The accurate detection and diagnosis of faults are critical for maintaining optimal operation and ensuring the reliability of industrial processes. Notably, the topic of online fault detection and diagnosis has recently presented a significant challenge. This work mainly deploys a neural network technique for the comprehensive detection and diagnosis of faults within the Tennessee Eastman Process (TEP) on a low-computational power system, the Raspberry Pi board. The devolved methodology showcases a remarkable level of accuracy (94.50%) in diagnosing the various TEP faults, affirming its robustness and effectiveness. To elevate the practical applicability of the proposed approach, a meticulous investigation into the implementation of the suggested approach on a Raspberry Pi 4 card was undertaken. The successful realization of this implementation not only highlights the adaptability of the approach but also paves the way for its seamless integration into practical industrial applications.

The key idea behind this study is to integrate a moving window dynamic PCA (MW-DPCA) methodology for fault detection within the Tennessee Eastman process (TEP) into a low-computational power system, the Raspberry Pi 4 card, for real-time application. Indeed, the paramount importance of real-time fault detection (FD) in intricate industrial processes presents a critical challenge. Various data-driven techniques have been developed to ensure safety, maintain operational stability, and optimize productivity in such processes. Principal Component Analysis (PCA) is a fundamental data-driven technique that utilizes dimensionality reduction to extract the most informative features from high-dimensional data, simplifying analysis and potentially revealing underlying fault patterns. However, PCA primarily focuses on static relationships and may miss crucial temporal dynamics for fault identification. This is where dynamic PCA (DPCA) excels. By incorporating lagged values of variables, DPCA captures the temporal evolution of features, enabling a more comprehensive understanding of process behavior and improving the detection of faults involving dynamic changes. In order to address the stochastic measurements, a moving average filter tool is also employed. The results obtained and the successful realization of this implementation demonstrate the adaptability of the approach and pave the way for its seamless integration into practical industrial applications.

This study presents an embedded system (ES) designed for fault detection and diagnosis in grid-connected photovoltaic (GCPV) systems using transient regime analysis. The primary aim of transient regime analysis is to facilitate real-time decision-making, especially during critical faults. A neural network classifier, incorporating a Genetic Algorithm for automated hyperparameter optimization, is developed for GCPV fault classification. These classifiers are seamlessly integrated into a Raspberry Pi 4 platform for fault diagnosis in GCPV systems. Both simulation and experimental results substantiate the ES's viability for fault diagnosis in the examined GCPV system, achieving high accuracy and enabling prompt decision-making to enhance the reliability and safety of GCPV systems.

The evolution of embedded systems has demonstrated their reliability as a solution for monitoring and controlling industrial systems, particularly in renewable energy conversion systems like photovoltaic (PV) energy. The increasing adoption of PV systems highlights the critical need for effective fault diagnosis to ensure their reliable operation. In this paper, we present a novel fault diagnosis approach utilizing Long Short-Term Memory (LSTM) networks optimized through Bayesian optimization techniques. Our methodology is implemented on a Raspberry Pi platform, demonstrating the feasibility of deploying sophisticated fault diagnosis algorithms in resource-constrained environments. Through extensive experiments, we demonstrate the effectiveness of our approach to accurately diagnose faults in grid-connected photovoltaic systems, thereby improving the reliability and efficiency of integrated environmental monitoring systems.The obtained results highlight the potential of combining advanced deep learning techniques with embedded systems to address complex diagnostic challenges, as demonstrated by achieving a 100% accuracy rate.

Grid Connected Photovoltaic (GCPV) sys-tems has been a rising research area in the industry fields. Therefore the high reliability, performance and safety operation of GCPV systems has become a high priority. Thus, it is important of developing an intel-ligent fault detection and diagnosis method that aims at increasing the efficiency of these systems. Therefore, the present study proposes an enhanced intelligent fault diagnosis approach. In the developed procedure, the measured normal and faulty data are applied together to extract the relevant feature, reduce the impact of noise, and then special features are fed to a neural network classifiers. To do that, a Multiscale Principal Component Analysis (MSPCA)-based Artificial Neural Network (ANN) method is proposed to provide the relia-bility and safety of the GCPV systems. From the GCPV measurements, features are appropriately extracted and scaled through multiscale principal component analysis. The ANN classifier is used in classifying twenty-one faults that can occur in GCPV systems operating under different working conditions. The diagnosis results show that the developed MSPCA-based ANN method not only able to detect faults, but also can effectively distinguish between different kinds of faults.

Safe production is of great significance in the process industry like the wind energy conversion (WEC) systems. An unexpected fault in part of the WEC system can damage the entire mechanical system, resulting in huge economic losses and even catastrophic failures. Therefore, this paper proposes an effective neural networks-based ensemble approach for fault de-tection and diagnosis (FDD) of WEC systems. The main contributions are twofold: first, an ensemble learning technique based on the combination of different neural network (ANN, CFNN, and GFNN) into one optimal model are developed in order to distinguish between the different WEC systems operating modes. Then, in order to enhance the results in terms of computation time and storage cost, a reduced version of the proposed neural network-based ensemble technique is presented. The main idea behind this proposal is to use the Hierarchical K-means (H-K-means) clustering to extract only the most significant samples from raw data. Then, the reduced data are introduced as inputs to the proposed neural network-based ensemble technique method to deal with the problem of fault classification. The experimental results demonstrated the feasibility and effectiveness of the proposed FDD techniques.

In this paper, new fault detection and di-agnosis (FDD) techniques dealing with uncertainties in wind energy conversion (WEC) systems are proposed. The uncertainty is addressed by using the interval-valued data representation. The main contributions are twofold: first, to simplify the Recurrent Neural Network (RNN) model in terms of training and computation time and storage cost as well, a reduced version of RNN is proposed. Reduced RNN is established on the H-K-means algorithms to treat the correlations between samples and extract a reduced number of observations from the training data matrix. The main idea behind using H-K-means algorithms for dataset size reduction is to simplify the RNN model in terms of training and computation time. Second, two reduced RNN-based interval-valued data techniques are proposed to distinguish between the different WEC system operating modes. Therefore, two reduced RNN-based interval centers and ranges and interval upper and lower bounds techniques are proposed to deal with the WEC system uncertainties. The presented results confirm the high feasibility and effectiveness of the proposed FDD techniques.

PV systems are prone to failure owing to aging and external/environmental factors. These failures can affect a range of system components, such as PV modules, connecting lines, and converters/inverters, re-sulting in decreased efficiency, performance, and even system failure. As a result, problem detection and diag-nosis (FDD) is an important issue in high-efficiency grid-connected PV systems. Deep learning techniques are the most well-known data-driven methodologies. The biggest advantage of deep learning algorithms, in diagnosis, are learning effectiveness, intelligent FDD becomes more effective. This paper therefore presents a comparative study of FDD based deep learning. The techniques include the Convolutional Neural Network (CNN) and Long Short Time Memory (LSTM). Finally, the FDD based frameworks are implemented using simulated PV data. The diagnosis results show that the CNN and LSTM-based fault diagnosis methods are able to detect and diagnose faults under different operating modes.

PV systems are subject to failures during their operation due to the aging effects and exter-nal/environmental conditions. These faults may affect the different system components such as PV modules, connection lines, converters/inverters, which can lead to a decrease in the efficiency, performance, and fur-ther system collapse. Thus, a key factor to be taken into consideration in high-efficiency grid-connected PV systems is the fault detection and diagnosis (FDD). The most well-known data-driven methods are Deep Learning (DL) approaches. The biggest advantage of DL algorithms, in diagnosis, are that they try to learn high- level features from PV data in a high-order, non-linear and adaptive manners. Then, the fault is classified using soft-max activation function. This work therefore presents a comparative study of FDD based DL techniques. These techniques include Artificial Neural Network (ANN), Recurrent Neural Network (RNN) and Long-Short Term Memory (LSTM). The DL techniques-based fault diagnosis are implemented using an emulated Grid-Connected PV (GCPV) system. The classification results for the pretrained DL models is exhibited and performance of the models are evaluated.

This paper proposes an effective fault detection and diagnosis (FDD) paradigm in Wind Energy Converter (WEC) Systems. The developed FDD frame-work merges the benefits of kernel principal component analysis (KPCA) model and bidirectional long short-term memory (BiLSTM) feature classifier. KPCA is used to extract and select the most effective features. While, BiLSTM is used for classification purposes. The proposed KPCA-based BiLSTM approach involves two main steps; feature extraction and selection and fault classification. It is tackled in such a way that KPCA model is developed in order to select and extract the more efficient features where the final features are fed to BiLSTM to distinguish between different working modes. Different simulation scenarios are considered in this study in order to show the robustness and performances of the developed technique when compared to the conventional FDD methods.

Abstract Automated classification of magnetic resonance brain images (MRIs) is a hot topic in the field of medical and biomedical imaging. Various methods have been suggested recently to improve this technology. In this paper, to reduce the complexity involved in the medical images and to ameliorate the classification of MRIs, a novel 3D magnetic resonance (MR) brain image classifier using kernel principal component analysis (KPCA) and support vector machines (SVMs) is proposed. Experiments are carried out using A deep multiple kernel SVM (DMK-SVM) and a regular SVM. An algorithm entitled SVM–KPCA is put forward. Its main task is to classify a brain MRI as a normal brain image or as a pathological brain image. This algorithm, firstly, adopts the discrete wavelet transform technique to extract features from images. Secondly, KPCA is applied to decrease the dimensionality of features. SVM is then applied to the reduced data. A K-fold cross-validation strategy is used to avoid overfitting and to ameliorate the generalization of the SVM–KPCA algorithm. Three databases are used to validate the suggested SVM–KPCA method. Three conclusions are obtained from this work. First, KPCA is highly efficient in increasing the classifier’s performance compared with similar algorithms working on the proposed database. Second, the SVM–KPCA algorithm performs well in differentiating between two classes of medical images. Third, the approach is robust and might be utilized for other MRIs. This proposes a significant role for computer aided diagnosis analysis systems used for clinical practice.

This study proposes a new representation of discrete Non‐linear AutoRegressive with eXogenous inputs (NARX) model by developing its coefficients associated to the input, the output, the crossed product, the exogenous product and the autoregressive product on five independent Laguerre orthonormal bases. The resulting model, entitled NARX‐Laguerre, ensures a significant parameter number reduction with respect to the NARX model. However, this reduction is still subject to an optimal choice of the Laguerre poles defining the five Laguerre bases. Therefore, the authors propose to use the genetic algorithm to optimise the NARX‐Laguerre poles, based on the minimisation of the normalised mean square error. The performances of the resulting NARX‐Laguerre model and the proposed optimisation algorithm are validated by numerical simulations and tested on the benchmark Continuous Stirred Tank Reactor.

In this study, the authors investigate the parametric complexity reduction of the Meixner‐like model for linear discrete‐time system representation. The use of the Meixner‐like functions is more suitable than the use of Laguerre functions and Kautz functions especially when the system have a slow initial onset or delay. The coefficients of the Meixner‐like model can be estimated recursively from input–output data by the new representation. Noting that the selection of an arbitrary pole for the Meixner‐like functions can raise the parameter number of the Meixner‐like model. However, when the pole is set to its optimal value, an optimal expansion of transfer functions is produced. Therefore an optimisation technique is developed to generate the optimal Meixner‐like pole, which is achieved by an iterative method, that consists in minimising the mean square error between the system output and the model output. Theoretical analysis and a numerical simulation show the efficiency of the approach.