Detection Of Fatigue Damage Research Articles

Study on magnetic memory detection of weld fatigue damage by using the relative entropy theory

In this paper, fatigue tests of different kinds of welds were carried out. The three magnetic components of MMM signal on the weld under different cycles were collected. The change law of MMM signal with the fatigue cycle was analyzed, and the relative entropy theory was introduced to establish the mapping relationship between the relative entropy and the fatigue cycle. The damage state of the weld was evaluated based on the change of the relative entropy. The results showed that the curves of initial MMM signal presented specific change law with the increase of fatigue cycle and a counter-trend change occurred at the local position after the weld failure. In addition, there was a good correspondence between the relative entropy with the fatigue stage of the weld. The value of relative entropy in the fatigue stable stage remained approximately constant and tended to be close to zero. In the final fatigue stage, the value of relative entropy increased exponentially. The relative entropy gradient and the magnetic index k were calculated to capture the time node of the deterioration of the weld more accurately and the k = 0.1 could be defined as the critical value to evaluated the stability of the weld. Hence, the relative entropy that was introduced could effectively characterize the fatigue damage degree of the weld, and realize the early warning of the weld damage.

A Unified Mixed Deep Neural Network for Fatigue Damage Detection in Components with Different Stress Concentrations

The article presents a mixed deep neural network (DNN) approach for detecting micron-scale fatigue damage in high-strength polycrystalline aluminum alloys. Fatigue testing is conducted using a custom-designed apparatus integrated with a confocal microscope and a moving stage to accurately pinpoint the instance of micron-scale crack emergence. The specimens are monitored throughout the duration of the experiment using a pair of high-frequency ultrasonic transducers. The mixed DNN is trained with ultrasonic time-series data that are obtained from two sets of specimens categorized by different stress concentration factors. To understand the effects of mixing the data from both types of specimens, a parametric analysis is performed by varying the amount of training data from each specimen to develop a series of mixed DNNs. The mixed DNN, when tested on unseen data from both specimens, exhibits an accuracy of over 95%. This article, therefore, demonstrates a successful alternative to customized DNNs for new types, geometries, or stress concentration factors in the materials under consideration.

Ultrasonic methods for the detection of near surface fatigue damage

Fatigue zones in a material can be identified using ultrasonic waves, as it has been shown that their propagation speed will reduce when travelling through such a zone. However, as fatigue damage is usually concentrated in a thin near-surface layer, through-thickness measurements result in very small changes of the average propagation speed across the full thickness, which are potentially difficult to reliably correlate to specific fatigue states. In this study, we have completed fatigue state assessments using Rayleigh waves, which travel on the surface of a material, to maximise those changes. We found that the use of Rayleigh waves amplifies the changes in speed, after propagation in the damaged region, by a factor of up to ten. The monotonic nature of the reduction in wave speed was verified against the theory using dislocation density measurements. Finally, a stiffness-reducing finite-element modelling technique, able to capture the effects of fatigue on the time of flight of longitudinal bulk and Rayleigh waves, was also derived and verified against the experimental measurements.

Dynamic response analysis of floating wind turbine platform in local fatigue of mooring

The moorings of the Floating Wind Turbine (FWT) platforms, long term suffering from the coupling loads of wind, waves and currents, are especially prone to structural fatigue. The purpose of this study is to mine effective information from the dynamic response of the FWT platform to achieve early damage detection for mooring health conditions. However, high nonlinearity of the FWT platform dynamic response, that is caused by the complexity of the working environment, hinders the accuracy of fatigue analysis and damage detection, Therefore, in this study, motivated by the accuracy of the chaotic features in quantifying nonlinearities and reliability of the convolutional neural network for feature extraction, an intelligent damage detection model, named Convolutional Neural Network-t-distribution Stochastic Neighbor Embedding (CNN-t-SNE), is proposed to automatically detect the damage magnitude of the moorings. Through analyzing the dynamics of FWT platform mooring from structure creep to failure, it is found that the yaw response is the most sensitive to structural damage. To examine the reliability of the proposed CNN-t-SNE method, the Lyapunov exponent and chaotic attractor quantify the nonlinearity of the features in the neural networks to indicate that the nonlinearity of the features decreases as the neural network layer deepens.

Numerical Study of Rayleigh Wave Interaction with Rolling Contact Fatigue Type of Defects

ABSTRACT Rolling Contact Fatigue or Damage (RCF/RCD) presents significant maintenance challenges to railroads across the globe. Quantifying RCF/RCD crack depths and density in rails is important for all railroads to manage their grinding programs effectively and efficiently and being able to conduct ultrasonic testing (UT) of rails for reliable detection of internal fatigue damage. This work focuses on the modeling of Rayleigh waves UT approach to detect and characterize RCF type of defects which can form as: vertical, oblique or branched shaped surface breaking defects in the rail head. Specifically, the transmission coefficient (Tc) of Rayleigh waves was studied using finite element analysis (FEA). The effect of crack tip geometry on Tc values is discussed. The results suggest that for oblique, and branch cracks, characterization based purely on the Tc can be challenging due to symmetric sinusoidal fluctuations in the Tc. A real crack using a micrograph image was also modeled to validate the Tc results for oblique and branched cracks. This points to the need for additional parameters to be identified for efficient and reliable characterization of RCF/RCD type of defects in rails.

Detection and Characterization of Fatigue Cracks in Butt Welds of Offshore Structures Using the Eddy Current Method

Abstract A reliable assessment of the actual condition of wind turbines is essential for their continuous operation, or if an extension of their service life is considered. To be able to make valid statements about the condition of the welded structure, an inspection concept for the reliable detection of fatigue cracks and damage on welds of the offshore tower and foundation structures is essential. In the present study, five different eddy current differential sensors were evaluated on cyclically fatigued DY butt welds under tensile load with a focus on the signal-to-noise ratio (SNR). Based on previous evaluations, two sensors were selected for semi-automatic weld testing. It is shown that due to the weld crossing necessary for the detection of fatigue cracks, air coils arranged parallel to the test surface have the highest SNR. This must be contrasted with the potential for crack depth determination. In this context, coils with different arrangements with respect to the test surface were analyzed. It is shown that groove depths can be differentiated based on the imaginary part of the measurement signal for groove depths of up to 8 mm, and actual fatigue cracks in welds with a crack depth of 0.5 mm were detected.

Assessment of Transfer Learning Capabilities for Fatigue Damage Classification and Detection in Aluminum Specimens with Different Notch Geometries

Fatigue damage detection and its classification in metallic materials are persistently challenging the structural health monitoring community. The mechanics of fatigue damage is difficult to analyze and is further complicated because of the presence of notches of different geometries. These notches act as possible crack-nucleation sites resulting in failure mechanisms that are drastically different from one another. Often, sensor-based tools are used to monitor and detect fatigue damage in critical metallic materials such as aluminum alloys. Through deep neural networks (DNNs), such a sensor-based approach can be ubiquitously extended for a variety of geometries as appropriate for different applications. To that end, this paper presents a DNN-based transfer learning framework that can be used to classify and detect fatigue damage across candidate notch geometries. The DNNs are built upon ultrasonic time-series data obtained during fatigue testing of Al7075-T6 specimens with two types of notch geometries, namely, a U-notch and a V-notch. The baseline U-notch DNN is shown to achieve an accuracy of 96.1% while the baseline V-notch DNN has an accuracy of 95.8%. Both baseline DNNs are, thereafter, subjected to a transfer learning process by keeping a certain number of layers frozen and retraining only the remaining layers with a small volume of data obtained from the other notch geometry. When a layer of the baseline U-notch DNN is retrained with just 10% of the total V-notch data, an accuracy above 90% is observed for fatigue damage detection of V-notch specimens. Similar results are also obtained when the baseline V-notch DNN is retrained and interrogated to detect damage for U-notch specimens. These results, in summary, demonstrate the data-thrifty quality of combining the concepts of transfer learning and DNN for fatigue damage detection in different geometries of specimens made of high-performance aluminum alloys.

Early fatigue damage evaluation based on nonlinear Lamb wave third-harmonic phase velocity matching

In this paper, third-harmonic phase velocity matching nonlinear Lamb waves are used for the early fatigue damage detection of aluminum alloys. An analysis is conducted regarding the relationship of third harmonic and third-order nonlinear parameters with fatigue accumulation. According to the analytical results, the amplitude of the third harmonic increases when the fatigue life falls below about 80%. However, when the fatigue life exceeds 80%, the third harmonic starts to decrease, and the third-order nonlinear parameters increase progressively before showing a downward trend. It can be found out that micro-cracks emerge with the accumulation of fatigue, thus reducing the amplitude of the third harmonic and causing the third-order nonlinear parameters to increase gradually and then decline. Furthermore, it is proposed to take the third harmonic amplitude and the third order nonlinear parameters as the indicators of early fatigue damage.

Feature extraction and neural network-based fatigue damage detection and classification

Feature extraction and neural network-based fatigue damage detection and classification

Inherent Magnetic Sensor for Estimation of Fatigue Damage in Type 304 Stainless Steel

Abstract In this study, the fatigue damage in type 304 stainless steel was estimated using an inherent magnetic sensor via plane-bending fatigue tests and the electromagnetic impedance method. The sensor was a magnetic composite material incorporating a ferromagnetic martensite phase generated in type 304 stainless steel by a surface finish process during the production stage. The output properties of this sensor as a function of the number of cycles were evaluated under various conditions. It was demonstrated that this sensor could detect fatigue damage starting from the zeroth cycle. The sensor output repeatability was evaluated, and the variation in the output between the inherent magnetic sensors was approximately 10% regardless of the sensor type and total strain amplitude. By using the two proposed estimation methods, the specific fatigue level and the number of cycles could be estimated with errors of 3–27%. These results indicated that the inherent magnetic sensor was suitable for use for fatigue damage estimation.

Non‐destructive fatigue damage detection of carbon fiber‐reinforced thermoplastics using positron annihilation method

AbstractTensile axial fatigue tests were carried out using carbon fiber‐reinforced thermoplastics (CFRTP) plate with a hole notch up to very high cycle fatigue (VHCF) regime, where number of loading cycles was 2 × 107. The fatigue test was periodically interrupted, and fatigue damage accumulation around the hole notch was non‐destructively detected by two positron annihilation methods. Positron annihilation lifetime spectroscopy was applied to obtain positron lifetime and Doppler broadening method to obtain S‐parameter. Both parameters increased with increasing fatigue loading cycles. Fatigue damage detected by two positron annihilation methods was attributed to the expansion of free volume in the polymer matrix. The estimated size of free volume was sub‐nanoscale according to the experimental value of the positron annihilation lifetime. SEM and μCT scan revealed that the macroscopic fatigue damages, such as surface cracking, transverse cracking and delamination, increased with increasing fatigue loading cycles, which indicated that sub‐nanoscale and macroscopic fatigue damages progressed simultaneously. Consequently, it can be concluded that the positron annihilation methods can be used for the non‐destructive fatigue damage detection of CFRTP.

Machine learning based thermal imaging damage detection in glass-epoxy composite materials

Machine learning (ML) based fatigue damage detection from thermal imaging in glass-epoxy composites is an important component of remote structural health monitoring used for safety assessment and optimization of composite structures and components. However, accurate characterization of fatigue damage hotspots in terms of size, shape, location, hysteretic heat, and local temperature deep inside the material using surface thermal images remains a challenge to date. This work aims at evaluating the theoretical accuracy level of hotspot characterization by training a ML model with artificially generated thermal images from 3D finite element models with increasing complexity. Modelling the fatigue damage as an intrinsic heat source allowed to significantly reduce the influence of thermal image noise and other uncertainties related to heat transfer. It is shown that ML can indeed accurately recover the heat influx, depth, and geometry of the heat source from the original thermal images of the composite materials with prediction accuracies in the range 85%–99%. The effect of training set size and image resolution on the prediction error is also presented. The findings reported in this work contribute to the advancement of accurate and efficient remote fatigue damage detection methods for fibre composite materials.

Potential energy as a new approach for detection and monitoring of fatigue damage in scarf adhesive joints modified with nanoparticles

In this study, fatigue results of the scarf adhesive joints (SAJs) were extensively analyzed regarding the effect of the adhesive materials (neat epoxy (NE), and SiC and Al2O3-nanocomposite), bond thickness (0.17 mm and 0.25 mm) on their dynamic properties, which include loss factor (tan δ), loss modulus ( E''), dissipated energy ( U d), potential energy ( U p), and storage modulus ( E'). Safe fatigue lives of the NE, SiC, and Al2O3-SAJs were estimated at different reliability levels. The experimental measurements revelated that reducing the adhesive layer thickness from 0.25 mm to 0.17 mm result in enhancing the tensile strength and stiffness of the SiC-SAJs by 9.6% and 4.4%, respectively, in return of 9.5% and 5.4% for the Al2O3-SAJ relative to the NE-SAJ. The strength and stiffness of the SiC-SAJs with 0.17 mm bond thickness are higher than those of the Al2O3-SAJ by 15.4% and 3.3%, respectively. The highest improvement of 22.0% in the fatigue strength was occurred for the SiC-SAJ with bond thickness of 0.17 mm in return of 4.8% for the Al2O3-SAJ. The U p was used for the first time (except author works) for monitoring and predicting the damage in the adhesive joints practically the change in the enclosed area of the hysteresis loop is too small to cause a large variation in the tan δ, E'', and U d. Damping factor of the NE-SAJs is marginally increased with the stress levels compared to the aggravated increase of both SiC and Al2O3-SAJs. Zhang model predicts well the potential energy and storage modulus specially at higher stress levels.

A new approach for fatigue damage detection in adhesive joints modified with nanoparticles under different temperatures

AbstractScarf adhesive joints (SAJs) in carbon fiber composites were fabricated with 5° taper angle and bondline thickness of 0.17 mm. The joints are modified with SiC‐nanoparticles and characterized through tensile and fatigue loading at room temperature (RT) and 50°C. The analysis of the hysteresis loop in terms of loss and storage moduli, potential and dissipated energies, and damping factor is a quite new topic and one of the important goals of this study. At RT, the fatigue limit of the SAJs was improved by 55% when the bondline thickness decreased from 0.25 to 0.17 mm and further enhanced by 22% with SiC incorporation. For the first time, the evolution of the potential energy versus fatigue cycles of the SAJs can be modeled for early predicting joint life and hence helps in preventing their sudden failure. At 50°C, the fatigue strength of the SiC‐SAJ is approximately equal to the unmodified joint at RT.

Advanced numerical simulations considering crack orientation for fatigue damage quantification using nonlinear guided waves

A numerical study including a fatigue crack trajectory simulation was undertaken by means of separating morphing and adaptive remeshing technology (SMART) Crack Growth in ANSYS, on the basis of which the simulations of nonlinear ultrasonic waves for fatigue damage detection using the precise fatigue crack trajectory was achieved. The simulated crack trajectory was first validated by experimental results in terms of crack initiation angle and number of fatigue cycles and was subsequently utilised for crack quantification based on second harmonic method. The results revealed that the nonlinearity in terms of the trend and magnitude with respect to crack length in the advanced simulation is closer to that in the experimental results than the common simulation approach where damage was modelled as a straight line crack. Given that the influence of crack orientation on contact acoustic nonlinearity (CAN) was taken into consideration, the developed advanced simulation could further enhance the capability of numerical modelling for simulating the interaction between nonlinear guided waves and fatigue crack, facilitating the fundamental investigation of CAN mechanism.

The estimation method of normalized Nonlinear Output Frequency Response Functions with only response signals under stochastic excitation

The nonlinearity degree of the system can be sensitively revealed by the methods based on Nonlinear Output Frequency Response Functions (NOFRFs). However, there is a lack of methods based on NOFRFs that work with only response signals under stochastic excitation, which limits the application of NOFRFs in practical engineering. In this study, a new concept of normalized NOFRFs is proposed, and a corresponding estimation method that employs only stochastic output response is established. Firstly, a method of estimating NOFRFs by Power Spectral Density (PSD) is established, and therefrom the normalized NOFRFs and a corresponding nonlinearity index are proposed. Then, by taking the PSD of the excitation signal as the input and the PSD of the response signal as the output, a general estimation method of normalized NOFRFs is built, and an output measured only estimation method of normalized NOFRFs is further deduced and established. In these two estimation methods, the assumption of white noise excitation is not required, and the estimated results are insensitive to the change of excitation intensity. Finally, the experiments of fatigue damage detection of three-point bending fatigue steel plate specimens and reducer box cases under stochastic excitation are carried out. The results demonstrate that, compared with the NOFRFs, the degree of fatigue damage can be effectively revealed by the normalized NOFRFs with output.

Magnetic characteristics and mechanism of 304 austenitic stainless steel under fatigue loading

This study examines the magnetic signal and mechanism of 304 austenitic stainless steel under fatigue. Fluxgate sensor is used to detect the magnetic signal during fatigue. Scanning electron microscope (SEM), x-ray diffraction (XRD), and Lorentz transmission electron microscope (LTEM) are used to study the microstructure’s evolution. On the basis of the microstructure results, a mechanism of magnetic characteristics is discussed and demonstrated. In this mechanism, martensite accumulated by phase transformation is magnetized by stress and enhances the magnetic signal. The magnetic signal’s trend and mechanism vary in three fatigue stages. In the first stage, the magnetic signal increases rapidly due to the increased content of strain-induced martensite. In the second stage, the magnetic signal increases steadily, and the relative strain-induced martensite content is only 0.6%. The increase in magnetic signal is primarily caused by the accumulated martensite being magnetized by stress. In the third stage, the magnetic signal increases rapidly due to the rapid development of fatigue cracks. Lastly, a differential magnetic signal processing method for detecting fatigue damage failure in 304 austenitic stainless steel is proposed; the proposed method can distinguish each of the fatigue stages. A warning of fatigue failure can be obtained when the results of the differential increase exceed the corresponding threshold value.

Fatigue damage detection and risk assessment via neural network modeling of ultrasonic signals

AbstractThis paper develops a data‐driven autonomous method for detection of fatigue damage and classification of the associated damage risk in mechanical structures, based on ultrasonic signal energy. The underlying concept is built upon attenuation of the signal and stability of the attenuation process. The attenuation provides pertinent information for damage quantification, whereas the stability represents resistance toward the fatigue damage growth. The proposed neural network (NN) model has been trained using the scaled conjugate‐gradient back‐propagation method. The NN model is capable of damage detection and damage classification into five classes of increasing risk. The wavelet transform tool has been used to reduce the noisy pattern of the ultrasonic signal energy by using the associated approximation coefficients. The results show that the proposed method of approximation signal energy can detect and classify the damage with an accuracy of up to ∼98.5%.

Numerical analysis on fatigue behavior of ultrahigh-performance concrete-Orthotropic steel composite bridge deck

Ultrahigh-performance concrete-orthotropic steel composite bridge deck is composed of the orthotropic steel deck and a thin ultrahigh-performance concrete (UHPC) overlayer. In the previous fatigue tests, two typical fatigue failure modes were found and identified. As a supplementary test after fatigue tests, air penetration method is capable of providing a reference to the quantitative and non-destructive damage detection of fatigue damage of UHPC. To further the previous study, a detailed numerical investigation is accomplished through complimentary finite element (FE) analysis. Compared with the solid element model, the refined shell-solid element model can better reflect the mechanical behavior. It is illustrated that the vertical stress can be adopted in assessing the fatigue strength of rib-to-diaphragm welded connection in the field test by means of nominal stress method. The combination of various factors would lead to fatigue shear failure of the short headed-studs. The fatigue strength of rib-to-diaphragm welded connection predicted by the hot spot stress method and the consistent nominal stress (CNS) method can basically meet the requirements of FAT90. The consistent nominal stress method can be used as the optimization method of nominal stress of fatigue detail. It is demonstrated that the fatigue life of UHPC can be estimated by S-N curves of ordinary concrete conservatively. The allowable equivalent maximum stress level can be taken as 0.55 for two million cycles of fatigue loading, and 0.52 for five million cycles of fatigue loading.

Fatigue damage detection of aerospace-grade aluminum alloys using feature-based and feature-less deep neural networks

Fatigue damage is one of the most common causes of failure in aerospace structural components. While numerical modeling and laboratory-scale experimentation provide much insight to the physics of failure evolution, it is extremely challenging to account for all variabilies that a component may be subjected to during on-field operation. Human-supervised continuous monitoring of such components using sensors, therefore, provides a much-needed alternative for reliable operation of these components. By leveraging the concepts in deep learning, such human supervision can be assisted with an automated pre-trained deep network for damage detection. To that end, this article studies two distinct deep neural network (DNN) architectures for fatigue damage detection in aluminum using ultrasonic time-series data obtained from a novel customized fatigue testing apparatus. The first DNN, called as a feature-based network, is built by using two predefined features viz. frequency domain amplitude and autocorrelation from the ultrasonic data as the inputs. The second DNN, called as a feature-less network, uses the ultrasonic data as-is without any pre-processing and relies on the black-box features generated during training. The capability of fatigue crack detection for both DNN architectures is evaluated at two distinct stages of fatigue failure. The feature-less network is observed to outperform the feature-based network with an accuracy of 94.26% and 98.94% for the two stages. The result indicates that feature-less DNNs, owing to their construction, can formulate better, albeit black-box features, and simplify the process of choosing customized signal processing methods for similar problems.