Search

Non-destructive inspection and analysis techniques are crucial for quality assessment and defect analysis in various industries. They enable for screening and monitoring of parts and products without alteration or impact, facilitating the exploration of material interactions and defect formation. With increasing complexity in microelectronic technologies, high reliability, robustness and thus, successful failure analysis is essential. Machine learning (ML) approaches have been developed and evaluated for the analysis of acoustic echo signals and time-resolved thermal responses for assessing their ability for defect detection. In the present paper different ML architectures were evaluated, including 1D and 2D convolutional neural networks (CNNs) after transforming time domain data into the spectral- and wavelet domains. Results showed that 2D CNNs processing data in wavelet domain representation performed best, however at the expense of additional computational effort. Furthermore, ML-based analysis was explored for lock-in thermography to detect and locate defects in the axial dimension based on thermal emissions. While promising, further research is needed to fully realize its potential. [ABSTRACT FROM AUTHOR]

This paper provides a framework based on machine learning approach in I4.0 environment to predict the remaining useful life of an aviation engine. For illustration purpose, an industrial case study is presented which applies machine learning algorithms to analyze the data collected using the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) simulation which includes run-to-degradation data for a turbofan engine. The results obtained from the study validate the proposed framework to identify prominent features and perform sequential analysis on unstructured data for predicting the remaining useful life of an aviation engine. Six machine learning models are applied to the dataset containing four subsets: FD001, FD002, FD003 and FD004 in C-MAPSS dataset each working on different degradation conditions for turbofan engine. For FD001, random forest had the lowest RMSE (11.59), and for FD002, FD003 and FD004, the lowest RMSE was given by LGBM classifier (12.78, 7.95 and 11.04), respectively. From the findings, it is observed that LGBM performs better with higher AUC 89% and lowest RMSE. The proposed framework can be applied to a wide range of failure prediction applications. Regardless of the underlying physics, ML-based data-driven methodologies can be used to analyze a wide range of systems. [ABSTRACT FROM AUTHOR]

Vibration analysis plays a pivotal role in the initial identification and reduction of defects in helical gear transmissions, underscoring the significance of precise fault detection. This research paper presents a comprehensive examination of vibration analysis as a means of detecting tooth wear faults in helical gear transmissions. The study utilizes the logistic regression (LR) algorithm and stochastic gradient descent (SGD) optimizer within a machine learning framework. A thorough examination of existing literature elucidated the importance of gear fault diagnosis and identified shortcomings in previous research efforts. To overcome these constraints, the suggested methodology incorporates sophisticated vibration analysis methodologies, techniques for enhancing data quality, and algorithms based on machine learning. Experimental trials, from the acquired vibration signals, conducted on a manufactured helical gear transmission system provide evidence of the effectiveness of the methodology, leading to a substantial enhancement in classification accuracy from 30.3% using the LR-based model to 97.8% using the LR-SGD-based model. The examination of the confusion matrix and ROC analysis provides additional evidence for the improved classification efficiency, as indicated by a substantial increase in the area under the curve from 0.653 to 0.997 in the LR and LR-SGD models, respectively. The results of this study demonstrate significant progress in the field of gear fault diagnosis, offering a sturdy and dependable methodology for practical implementations. Future research directions may encompass the augmentation of the dataset, investigation of alternative machine learning algorithms, and integration of supplementary diagnostic techniques to further amplify fault diagnosis capabilities. [ABSTRACT FROM AUTHOR]

Industry 4.0 brings about a convergence between traditional industrial practices and modern computational techniques. The manufacturing sector is expected to make a greater number of decisions and actions based on the advanced computational analysis of the measurable machinery data. Predictive maintenance of the CNC lathe spindle unit helps the machine avoid unexpected downtimes and meet closer machining tolerances. In this paper, the vibration signals acquired from an experimental run-to-failure test rig are first processed to extract health degradation features, which are then subjected to a neighborhood component analysis-based regression feature selection criteria. Finally, the selected spindle health degradation features are used to train a support-vector machine (SVM) algorithm to evolve a remaining useful life (RUL) estimation model. The SVM hyperparameters that strongly affect the prediction performance are tuned using the Bayesian optimization approach. The evolved predictive model is tested using an independent lathe spindle health degradation data set to obtain the root-mean-square error for predicted and actual RUL. The overall predicted RUL is having an acceptable agreement with the actual RUL. An RMSE equal to 206.23 is obtained as a quantitative measure of prediction accuracy for the Bayesian optimized SVM model for the given dataset. In industrial practice, the evolved SVM predictive model can be employed for the real-time RUL estimation of a similar mechanical system. The proposed predictive model with the integrated feature extraction, selection, and prognosis algorithm can be employed on a real-time spindle health monitoring and predictive maintenance platform for maintenance decision-making. [ABSTRACT FROM AUTHOR]

This paper presents a classification-based failure analysis of medium voltage underground power cables using voltage measurements with an echometer. The study focuses on identifying different types of cable faults, such as short circuits and open circuits, based on voltage signatures obtained through echograms. A comprehensive dataset of voltage measurements and echograms from various fault scenarios is collected and analyzed. Machine learning techniques are applied for classification, enabling accurate fault detection and localization. The proposed approach offers a reliable and efficient method for cable fault analysis, contributing to improved maintenance and reduced downtime in medium voltage underground power cable networks. The results demonstrate the effectiveness of using voltage measurements with an echometer for classifying cable faults, making it a valuable tool for fault diagnosis and prompt corrective actions. [ABSTRACT FROM AUTHOR]

The paper studies time series of dynamical systems for failures, applying data-driven machine learning techniques, such as clustering and tipping point analysis. Artificial data with known properties and real systems case studies are considered, with diverse patterns of time series. Applicability of various techniques is discussed. The proposed methodology may be useful in industrial and geophysical applications, where sensor records are available for data-driven failure analysis. [ABSTRACT FROM AUTHOR]

Layup optimization of the composite laminates is a very complex problem due to the convoluted multidimensional solution space which is usually explored by addressing different heuristic methods from which the most reliable are the genetic algorithms (GA). The optimization process converges by evaluating a lot of layup configurations which imply that the evaluation should be not only robust but also very fast. The most accurate numerical tool used to simulate the mechanical behavior of the composite laminates is the finite element analysis (FEA) which unfortunately is a computational intensive method. Some studies proposed very fast FEA models specially designed for the layup optimization with the lower bound of the execution time determined by the global linear system solving. Other studies pushed this bound even lower using classical machine learning techniques trained with prior observations (layup configurations) evaluated with FEA. It has been shown that the trained models can successfully replace the computational intensive FEA. The results are very important because the optimization time is dramatically reduced, while the estimation errors induced by the statistical models are acceptable. In this paper, we propose different deep neural network architectures such as multilayer perceptron (MLP) and convolutional and recurrent neural networks (CNN and RNN) that significantly reduce the estimation errors. For example, the classification error reduces from 2% to zero compared to previous studies, for the same numerical example. Also, we use different sets of predictors which allow the failure estimation for each layer in the composite laminate opposite to the previous studies which model the failure response only for the whole structure. [ABSTRACT FROM AUTHOR]

The given document provides information about various events and conferences taking place in different locations around the world. These events cover a wide range of topics, including forensic engineering, steel industry, ceramic and glass technology, digital transformation in the oil and gas industry, thermal spray and PVD coatings for aerospace, materials protection and performance, geotechnical engineering, mechanical engineering, materials research, polymer engineering for energy, and more. The document includes details about the themes, program tracks, and contact information for each event. [Extracted from the article]

Remaining useful life (RUL) prediction plays a significant role in prognostics and health management systems. While three different approaches have been utilized to estimate the RUL, hybrid-based methodologies yield more accurate results in this field. This study aims to introduce a hybrid prognostic approach based on deep learning methods, including long short-term memory (LSTM) and convolutional neural network (CNN). In most of the combined models, CNN is using to extract the features, and then, these LSTM be fed by extracted information from CNN, but in the hybrid model, both LSTM and CNN use organically to enhance the prediction ability. Besides, the time window (TW) is utilized to provide sequential data by sliding it on input data. To evaluate the proposed model's accuracy and speed, the KPCA algorithm is used to determine the dependency of the model on extracted features. The proposed model is validated on the data developed by NASA's commercial modular aero-propulsion system simulation (C-MAPSS) dataset. The results have illustrated that removing less important features has no effect on the proposed model. [ABSTRACT FROM AUTHOR]

This article aims to predict the mid-span displacements of reinforced concrete (RC) columns exposed to blast loading with machine learning models. Machine learning models including the gradient boosting decision tree (GBDT) model and the random forest (RF) model are developed. The model hyperparameters in the machine learning models are globally optimized using Bayesian optimization method. The dataset used to train and test the models is constructed by collecting data from shock-tube-simulated experiments. Results show that the performance of the optimized models is significantly improved. By comparing the performance metrics of two models, it is evident that the GBDT model is superior to the RF model in predicting mid-span displacements. The optimized GBDT model is then used to predict the displacements of columns during field tests, and the experimental results and predictive results are relatively consistent, which verifies the effectiveness of GBDT model with Bayesian optimization in damage assessment for RC columns. [ABSTRACT FROM AUTHOR]

Accurately identifying bearing defects is crucial for guaranteeing the dependability and effectiveness of industrial systems. Although the use of vibration signals for diagnosing bearing faults is highly important, there are still persistent obstacles, especially when it comes to detecting minor damage in the early stages. Relying solely on time-domain analysis for statistical feature extraction in complex multi-fault scenarios may lack robustness. The computational difficulties of frequency-domain and time–frequency approaches impede the real-time identification of emergent errors, despite their effectiveness. Although machine learning holds potential, its reliance on attributes that are not generated from the time domain poses a difficulty. Therefore, the main aim of this study is to fill these deficiencies by examining the utilization of temporal vibration signals for the purpose of diagnosing bearing defects. The vibration signals originated from the Case Western Reserve University. A total of fourteen time-domain features were derived from the vibration signal, encompassing root mean square, kurtosis, and skewness. The study employed two feature selection strategies, specifically Information Gain and Fast Correlation-Based Filter (FCBF), to identify the most important seven features for training machine learning models, including k-Nearest Neighbor (kNN), Support Vector Machines, and Naïve Bayes. Based on the acquired data, the kNN-based FCBF model (kNN-FCBF) approach exhibited superior classification outcomes in comparison to alternative methods. The evaluation metrics, including Area Under the Curve (AUC), Accuracy (AC), F1-score, Precision, and Recall, demonstrated a robust performance. The AUC attained a value of 99.1%, AC was assessed at 97%, F1-score reached 96%, Precision was 96%, and Recall earned a score of 95.7%. The benefits of the kNN-FCBF model were emphasized by a comparison analysis with prior studies. The kNN-FCBF algorithm provides a straightforward and precise approach that is less intricate and computationally affordable, while still achieving high levels of accuracy. [ABSTRACT FROM AUTHOR]

Accurately predicting stress–strain curves is essential for understanding the plastic behavior of metallic materials. This study investigates the effectiveness of machine learning (ML) methods in predicting stress–strain curves for aluminum alloys at different temperature levels. Specifically, three ML techniques, Gaussian process regression (GPR), neural network (NN), and boosted trees (BST), were utilized to predict the stress–strain response of Al6061-T6 at temperatures ranging from 25 to 300 °C. The performance of these ML models was evaluated using actual strain–stress measurements obtained from uniaxial tensile testing on Al6061-T6. A fivefold cross-validation approach was applied to train the models under investigation. Optimal parameters for the ML techniques were obtained during the training phase using the Bayesian optimization method to minimize mean absolute error. Four statistical metrics were employed to assess the accuracy of the predictions. The results of this study demonstrate the potential of machine learning methods in accurately predicting strain–stress measurements of materials. Additionally, the NN model outperformed the other models, achieving an average mean absolute error percentage of 0.213 and a coefficient of determination R2 of 0.998. Furthermore, it was observed that crack initiation mechanisms varied with temperature; particle fracture dominated at temperatures up to 200 °C, while interfacial decohesion prevailed at 300 °C. [ABSTRACT FROM AUTHOR]

Solar panels have grown in popularity as a source of renewable energy, but their efficiency is hampered by surface damage or defects. Manual visual inspection of solar panels is the traditional method of inspection, which can be time-consuming and costly. This study proposes a method for detecting and localizing solar panel damage using thermal images. The proposed method employs image processing techniques to detect and localize hotspots on the surface of a solar panel, which can indicate damage or defects. The findings of this study show that the proposed method is effective in detecting and localizing solar panel damage and can reduce inspection time and cost. This study proposes a method for detecting and localizing solar panel damage using thermal images. The proposed method employs image processing techniques to detect and localize hotspots on the surface of a solar panel, which can indicate damage or defects. The findings of this study show that the proposed method is effective in detecting and localizing solar panel damage and can reduce inspection time and cost. The proposed method has the potential to improve the efficiency and lifespan of solar panels while also contributing to the wider adoption of renewable energy. This research suggests a way for detecting and localizing solar panel damage using thermal imaging, which could get rid of the requirement for manual visual examination. The suggested technology detects and localizes hotspots on the surface of solar panels, which indicate faults or damage. This method can increase the efficiency and longevity of solar panels, hence promoting the use of renewable energy. Future improvements, such as incorporating AI and ML algorithms and advances in thermal imaging technologies, could improve the accuracy of this method even further. [ABSTRACT FROM AUTHOR]

The detection of rolling bearing faults of rotating machines is very important in reducing production losses, financial losses, and accidents in the manufacturing industry. Therefore, various methods have been developed so far for the detection of rolling bearing faults. Recently, transfer learning-based methods are getting popularity in the area of artificial intelligence for this purpose. In this study, the performance of a transfer learning algorithm for fault classification using scalogram images has been studied at different load conditions of the collected dataset, size of the dataset, and training–testing ratio of the dataset for the benefit of future researchers. [ABSTRACT FROM AUTHOR]

Jonathan Hopkins, CC BY-ND The experimental study lays the foundation for AI materials that can be applied in the construction of buildings, airplanes, and imaging technologies. For more information: https://engineering.lehigh.edu/news/article/cooler-approach-making-new-materials-can-stand-heat AI Capable of Predicting Properties of Multifaceted Metamaterials In a new study by the University of Amsterdam's Institute of Physics and research institute AMOLF, the researchers tested how well artificial intelligence (AI) can estimate the properties of so-called combinatorial mechanical metamaterials. Courtesy of A. Kuzmin/University of Latvia and A. Smekhova/HZB To investigate the local environment of individual components, the team used multi-edge x-ray absorption spectroscopy (EXAFS) at BESSY II and then the reverse Monte Carlo method to analyze the collected data. High-entropy alloys are a new class of alloys that are composed of four or more metallic elements in approximately equal amounts. [Extracted from the article]

Rotary machines are key equipment in many industrial sectors, from mining operations to advanced manufacturing. Among the critical components of these machines are bearings, gearboxes, rotors, among others. These components tend to present failures, which can be catastrophic, with economic, safety and/or environmental consequences. Among the most established methods for classifying bearing faults, the envelope method has been widely used, with relative success, for several years. However, this method and its variations are difficult to automate and require extensive experience on the part of the analyst. We report that, while traditional methods (e.g., envelope) successfully classified bearing failures less than 45% of the time, machine learning methods were successful more than 62% of the time, and in some cases reaching 67%. This work differs from others in the sense that it uses all the available measurements from a well-known database, not just a subset. In addition, the measurements are taken at the motor base, which are more difficult to classify, and avoid using different segments of the same signal in both training and validation, thus reducing the possibility of overfitting. As a consequence, the results obtained are apparently inferior to those reported elsewhere, but probably closer to what one might expect in practical applications. [ABSTRACT FROM AUTHOR]

The center shaft of rotary steering spindle system is bendable under bias force. A severe partial load effect occurs among rollers, the inside and outside circles of the first cantilever bearing. Simulation analysis was conducted by loading boundary condition of the spindle under bias force. Furthermore, three different types of deep cavity rollers, which were cylindrical, conical, and spherical, respectively, were analyzed by finite element method. The effects of deep cavity angles, radius, and offset on mechanical properties of bearing were studied. The data obtained by simulation analysis were trained and predicted by Back Propagation (BP) neural network, and then the BP neural network model was incorporated into fmincon function. Thereby, structure optimization of rollers was established based on BP neural network model and fmincon function. The results show that structure of the conical deep cavity roller gets optimal mechanical performance. After being optimized, maximum stress of edge region and elliptical area decreases, respectively, by 22 and 17% than before, indicating that structure optimization method of the neural network and fmincon function can be used in optimization of deep cavity rollers. This method can quickly search for the optimal solution with sufficient engineering accuracy, ease of use, and adaptability. [ABSTRACT FROM AUTHOR]