4280 mammograms depicting suspicious soft-tissue mass-type lesions. A region of interest (ROI) is extracted from each image based on lesion center. Among them, 2480 and 1800 ROIs depict verified benign and malignant lesions, respectively. The image dataset is randomly split into two subsets with a ratio of 9:1 five times to train and test two ResNet50 models with and without using CBAM. Results: Using the area under ROC curve (AUC) as an evaluation index, the new CBAM-based ResNet50 model yields AUC = 0.866 ± 0.015, which is significantly higher than that obtained by the standard ResNet50 model (AUC = 0.772 ± 0.008) (p < 0.01). Conclusion: This study demonstrates that although deep transfer learning technology attracted broad research interest in medical-imaging informatic fields, adding a new attention mechanism to optimize deep transfer learning models for specific application tasks can play an important role in further improving model performances.

Wear mechanism of a cutting tool is highly complex in that the processes of tool wear results from interacting effect of machining configurations. Various output generated by the study and analysis of each tool is extremely useful in analyzing the tool characteristics in general and to make efforts to obtain the estimated tool life in particular. The gradual process of tool wear has adverse influence on the quality of the surface generated and on the design specifications in the work piece dimensions and geometry, and causes, at the worst case, machine breakdown. Advanced manufacturing demands proper use of the right tool and emphasizes the need to check the wear rate. A scientific method of obtaining conditions for an optimal machining process with proper tools and control of machining parameters is essential in the present day manufacturing processes. Many problems that affect optimization are related to the diminished machine performance caused by worn out tools. One of the indirect methods of tool wear analysis and monitoring is based on the acoustic emission (AE) signals. The generation of the AE signals directly in the cutting zone makes them very sensitive to changes in the cutting process and provides a means of evaluating the wear of cutting tools. Wear parameters obtained in the process are analyzed with the output generated by using Multi Layer Perceptron (MLP) based back propagation technique and Adaptive Neuro Fuzzy Interference System (ANFIS). The results obtained from these methods are correlated for the actual and predicted wear. Experiments have been conducted on EN8 and, EN24 using Uncoated Carbide, Coated carbide and Ceramic inserts (Kennametal, India make) on a high speed lathe for the most appropriate cutting conditions. The AE signal analysis (considering signal parameters such as, ring down count (RDC), rise time (RTT), event duration (ED) and energy (EG). Flank wear in tools and corresponding cutting forces for each of the trials are measured and are correlated for various combinations of tools and materials of work piece.