Deep learning based characterization of nanoindentation induced acoustic events

Acoustic emission signals from concrete compression damage process are non-stationary and their arrival time is unknown. Therefore, the definition on the acoustic emission event in the time series is not clear. In this paper, the acoustic emission signals associated with concrete specimen under uniaxial compression are acquired by full-digital acoustic emission system. The overlapping phenomenon of acoustic emission signal is presented and discussed by the analysis of time interval between adjacent hits. Acoustic emission signals are classified into three categories according to their waveform characteristics. And the result shows that as the loading process continues, more and more amount of continuous type of acoustic emission signals appeared, indicating the acoustic emission signal does not carry the features of burst type, which will introduce obvious errors in the calculation of acoustic emission event. In contrast, the average signal level used in the analysis of continuous type of acoustic emission signal shows a good regularity with the damage process of concrete.

For the two-dimensional localization of acoustic emission events (AE), artificial intelligence (AI) was applied. AE events were generated by breaking pencil leads on a granite plate measuring with dimensions of 50 cm × 50 cm × 6 cm. A total of 128 measurements were carried out at each of the 81 positions on the top surface of the granite plate. The recorded signals with 16 channels were processed using continuous wavelet transformation (CWT). The resulting RGB images served as input data for a convolutional neural network (CNN). The transformed images of the measured AE signals were converted into a four-dimensional input by considering them as deeper input, allowing for improved data capture. The result was analysed using a binary classifier according to the one-versus-all method. However, due to limited computational capacity, processing the four-dimensional data posed a challenge, resulting in significant programming effort. The results of this investigation provide insights into the potential of using AI for the localization of AE events.

Influence of water activity on acoustic emission of flat extruded bread

According to stress damage acoustic emission signal of loading honeycomb sandwich composite, taking the destructive compression test. Analyzing mechanical properties of the honeycomb sandwich composite material compression, establishing the correlation diagram between load and acoustic emission signal; the distribution of each phase signal characteristic parameter is studied with statistics. Taking three times filtering based on each phase signal characteristics. With increasing the load under loading condition, the damage of composite material increases gradually. The whole process can be divided into four stages, the first stage, There is small amount of the acoustic emission signal and the signal is generally low; Secondly, the number of acoustic emission events are relatively small and stable with a small increasing in characterization of signal; Thirdly, acoustic emission events grows suddenly, the average energy released by it increases; Fourthly, there will be arising a large number of signal which characteristic parameter is large. The experimental results show that it is effective to analyze the material damage and its evolution based on the analysis of acoustic emission parameter in compression process. There are coincides between the actual fracture location and the signal distribution obtained after filtering which is based on the characteristics of each phase parameters. The paper can provide reference for damage forecast and health monitoring of honeycomb sandwich composite.

This paper presents an efficient deep learning framework for long-term monitoring of acoustic events from hydrophone big data. The large-scale noisy ONC (Ocean Networks Canada) data may contain rare acoustic events, which can be automatically recognized by utilizing a deep convolutional neural network. Few works have been reported in the area of deep learning for the recognition of different kinds of marine mammal calls which however is crucial for many applications such as marine navigation. In this proposed scheme, deep learning feature sets are adopted and processed by a support vector machione (SVM) classifier. The proposed method is tested with 28685 minutes of data, spanning a single year with 5573 whale calls/acoustic events, and using a human operator's annotations. The experimental results show that the average accuracy rate of recognition using deep feature learning are 98.69% (two-class) and 94.48% (multi-class), respectively, for the proposed recognition scheme, which outperforms the MFCC-based method.

SiC monofilament-reinforced copper is a potential heat sink material for the divertor of future fusion reactors, where heat loads up to 15 MW/m2 must be removed. The composite combines high mechanical strength (300 MPa at 300℃) with sufficient thermal conductivity (at least 200 W/mK) at temperatures up to 550℃. The bonding between fibres and matrix is essential for the mechanical behaviour. We investigated two different fibre types (SCS0 and SCS6, Specialty Materials Inc.) with different interlayers in a galvanic copper matrix. Tensile tests in combination with acoustic emission analysis showed the dependence of fracture behaviour of the composite material on the fibre type. The average ultimate tensile strength of composite samples with 20% fibre reinforcement was in the range of 600 MPa, with a Young's modulus of 160 GPa. The tensile tested composite with SCS6 fibres generated roughly twenty times the number of acoustic emission events in comparison with the SCS0 fibres composite. Acoustic emission events in tensile test specimens with SCS6 fibres are correlated with the failure of the carbon-enriched surface layer of the fibre. In the case of SCS0 fibre-reinforced specimens, acoustic emission events were only caused by fibre failure. Fracture area analysis after tensile tests showed the failure of the SCS6 fibres at the interface between the two outer carbon layers and no interface failure for the composite with SCS0 fibres.

In uniaxial compression tests performed on Aue granite cores (diameter 50 mm, length 100 mm), a steel loading plate was used to induce the formation of a discrete shear fracture. A zone of distributed microcracks surrounds the tip of the propagating fracture. This process zone is imaged by locating acoustic emission events using 12 piezoceramic sensors attached to the samples. Propagation velocity of the process zone is varied by using the rate of acoustic emissions to control the applied axial force. The resulting velocities range from 2 mm/s in displacement‐controlled tests to 2 μm/s in tests controlled by acoustic emission rate. Wave velocities and amplitudes are monitored during fault formation. P waves transmitted through the approaching process zone show a drop in amplitude of 26 dB, and ultrasonic velocities are reduced by 10%. The width of the process zone is ∼9 times the grain diameter inferred from acoustic data but is only 2 times the grain size from optical crack inspection. The process zone of fast propagating fractures is wider than for slow ones. The density of microcracks and acoustic emissions increases approaching the main fracture. Shear displacement scales linearly with fracture length. Fault plane solutions from acoustic events show similar orientation of nodal planes on both sides of the shear fracture. The ratio of the process zone width to the fault length in Aue granite ranges from 0.01 to 0.1 inferred from crack data and acoustic emissions, respectively. The fracture surface energy is estimated from microstructure analysis to be ∼2 J. A lower bound estimate for the energy dissipated by acoustic events is 0.1 J.

Aluminium Alloy AA 2219 is the principal metal for the production of propellant reservoir used in launch vehicles. The propellant tanks are often proof tested with acoustic emission technique (AET) to ensure its health. Acoustic Emission Testing during the structural health monitoring and proof testing is complex and unrealistic occasionally as it is performed in the noisy environment. Identification of signature corresponds to crack and its extraction from noise signatures is a major challenge in AET. Wavelet Packet Transform is an efficient mathematical tool for the analysis of AE signals. This paper recommends a novel combination of normalized cross correlation, Wavelet Packet Transform and Continuous wavelet transform to detect and extract the event related to failure. Experiments were carried out on AA 2219 tensile coupons at different threshold conditions. The recorded AE hits contain signals related to different events such as atmospheric noise, rubbing noise and other noise signals along with the signals from cracks. By applying the fine tuned wavelet packet transform technique in combination with CWT, the extraction of denoised single event related to crack was executed. Based on the frequency and the wavelet coefficient the crack related hits and the noisy hits are categorized.

The subject matter of this presentation will refer to the preliminary development of a non-destructive and non-invasive diagnostic method using low-frequency acoustic emissions to characterize and diagnose joint and muscle disorders. Low-frequency acoustic emissions have had limited effectiveness when used in the past to characterize and diagnose joint and muscle disorders because Fourier transform techniques were not capable of describing when a specific frequency occurred during an acoustic emission event. With the development of techniques that decompose a signal into time-frequency representations using wavelet analysis it is now possible to describe complex transient signals. While the continuous wavelet transform (CWT) method has already made an impact within many areas of the medical sciences such as EEG analysis, ECG analysis, and DNA-sequence analysis, the method is now becoming popular for analyzing muscle and joint acoustic emissions from the hip, knee, and elbow joints. This paper will present the results of our application of the CWT to the analysis of acoustic emissions from the shoulder joint.

Scale effect and acoustic emission characteristics of rock damage are two important topics in rock engineering research. In this paper, considering different rock scales and establishing the 3D numerical rock models by means of PFC software platform firstly, the mechanical properties and acoustic emission characteristics were studied. Finally, based on the acoustic emission, characteristic discussed the damage evolution law of the rock models. The research result shows that the multi-scale mainly affects the peak strength, peak strain, and elastic modulus of the mechanic properties, affects the maximum acoustic emission event number, strain range of serious acoustic emission events of the acoustic emission characteristics, affects the damage rapid increase and damage failure stages of the damage evolution process. The stress-strain curves, acoustic emission events curves, and damage variable curves can be divided into three stages and different scales have different effect on these stages.

In this study, self-healing of mortar was achieved by “built-in” carbonation of soluble Na2CO3 and Ca(OH)2. The effect of carbonate and calcium ions, available either internally or externally by conditioning the specimens with Ca(OH)2 and Na2CO3 solution, on the formation of calcite in cracks was investigated. The acoustic events were monitored and compared in the loading process before and after healing. Furthermore, a calibration test was carried out to distinguish the characteristic acoustic emission events of the fracture of the matrix and of healing products. It was found that the distribution of acoustic energy with FMA (frequency at maximum amplitude) and the hits with duration show a consistent trend with that of calibration. The change of tensile to shear cracking ratio in reloading illustrates a self-healing effect of cracks. Meanwhile, X-ray diffraction analysis indicates more calcite formed in the crack of self-healing specimens. The pretreated ceramsite-containing specimens exhibit the predominate self-healing effect due to the internally available CO32– and sufficient Ca2+ as compared to the others.

Due to the advantage of automatically extracting features from raw data, deep learning (DL) has been increasingly favored in the field of machine fault diagnosis. However, DL exposes the problems of large sample size and long training time, and in actual working conditions, the amount of labeled fault data available is relatively small, so a DL model of good generalization and high accuracy is difficult to be trained. In order to solve these problems, a deep transfer convolutional neural network (DTCNN) is proposed in this research. ResNet-50 is selected as the pre-trained model of deep convolutional neural network, and is transferred to solve the problem of bearing fault classification based on the idea of transfer learning. Firstly, raw fault signals are converted into time-frequency images by using continuous wavelet transform (CWT). Then, the images are further converted into RGB formats, which are used as the input of DTCNN. Finally, an end-to-end fault diagnosis model based on DTCNN is designed. The proposed method is validated on two datasets collected from motor bearing and self-priming centrifugal pump, respectively. Most sub-datasets from motor bearing show the prediction accuracies near 100%, and in the self-priming centrifugal pump dataset, we achieve improvement in accuracy from 99.48%±0.1966 to 99.98%±0.0332. The experimental results demonstrate that the proposed method outperforms other DL methods and traditional machine-learning methods.

In this paper a radial-basis-function neural network is used to localize acoustic emission events in ship hulls. It is shown that using a tiny network configuration and a small set of robust features, selected automatically by the K-means algorithm from a superset of 90 signal parameters, the location of a single event can be classified efficiently into three typical areas found in ship hulls. In simulation experiments, where a stiffened plate model is partially sunk into the water, the localization rate of acoustic emission events in a noise-free environment is 100% using four sensors and only four features per sensor. In the proposed method, the feature set is adapted and estimated automatically in cases of noisy environments. Robust acoustic emission localization rates, greater than 90%, are achieved using less than ten features per sensor in case of additive white Gaussian noise at 0 dB SNR or more.

In order to investigate the effects of elliptical defects on rock failure under ultrasonic vibrations, ultrasonic vibration tests and PFC2D numerical simulations were conducted on rocks with single elliptical defects. The research results indicated that the fracture fractal dimension, axial strain, and crack depth of specimens with elliptical defects at 45° and 90° were the smallest and largest, respectively. The corresponding strain and fractal dimension showed a positive linear and logarithmic function relationship with time. The maximum crack depth of 46.50 mm was observed on the specimens with an elliptical defect angle of 90°. Specimens with elliptical defects at 0°, 30°, 75°, and 90° exhibited more dense and frequent acoustic emission events than those with elliptical defects at 15°, 45°, and 60°. During the ultrasonic vibration process, the maximum total energy (87.86 kJ) and energy consumption coefficient (0.963) were observed on specimens with elliptical defect angles of 30° and 45°, respectively. The difference in the stress field led to varying degrees of plastic strain energy in the specimens, resulting in different forms of crack propagation and triggering differential acoustic emission events, ultimately leading to specimen failure with different crack shapes and depths. The fractal dimensions of elliptical defect specimens under ultrasonic vibration have a high degree of consistency with the changes in axial strain and failure depth, and the fractal dimension of defect specimens is positively correlated with the degree of failure of defect specimens.

Temporal and spatial distribution of microfractures in granites of different structures under triaxial compression and its significance in seismology