Search

The acquisition of large atomic-force-microscopy (AFM) scans at nanoscale resolutions can take hours and produce datasets with millions of pixels, which is time consuming and computationally expensive to analyze. In this paper, we present an approach to speedup this process by using a computer-vision algorithm, namely the Noise2Noise algorithm, to reconstruct high-resolution, low scan speed AFM data from high-speed, noisy, sparsely sampled AFM data. This algorithm is trained on various noise types to reproduce different sources of experimental noises encountered during the acquisition of AFM data. Our results demonstrate that a sparse, uniform AFM scan of 20 × 20 μ m at 128 × 128 pixel resolution can be processed within seconds, and the output image is comparable to a higher quality raw AFM data scan which required 30 minutes or more to collect and process manually, reducing not only the acquisition and analysis time, but also the size of the data being collected. • High resolution, low scan speed AFM images are reconstructed from higher scan speeds data. • The Noise2Noise algorithm is used to map noisy AFM images to clean AFM images. • Several noise modes are used to train the algorithm to mimic realistic noises in AFM data. • Process of raw and fast scan data corrections leads to up to 70% improvement in image quality. [ABSTRACT FROM AUTHOR]

Highlights • A deep learning method for steel seismic fuse damage inspection is investigated. • Micromechanical fracture index is defined to quantify the fuse damage condition. • A saliency-based method is adopted for damage feature pattern visualization. • The effectiveness of the proposed method is demonstrated in two case studies. Abstract This paper presents a deep learning based structural steel damage condition assessment method that uses images for post-hazard inspection of ultra-low cycle fatigue induced damage in structural steel fuse members. The deep learning model – Convolutional Neural Network (CNN) model, can be trained to represent the high dimensional features in huge amount of raw data which traditional mathematical models are unable to describe. A saliency-based visualization method is employed to visualize the feature-related pattern recognized by the deep learning model. To quantify the damage condition of the inspected structure fuse members, a micromechanical fracture index is defined as the damage index and used for labeling the images in the training data set. To provide large training dataset, cumulative plastic strain contour plot images generated through finite element (FE) simulation are adopted for the training data. Parametric studies were performed to validate and optimize the deep learning based damage condition assessment method. The method and findings are further examined using real experimental images collected from cyclic testing of a steel notched plate specimen. The results suggest that the proposed method provides a promising automation tool for rapid inspection of structural steel fuse member damage condition. [ABSTRACT FROM AUTHOR]

Federated learning (FL) guaranteeing data privacy is of great interest in decentralized fault diagnosis. However, limited research attention has been paid to the dynamic domain-shift issue due to varying working conditions. This paper proposes an active federated transfer algorithm based on broad learning to address the domain shift issue in FL. First, a central server dispatches a global model to the source clients for collaborative modeling. Subsequently, the global model is initialized with a federated averaging strategy. Next, the initialized global model is used to annotate emerging signals from the target clients based on an active sampling strategy proposed. Finally, an asynchronous update scheme is designed to adapt the global model to the target domain. The performance of the AFTBL algorithm is validated with three datasets, including 24 centralized- and decentralized-modeling tasks. The computational results indicate that the proposed algorithm is more accurate and efficient than the prevalent algorithms. • The cross-domain incremental federated learning problem is investigated. • An active federated broad transfer algorithm is proposed for fault diagnosis. • Automated data annotation and selection are proposed for incremental model update. [ABSTRACT FROM AUTHOR]

• The uncertainties due to the cyberattacks are tackled with a low computation burden deep learning algorithm. • A new IoT structure for monitoring the induction motor is developed. • A graphical visualization is performed to identify the faults of the induction motors. • Parameterizing the DNN of the fault recognition model is investigated. • The results confirm the robustness and resiliency of the designed model. Maintaining electrical machines in good working order and increasing their life expectancy is one of the main challenges. Precocious and accurate detection of faults is crucial to this process. Induction motors (IMs) are among these machines widely utilized in various fields including industrial and domestic applications that require effective detection of their status. This paper proposes a novel fault recognition and correction (FRC) scheme based on the internet of things (IoT) and deep learning for IMs. In the developed system, vibration signals during motor operation are used to generate bearing fault features, which are inputted to the designed deep learning model to successfully identify bearing faults. The robustness of the proposed approach is tested against a false data injection (FDI) attack. Further, the proposed deep learning approach has been assessed and compared with other state-of-the-art algorithms available in the literature. Experimental testing has been carried out on a real IM to perform the suitability of the developed fault detection and correction scheme. Compared to other fault recognition techniques, the proposed method proves to be more effective. In essence, the results verified the robustness of the novel proposed strategy against the FDI attack making it possible to recognize faults with confidence and improve decision-making to determine the motor's status. [ABSTRACT FROM AUTHOR]

In this paper, we use multivariate machine learning-based predictors to replace missing data and propose a mechanism to evaluate and track correctness by estimating its confidence level whenever successive missing data points occur. The proposed solution relies on the idea of confidence attribution, which assigns a value to every measurement, indicating how much it is believed to be accurate based on the difference between measured and predicted data. When data is missing, we perform data imputation using the predicted value and estimate confidence. We estimate confidence based solely on parameters used for confidence attribution and information acquired during the predictor's training. We evaluate the solution with two real datasets, one collected from a solar farm and another from a collection of wind turbines. The results show that the accuracy of multivariate models can decrease significantly when input data goes missing, demonstrating the need for the proposed confidence tracking mechanism. • Imputing successive missing data can impact the precision of multivariate predictors. • Multiple missing variables impact the precision of multivariate predictors. • Tracking the prediction error for each imputation based on the model properties. • A mechanism to evaluate the validity of imputing data based on error thresholds. [ABSTRACT FROM AUTHOR]

• Early prediction for mammography. • To diagnose the breast cancer at the earliest to reduce the mortality rate. • This paper proposes an expert system for the diagnosis of breast cancer disease using DE-RBF-KELM method. • The proposed method is effectively classifying the abnormal classes of mammograms. Breast cancer disease is a major public health problem among women worldwide. This article proposes an expert system for the diagnosis of breast cancer disease based on an evolutionary algorithm known as Differential Evolution (DE) of a Radial-Based Function Kernel Extreme Learning Machines (RBF-KELM). In the structure of the RBF-KELM, there are two adjustable parameters of the RBF-kernel which are the penalty parameter C and the RBF-kernel's parameter (σ). These parameters play a major role in the efficiency of RBF-KELM. In this study, the optimal values of these parameters have been obtained using a differential evolution (DE) algorithm. To validate the effectiveness of the suggested approach, DE-RBF-KELM was examined on the two datasets: The Mammographic Image Analysis Society (MIAS) and the wisconsin breast cancer database (WBCD) and the results were satisfactory compared to conventional approaches. [ABSTRACT FROM AUTHOR]

• The proposed 4 x -expert system is used to create a prediction system for osteoporosis suspected patients. • It is designed using multi model machine learning algorithms. • Which uses and displays the results from all the systems along with performance metrics. • Thus 4x-expert systems, helps in better diagnosing osteoporosis cases with reduced computational cost. Osteoporosis occurs due to micro-architectural deterioration of the bone tissues with an increased risk of bone fragility, which can cause fractures in the bone without much pressure applied to it. The T-score of a person's bone density report can be used to calculate the difference between BMD to that of healthy bones. Currently, osteoporosis is detected using conventional methods like DXA scans or high computational power requiring FEA tests. Considering individual approaches and mono-prediction techniques leads to omission of micro-fractional prediction parameters. In this paper, we have proposed a 4 x -expert system for suspected osteoporosis patients, which is designed using multi model machine learning algorithms for improving prediction and accuracy through the various computational process. The experiment results shows, that the 4 x -expert system covers the extensive prediction and accuracy of any suspected bone disorder patients, ranging from 75% to 97%. [ABSTRACT FROM AUTHOR]

• Proposing a new method for power quality disturbances recognition using ACMP. • The proposed method has high noise rejection capability. • Less required memory storage due to using of infinite feature selection. • High accuracy in PQDs recognition by using grasshopper optimized SVM. This paper presents a new method for identifying different types of power quality disturbances (PQDs). In this article, adaptive chirp mode pursuit (ACMP) is used to extract useful features and the greedy algorithm that uses the similar principles of the matching pursuit method is adopted in the ACMP. Besides, the ACMP takes advantage of the sparse matrices, which reduce the computational cost. The infinite feature selection as the filter-based algorithm is applied for the elimination of improper features. Also, the grasshopper optimization algorithm (GOA) is used to optimize the parameters of the SVM as the classifier. The obtained results of the simulations and real disturbances show the high accuracy and speed of the proposed algorithm, which makes it possible to use it in power quality measuring and analysis devices. Also, the proposed method has noise rejection capability, which is another advantage that justifies its use in measuring and analyzing PQDs. [ABSTRACT FROM AUTHOR]

• A model of general range measurement model is proposed. • The model obtains the initial parameters through layer-by-layer unsupervised learning. • The model uses the back-propagation algorithm to fine-tune the entire model. • The model uses SoftMax for classification and recognition. The focus of this article is that it is difficult to effectively measure the universal distance through the use of conventional parameters. This paper uniquely proposes the use of depth-limiting machine learning methods to model the universal distance measurement model, which can effectively perform multi-modal online measurement of the data measurement machine learning model. The feature extraction is unique to this article. The model uses the Softmax software to fine-tune the logarithm to verify the correctness of the propagation algorithm. It can effectively verify the measured distance and obtain multimodal data online. Recognition accuracy and strong robustness are also one of the characteristics of this article. This article has high academic value and innovative features, and is of great significance for the study of machine learning models. [ABSTRACT FROM AUTHOR]

• A new method for the in-process measuring the size and roundness of workpiece with discontinuous surface. • The proposed method is based on the machine learning technology. • An efficient approach for replacing the invalid data of the groove is also presented. • The roundness evaluation model established meets geometrical product specifications (GPS). • Experimental results validate the feasibility of the proposed on-line measuring method. On-line metrology is widely applied in manufacturing which not only contributes to improving the working efficiency but also helps to ensure the surface quality of machined parts. However, for cylindrical workpieces with discontinuous surface such as spline shaft, the accurate measurement method is still the key challenge. This paper proposes a novel in-process measuring method of geometrical characteristics of discontinuous surface based on extreme value memory strategy and predictive regression algorithm. The invalid measured size data in the keyways can be effectively replaced by the computed data. Based on the in-process measuring method and the established evaluation mathematical models, the roundness of the manufacturing workpieces can be obtained by using the continuous and valid measurement data. Experimental investigations for machining multi-keyway shaft are carried out based on the proposed method. The experimental results show that the proposed measurement method has a high accuracy for discontinuous surfaces. [ABSTRACT FROM AUTHOR]