Search

Accurate prediction of the remaining useful life (RUL) of proton exchange membrane fuel cells (PEMFCs) is essential for maximizing their operational lifespan. However, existing methods often face limitations in two key areas: long-term prediction (beyond 168 h, or one week) and adaptability to varying operating conditions. To address these challenges, we propose a novel self-adaptive digital twin (SADT) model for RUL prediction of PEMFCs. Our approach uniquely integrates a deep convolutional neural network to generate robust health indicators (HIs) that maintain consistent monotonicity across diverse operating conditions. Additionally, we introduce a novel quantile Huber loss (QH-loss) function to enhance prediction accuracy and incorporate a transfer learning technique to improve adaptability under varying operational scenarios. Experimental results on PEMFC degradation datasets demonstrate that our method outperforms state-of-the-art techniques in long-term prediction accuracy, highlighting its potential to significantly extend fuel cell lifetimes. • A novel self-adaptive digital twin for predicting RUL of the PEMFC. • Transfer Learning enhanced RUL prediction accuracy and adaptability. • Universal monotonic HI applicable to various PEMFC operating conditions. [ABSTRACT FROM AUTHOR]

Accurately predicting the degradation and remaining useful life (RUL) of Proton Exchange Membrane Fuel Cells (PEMFCs) is crucial for enhancing their reliability, particularly in automotive and energy sectors. Traditional models often fail to capture the complex, non-linear degradation patterns of PEMFCs, resulting in suboptimal predictions. This study addresses these technological gaps by introducing a novel approach, the Self-Attention Temporal Dual Discriminator Generative Adversarial Network (SAT-DD-GAN). This model is designed to overcome the limitations of existing methods by integrating advanced Long Short-Term Memory (LSTM) networks for time-series analysis, a self-attention mechanism to focus on critical degradation signals, and dual discriminators to improve prediction accuracy and robustness. By addressing the gap in handling long-term dependencies and complex degradation patterns, the proposed SAT-DD-GAN bridges the disconnect between traditional models and the real-world degradation behavior of PEMFCs. The SAT-DD-GAN achieved groundbreaking predictive performance, with the lowest RMSE (0.983E-3 for static load, 1.012E-3 for dynamic load), MAPE (0.036E-1 and 0.253E-1), MAE (0.776E-3 and 0.806E-3), and the highest R2-score (0.9998 and 0.9983), alongside zero relative error for both datasets. These results demonstrate the model's ability to capture the intricate degradation dynamics more effectively than existing state-of-the-art methods. By filling the technology gap of accurately modeling PEMFC degradation under varying load conditions, this study establishes a new standard for PEMFC prognostics and paves the way for future innovations in optimizing fuel cell performance and longevity. • Introduced a novel SAT-DD-GAN model for accurate prediction of PEMFC stack voltage degradation and Remaining Useful Life (RUL). • Innovatively combined LSTM, self-attention, and dual discriminator mechanisms for enhanced predictive robustness. • Achieved superior accuracy, with RMSE as low as 0.000983 (static load) and R2-score of 0.9998 (static load). • Demonstrated advanced handling of static and dynamic loads, outperforming recent models in key prediction metrics. • Set a new benchmark in predictive maintenance for PEMFCs, offering potential advancements in neural network-driven prognostics. [ABSTRACT FROM AUTHOR]

Durability is one of the important factors limiting the large-scale commercial application of proton exchange membrane fuel cells (PEMFC) systems, and accurate lifetime prediction of the PEMFC is beneficial to enhance their durability and reliability. In this paper, a fusion prognostic model based on bidirectional temporal convolutional network, bidirectional gated recurrent unit, and extreme learning machine (BiTCN-BiGRU-ELM) is proposed to realize short-term degradation prediction and remaining useful life estimate of the PEMFC system. The nature-inspired metaheuristic of the crested porcupine optimizer algorithm is introduced to optimize the parameters of the proposed BiTCN-BiGRU-ELM method. The effectiveness of the hybrid method is verified under three current conditions: steady state, quasi-dynamic, and dynamic. The fusion model can significantly improve prediction performance compared to traditional machine learning methods. This method is of great significance for online lifetime prediction and health management of PEMFC. • A PEMFC degradation prediction model of BiTCN-BiGRU-ELM was proposed. • Improved prediction performance of ELM by BiTCN-BiGRU feature extraction. • Optimal model parameters were determined using a CPO algorithm. • The effect of different training lengths on prediction accuracy was investigated. • This fusion model can be applied to RUL's prediction under dynamic conditions. [ABSTRACT FROM AUTHOR]

To achieve the 8000-h proton exchange membrane fuel cell stack (PEM FCS) life target set by the U.S DoE and promote fuel cell hybrid electric vehicles (FCHEVs) massive introduction in the automotive market, using multi-fuel cell stack (MFCS) systems instead of single-fuel cell stack systems seems to be an interesting solution that deserves to be explored. MFCS systems' concept combines several small FCSs modules instead of using a single high-powered FCS module. The modularity in such systems can be exploited through energy management to improve their durability and extend their good energy-efficiency power range. However, FCSs' multiplicity makes it challenging to implement effective energy management strategies (EMSs). This paper proposes a remaining useful life (RUL) prognostic-based EMS to extend MFCS systems' lifetime while keeping their hydrogen consumption reasonable. For this purpose, a prognostic algorithm is developed to predict PEM FCSs' RUL in real-world automotive application scenarios. Then a rule-based EMS allocates the demand between stacks using prognostic results. The proposed strategy's performance is evaluated on a hybrid MFCS/battery system using Matlab/Simulink's environment. Simulation results show that implementing the proposed strategy instead of conventional EMSs can extend MFCS systems' lifetime by at least a factor of 2.35 while keeping their hydrogen consumption reasonable. © 2001 Elsevier Science. All rights reserved. • A post-prognostic decision-making strategy is proposed for MFCS systems. • RUL predictions are performed in real-world automotive scenarios. • The MFCS system's lifetime is significantly extended with the proposed EMS. • RUL is a suitable parameter for adaptive energy management decision-making. [ABSTRACT FROM AUTHOR]

Prediction of Remaining Useful Life (RUL) for Rolling Element Bearings (REB) has attracted widespread attention from academia and industry. However, there are still several bottlenecks, including the effective utilization of multi-sensor data, the interpretability of prediction models, and the prediction across the entire life cycle, which limit prediction accuracy. In view of that, we propose a knowledge-based explainable life-cycle RUL prediction framework. First, considering the feature fusion of fast-changing signals, the Pearson correlation coefficient matrix and feature transformation objective function are incorporated to an Improved Graph Convolutional Autoencoder. Furthermore, to integrate the multi-source signals, a Cascaded Multi-head Self-attention Autoencoder with Characteristic Guidance is proposed to construct health indicators. Then, the whole life cycle of REB is divided into different stages based on the Continuous Gradient Recognition with Outlier Detection. With the development of Measurement-based Correction Life Formula and Bidirectional Recursive Gated Dual Attention Unit, accurate life-cycle RUL prediction is achieved. Data from self-designed test rig and PHM 2012 Prognostic challenge datasets are analyzed with the proposed framework and five existing prediction models. Compared with the strongest prediction model among the five, the proposed framework demonstrates significant improvements. For the data from self-designed test rig, there is a 1.66 % enhancement in Corrected Cumulative Relative Accuracy (CCRA) and a 49.00 % improvement in Coefficient of Determination (R2). For the PHM 2012 datasets, there is a 4.04 % increase in CCRA and a 120.72 % boost in R2. [Display omitted] • Advocate explainable knowledge-data integration for RUL prediction throughout the life cycle, leveraging multi-sensor data. • Introduce a novel unsupervised learning framework that integrates IGCA and CMSACG to construct HI. • A two-stage prediction framework, fusing MCLF and BR-GDAU, is proposed for predicting the life cycle RUL. • The effectiveness of the proposed framework is validated using both Self-Designed Experiment-Derived and PHM 2012 Prognostic Challenge Bearing Datasets. [ABSTRACT FROM AUTHOR]

The durability is a key factor of fuel cell in large-scale application of automobiles. It is great important to real the lifetime and health status of fuel cells, especially in vehicle driving conditions. Based on the experimental data, this paper used the model-based method and data-driven method to predict the life of fuel cells, respectively. The model-based method established the equivalent circuit model according to the mapping law of electrochemical impedance spectroscopy (EIS) and the parameters of the equivalent circuit model were identified by newly considering the phenomenon of voltage recovery, revealing the strong linear correlation between impedance and voltage. The data-driven method selected a combinatorial optimization PSO-LSSVM algorithm and compared it with the prediction results of MATLAB's time series toolbox. The results show the mean square error (MSE), End of Life (EOL) of the model-based method, the time series toolbox and PSO-LSSVM algorithm are 0.019, 8.2 × 10−3 and 4.42 × 10−5, the 260h, 438h and 427h, respectively. Comparing with the real EOL of 421h, the error of PSO-LSSVM algorithm is only 6 h, which means the remaining useful life can be predicted with high accurately. Thus this study provides a reference for fault identification and life prediction of fuel cell health management. • The equivalent circuit model is established to realize the segmented voltage prediction and analysis of fuel cells. • The MSE of combinatorial optimized algorithms PSO-LSSVM to predict voltage is only 4.42 × 10−5. • The method in paper provides a reference for fault identification and life prediction of PEMFC's health management. [ABSTRACT FROM AUTHOR]

Proton exchange membrane fuel cells (PEMFCs) offer high energy conversion rate and low pollution, while low durability of PEMFCs hinder the further development. Accurately predicting degradation trends and remaining useful life (RUL) are crucial to timely detecting problem and improving PEMFCs durability. In this paper, an enhanced transfer learning (Add-TL) based on adder network (AdderNet) and deep domain confusion is proposed for the prediction of PEMFCs. To reinforce the ability of extracting key features from extensive data, AdderNet with attention mechanism is applied. Furthermore, the fused deep domain confusion employs existing data to achieve higher prediction accuracy. The experimental results demonstrate that Add-TL outperforms other classical methods in predicting degradation trend, the optimal adjusted R-square (R2) values for the two datasets are 0.99879 and 0.99906, respectively. When Add-TL is applied to RUL prediction, the accuracy scores for two datasets reach 0.995, which further validating the applicability of Add-TL in PEMFCs prediction. • AdderNet applied for the first time to predict degradation trends and RUL of PEMFCs. • The method proposed in this paper integrates deep learning and transfer learning. • The incorporation of transfer learning the connections between data, aiming to improve the accuracy of RUL prediction. • Add-TL predicted static and dynamic conditions, providing a comprehensive validation. • Add-TL achieves best performance when training on datasets with varying lengths. [ABSTRACT FROM AUTHOR]

This paper proposed a prediction algorithm for the degraded actuator taking into account the impact of estimation error of hidden index in the closed-loop system. To this end, a unified prediction framework is established to evaluate the hidden degradation information and recursively update the degradation model parameters simultaneously. The advantage is that the prediction framework can comprehensively compensate the estimation error of hidden degradation index caused by system uncertainty. To jointly estimate the degradation information in avoidance of the impact of system uncertainty, a modified adaptive Kalman filter is designed, and the proof of stability is provided. With the priori estimate from the filter, the degradation model parameters are updated by the inverse filtering probability based on Bayes' theorem. It is followed by the computation of the remaining useful life (RUL) prediction utilizing aforementioned hidden degradation information and the latest degradation model. The effectiveness of the proposed RUL prediction algorithm is demonstrated by the degraded actuator in the continuous casting process. • A novel prediction framework is designed for degraded actuator in closed-loop system. • Adaptive degradation model is established by modified adaptive Kalman filter. • The estimation error of hidden degradation index is integrated. • The stability of optimized adaptive Kalman filter is proven. • Recursive prediction algorithm is performed without full life data. [ABSTRACT FROM AUTHOR]

Schedule maintenance activities for hydrogen fuel cell vehicles are aimed to keep them in better condition in long-term operation. In this paper, we develop a reinforcement learning based schedule maintenance strategy, which is used to comprehensively consider the balance between safety and maintenance cost, so as to find the optimal maintenance strategy. A multi-level framework is established with remaining useful life of key components, fault tree analysis, and logistics cost model to generate an exploration environment that can express the operational stability, the failure rate of components, and the cost of repair and storage. The trained agent combines a deep neural network to explore the optimal strategy under the dynamic reliability of key components. Finite steps are the constraints. Safety rates, maintenance cost, and episode of operation are incorporated into a multi-objective reward function. The hydrogen supply circuit of fuel cell vehicle is simulated via Python. The trained agent is compared with traditional time-based schedule maintenance strategy and corrective maintenance strategy. The results show that the reinforcement learning based schedule maintenance agent has optimized the total reward, cost control and accident rate by 77%, 59% and 5%, respectively, compared with the traditional time-based schedule maintenance strategy. In the safe operation management of fuel cell vehicles, efficient and stable decision-making ability has been achieved. • An adaptive maintenance strategy of deep reinforcement learning is proposed. • An exploration environment of FCV is established. • The deep Q network is proposed to solve the maintenance problem. • The efficient and stable decision-making ability is much better than time-based scheduled maintenance strategy. [ABSTRACT FROM AUTHOR]

Predicting the Remaining Useful Life of a machine components or systems is a pivotal technology in condition-based maintenance and essential for ensuring the reliability and safety in various production-engineering applications. The influx of extensive industrial data has notably enhanced the efficacy of data-driven Remaining Useful Life prediction models, especially deep learning models. One of the promising deep learning model architectures is the Transformer-based model with the Self-Attention mechanism at its core. However, inherent limitation arise when applying Self-Attention to high- frequency time-series data with large window sizes. Due to its high computational complexity, hardware limitations hinder the practical implementation of Transformer with Self-Attention in production-engineering applications. This study looks into the utilization of alternative Attention modules with reduced complexity, making it more applicable to high-frequency time-series data. In order to allow comparability, this study uses the well-known C-MPASS dataset for benchmarking Remaining Useful Life approaches. Although this dataset does not consist of high-frequency data, it demonstrates the usefulness of alternative Attention modules without noteworthy losses in model accuracy. [ABSTRACT FROM AUTHOR]

The remaining useful life (RUL) is one of the critical factors for proton exchange membrane fuel cells (PEMFCs), as it is hindering the commercialization of PEMFCs in various fields. In this research, a novel deep learning (DL) algorithm, i.e. the Non-stationary Transformer is newly applied to the RUL prediction of a PEMFC stack. For better prediction accuracy, the discrete wavelet transform (DWT) is utilized to denoise the original data before the normalization process. Prediction results of the proposed Non-stationary Transformer-based PEMFC RUL prediction method are analyzed under different lengths of training datasets, under different time steps ahead prediction, and for different PEMFC End of Life (EoL) threshold values and compared to those of other DL-based prediction methods including the long short-term memory (LSTM)-based and echo state network (ESN)-based methods. Results show that the proposed method outperforms the LSTM-based and ESN-based methods for all different cases studied in this research. • The Non-stationary Transformer is newly applied to PEMFC RUL prediction method. • The DWT method is utilized to denoise original data to accurate the prediction. • The proposed RUL prediction method is compared to LSTM and ESN-based methods. • RUL prediction methods are analyzed under various dataset lengths and time steps. • The proposed RUL prediction method outperforms LSTM and ESN-based in accuracy. [ABSTRACT FROM AUTHOR]

Remaining useful life (RUL) prediction and degradation assessment are pivotal components of prognostic and health management (PHM) and represent vital tasks in the implementation of predictive maintenance for bearings. In recent years, data-driven PHM techniques for bearings have made substantial progress through the integration of deep learning methods. However, modeling the temporal dependencies inherent in raw vibration signals for both degradation assessment and RUL prediction remains a significant challenge. Hence, we propose a joint optimization architecture that uses a temporal convolutional auto-encoder (TCAE) for the degradation assessment and RUL prediction of bearings. Specifically, the architecture includes a sequence-to-sequence model to extract degradation-sensitive features from the raw signal and utilizes temporal distribution characterization (TDC) and a nonlinear regressor to determine the degradation stages and predict RUL, respectively. Our framework integrates the tasks of degradation assessment and RUL prediction in a unified, end-to-end manner, using raw signals as input, resulting in high RUL prediction accuracy (RMSE = 0.0832) on publicly available and self-built datasets. Our approach outperforms state-of-the-art methods, indicating its potential to significantly advance the field of PHM for bearings. • A novel temporal convolutional auto-encoder (TCAE) is proposed. • TCAE jointly optimizes remaining useful life prediction and degradation assessment. • Degradation-sensitive features are extracted with joint supervised and unsupervised learning. • Experimental case studies verified validity and superiority. [ABSTRACT FROM AUTHOR]

In order to realize the remaining useful life (RUL) prediction of mechanical equipment under different operating conditions, a domain adaption residual separable convolutional neural network (DRSCN) model is proposed in this paper. In the DRSCN model, instead of the traditional convolutional layer, a residual separable convolutional module is developed to improve the feature extraction ability of the model. Moreover, a multi-kernel maximum mean discrepancy metric function and an adversarial learning mechanism are embedded in the DRSCN model to enhance its ability to resist domain shifts, thus improving the cross-domain RUL prediction accuracy of the model. The effectiveness of the DRSCN model is verified on an aircraft engine dataset. The experimental results show that the proposed model can realize high-accuracy RUL prediction. [Display omitted] • A DRSCN model is proposed for cross-domain RUL prediction. • MMD is embedded in the DRSCN model to enhance its generalization performance. • Adversarial learning mechanism is adopted to realize domain adaption of the model. • A residual separable convolution module is proposed to replace convolutional layer. [ABSTRACT FROM AUTHOR]

We introduce a general path Gamma-Gamma model for degradation measures, related to different inspection times functions, obtaining flexible forms of degradation paths. One important contribution of the proposed model is the way the degradation rate is modeled. It is composed of two random components: one random effect quantifying the specific features of each device and a dynamic effect, common to all devices, measuring the impact of the environment on the degradation. The model is identifiable under mild constraints. Besides producing gains regarding the interpretability of the parameters, this decomposition generates a parsimonious model, reducing computational time. The relation between degradation and failure time is obtained, allowing a computational approximation for the failure time distribution. The model performance is evaluated through simulation, helping to guide the prior specifications to model identification. The proposed model is applied to analyze fatigue crack growth data. We compare the proposed model with the traditional linear Weibull model and with a dynamic linear Normal model. Results show that the proposed methodology is competitive in predicting failure times and estimating the remaining useful life. • We develop a dynamic general path Gamma-Gamma model for positive degradation measurements. • The degradation rate is a function of unit-specific and environmental effects. • The environmental effects evolve over time considering a Markovian equation. • Different functions relate the degradation measures and inspection times. • Analysis of simulated and true datasets pointed out that it is a competitive model. [ABSTRACT FROM AUTHOR]

This paper focuses on the implementation of predictive maintenance processes using both physics-based and data-driven methods, specifically applied to a Salvagnini panel bender. The primary objective is to enhance production data analysis through digital twin predictions, enabling the calculation of process parameters that cannot be directly measured. Two key aspects are addressed in this study: Firstly, two global measures of the effective cumulative load on a machine during its lifetime are formulated. These measures can be used as criteria for estimating the health condition of a machine and for comparing different machines. To achieve this, dimensionless parameters are derived from actual production data. Secondly, a remaining useful life (RUL) model is developed for two machine components relying on production data analysis and digital twin predictions. The first component under consideration is a hybrid hydraulic actuator. Internal leakage prediction for this component is performed based on its data history, resulting in a corresponding trend curve. Additionally, wear of critical bearings is calculated using an analytical model that depends on the production history log. Through the proposed hybrid approach, this paper aims to enhance the predictive maintenance process by leveraging both physics-based and data-driven methodologies. The findings from this study can offer valuable insights for optimizing maintenance strategies and improving the overall efficiency of similar manufacturing systems. [ABSTRACT FROM AUTHOR]

Within the paradigm of industry 5.0, manufacturing systems are seeking for human-centred production, where the operator finds high-level supervision tasks. In this context, low-level decision making should be performed by machines themselves. In this paper, a hybrid prognosis algorithm is developed to automatically inspect the cutting edges of drill-bits and to predict their Remaining Useful Life (RUL) and the associated probability density function. The solution relies on the automatic measurement of flank wear through convolutional filtering and edge detection. Prognosis exploits particle filter, which updates multi-layer perceptron with online data, to adaptively predict drill-bits RUL. The solution reduces the experimental preliminary run-to-failures needed for training standard machine learning algorithms, exploiting them in a real-time adaptive scenario, while predicting tool RUL under untested and variable cutting process operations. The algorithm uses direct wear observations, taken during set-up times (e.g., tool changes, workpiece change), thus not interfering with the process. [ABSTRACT FROM AUTHOR]

In recent years, predictive maintenance tasks, especially for bearings, have become increasingly important. Solutions for these use cases concentrate on the classification of faults and the estimation of the Remaining Useful Life (RUL). As of today, these solutions suffer from a lack of training samples. In addition, these solutions often require high-frequency accelerometers, incurring significant costs. To overcome these challenges, this research proposes a combined classification and RUL estimation solution based on a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. This solution relies on a hybrid feature extraction approach, making it especially appropriate for low-cost accelerometers with low sampling frequencies. In addition, it uses transfer learning to be suitable for applications with only a few training samples. [ABSTRACT FROM AUTHOR]

Hydrogen fuel has shown promise as a clean, alternative fuel aiding in the reduction of fossil fuel dependence within the transportation sector. However, hydrogen refueling stations and infrastructure remains a barrier and are a prerequisite for consumer adoption of low-cost and low-emission fuel cell electric vehicles (FCEVs). The costs for FCEV fueling include both station capital costs and operation and maintenance (O&M) costs. Contributing to these O&M costs, unscheduled maintenance is presently more costly and more frequent than for similar gasoline fueling infrastructure and is asserted to be a limiting factor in achieving FCEV customer acceptance and cost parity. Unscheduled maintenance leads to longer station downtime, therefore, causing an increase in missed fueling opportunities, which forces customers to seek refueling at other operable stations that may be significantly farther away. This research proposes a framework for a hydrogen station prognostics health monitoring (H2S PHM) model that can minimize unexpected downtime by predicting the remaining useful life for primary hydrogen station components within the major station subsystems. The H2S PHM model is a data-driven statistical model, based on O&M data collected from 34 retail hydrogen stations located in the U.S. The primary subcomponents studied are the dispenser, compressor, and chiller. The remaining useful life calculations are used to decide whether or not maintenance should be completed based on the prediction and expected future station use. This paper presents the background, method, and results for the H2S PHM model as for a means for improving station availability and customer confidence in FCEVs and hydrogen infrastructure. • Hydrogen station reliability is a key contributor to market success. • Mean fills between failure and station availability are lower than conventional technologies. • A prognostics health monitoring system can improve hydrogen station availability. • Estimated remaining useful life of primary components assists in maintenance decisions. [ABSTRACT FROM AUTHOR]

Accurate prediction of remaining useful life (RUL) can ensure the safety and reliability of power batteries during operation, reduce the failure rate and operating costs, and enhance user experience. However, battery degradation is a complex, nonlinear dynamic process that is difficult to fully comprehend and predicting RUL remains a significant challenge. To address this issue, the hybrid data-driven prediction model PCA-CNN-BiLSTM was proposed in this paper, which combines principal component analysis (PCA), convolutional neural network (CNN), and bi-directional long short-term memory (Bi-LSTM) network. PCA was applied to downscale and whiten the health factor (HF) to maximize the extraction of important features of lifespan decay, while reducing the correlation between features. The convolution kernel of the CNN was used to explore the local region feature information of the input information and search for the common patterns among the neighboring data. Additionally, the model parameters and computational efforts were reduced through pooling. Finally, battery RUL prediction was achieved using Bi-LSTM, which has the advantages of effectively enhancing model accuracy and reducing the risk of over-fitting by taking into account both past and future data. The performance of the proposed model was evaluated utilizing NASA and CALCE's battery datasets, and the results suggest that it exhibits a high level of accuracy across various datasets. Compared to other methods, the PCA-CNN-BiLSTM method has the best performance indicators for predicting battery RUL, including RMSE, MAE, MAPE, RUL e and DOL. This indicates that the proposed model has better fitting performance, accuracy, robustness, and generalization ability. [Display omitted] • Propose a novel battery RUL prediction method based on hybrid data-driven model. • Nineteen health factors were extracted, including segmented discharge time and temperature. • The effectiveness of the PCA-CNN-BiLSTM was verified on multiple public datasets. [ABSTRACT FROM AUTHOR]

A robust model based on stochastic processes with Copula function is developed for remaining useful life (RUL) prediction of subsea pipeline subject to corrosion-fatigue degradation. Gamma-based stochastic degradation process is used to simulate the corrosion degradation of pipeline, and fatigue crack propagation is estimated with Wiener process. The Particle Filter (PF) and Kalman Filter (KF) are utilized to update the model parameters. Then, an improved joint distribution model incorporating marginal distributions and copula function is presented to capture the complex dependencies between corrosion and fatigue. By establishing an acceptable pipeline failure threshold, a comparison is conducted among three stochastic process models to assess the practicality and effectiveness of the developed robust model. The results indicate that copula-based model has superiority in predictive accuracy with a deviation of 0.5. Considering interaction of corrosion-fatigue has a substantial impact on improving the accuracy of RUL. The present model can significantly contribute to the proactive maintenance and management of subsea pipelines, thereby enhancing operational efficiency and reducing potential risks. [ABSTRACT FROM AUTHOR]

• A new bearing prognosis framework is built upon the time-domain signal preprocessing and physics-based wavelet neural network (WNN). • Empirical mode decomposition (EMD) and statistical time series analysis are adopted as performance multipliers of WNN. • The framework enables the physical interpretation of bearing fault diagnosis and yields excellent predictive performance. • Comprehensive case studies using multiple datasets demonstrate the feasibility of the framework. Rolling bearings are important components in rotating machinery in various industries. Conducting intelligent prognostics for bearing remaining useful life (RUL) prediction plays an important role in the health management and reliability assessment of these machinery systems. While the bearing fault prognosis using measured system response through machine learning techniques has attracted significant attention and demonstrated promising potential, completely data-driven approaches face some challenges in understanding complex domains with data efficient learning. In this research, we develop an integral bearing fault prognosis framework informed by the well-designed time-domain signal preprocessing to conduct the bearing RUL prediction. This framework is built upon physical feature-oriented signal preprocessing and an associated wavelet neural network (WNN). Sequential procedures of time-domain analyses are proposed to extract the physical features of bearing degradation. Empirical mode decomposition (EMD) is specifically chosen owing to its capability of handling bearing fault signals that are nonstationary with underlying nonlinearities. The WNN is a new class of neural networks that combines the classic neural network and wavelet analysis. Here in this research a WNN model built upon B-spline mother wavelet to suit the preceding EMD-based signal preprocessing is constructed to process the bearing degradation features extracted and then identify their correlation with bearing RUL. The combination of EMD-based signal preprocessing and B-spline mother wavelet in WNN enables the network to learn the input-output correlation in a physical sense. Case studies formulated upon multiple bearing datasets and multiple benchmark methods are carried out to systematically validate the proposed framework. The results consistently demonstrate prediction accuracy and performance robustness. [ABSTRACT FROM AUTHOR]

• Physics-Constrained Bayesian Neural Network. • Failure prognostics of lithium-ion batteries. • Comparative study of physics-based, data-driven, and physics-constrained models. In order to predict the remaining useful life (RUL) of lithium-ion batteries, a capacity degradation model may be developed using either simplified physical laws or machine learning-based methods. It is observed that even though degradation models based on simplified physical laws are easy to implement, they may result in large error in the application of failure prognostics. While data-driven prognostics models can provide more accurate degradation forecasting, they may require a large volume of training data and may invoke predictions inconsistent with physical laws. It is also very challenging for existing methods to predict the RUL at the early stages of battery life. In this paper, we propose a Bayesian physics-constrained neural network for battery RUL prediction by overcoming limitations of the current methods. In the proposed method, a neural differential operator is learned from the first 100 cycles of data. The neural differential operator is modeled with a Bayesian neural network architecture that separates the fixed history dependence from the time dependence to isolate epistemic uncertainty quantification. Using the battery dataset presented in the paper by Severson et al. as an example, we compare our proposed method with a simplified physics-based degradation forecasting model and two data-driven prognostics models. The results show that the proposed physics-constrained neural network can provide more accurate RUL estimation than the other methods with the same group of training data. Most importantly, the proposed method allows for RUL prediction at earlier stages of the battery life cycle. [ABSTRACT FROM AUTHOR]

Accurately evaluating the remaining useful life (RUL) of aircraft engines is crucial for ensuring operational safety and reliability, and serves as a critical foundation for making informed maintenance decisions. In this paper, a novel prediction framework is proposed for forecasting the RUL of engines, which utilizes a dual-frequency enhanced attention network architecture built upon separable convolutional neural networks. First, the information volume criterion (IVC) index and information content threshold (CIT) equation are designed, which are applied to quantitatively quantify the degradation features of the sensor and remove redundant information. In addition, this paper introduces two trainable frequency-enhanced modules, namely the Fourier transform module (FMB-f) and the wavelet transform module (FMB-w), to incorporate physical rules information into the prediction framework, dynamically capture the global trend and local details of the degradation index, and further improve the prediction performance and robustness of the prediction model. Furthermore, the proposed efficient channel attention block generates a unique set of weights for each possible vector sample, which establishes the interdependence among different sensors, thereby augmenting the prediction stability and precision of the framework. The experimental demonstrate that the proposed RUL prediction framework can deliver accurate RUL predictions. • The IVC and CIT are used to quantify and eliminate degradation information. • A dual-frequency block(FEB-f and FEB-w) with trainable parameters is proposed. • FEB-w and FEB-f jointly trained with SDFEPN for dynamic coordination. • The integration of physical rules into the SDFEPN enhances the interpretability. [ABSTRACT FROM AUTHOR]

Estimating the remaining useful life (RUL) of proton exchange membrane fuel cells (PEMFCs) is key to predicting and increasing their durability. This paper investigates six deep learning techniques – long short-term memory (LSTM) networks, gated recurrent units (GRU), 1-D convolutional neural network (CNN)-LSTMs, 1D–CNN–GRUs, 1D–CNN–Bidirectional-LSTMs, and 1D–CNN–Bidirectional-GRUs – for RUL and long-term degradation trend prediction for a five-cell PEMFC stack operated with and without current ripples. The hyperparameters of all models are optimized for minimum root-mean-square error (RMSE) of the stack voltage using tree-structured Parzen estimators (TPE) with hyperband pruning. The obtained prediction results are benchmarked using the coefficient of determination (R2), RMSE, and mean absolute percentage error (MAPE) in stack voltage predictions and relative error (RE) in RUL estimates. The comparative analysis for models trained on 50% datasets reveals that the best R2, RMSE, MAPE, and RE of 98.87E-2, 1.258E-3, 0.285E-1, and 0.0425E-1 are offered by the LSTM model for the stack without current ripples. Further, for the stack with current ripples, the best R2, RMSE, MAPE, and RE of 99.482E-2, 1.77E-3, 0.444E-1, and 0.089E-1 are exhibited by the 1D–CNN–Bidirectional-GRU model. The LSTM model also reveals excellent transfer learning performance on the stack with current ripples with the best R2, RMSE, MAPE, and RE of 98.91E-2, 2.455E-3, 0.362E-1, and 0.044E-1, respectively. [Display omitted] • Six deep learning models are developed for PEMFC RUL and degradation prediction. • Hyperparameters of each model are optimized using TPE for static and dynamic loading. • Static loading models exhibited excellent transfer learning performance. • All tuned models performed superior to recent reported approaches in RMSE. • LSTM and 1D–CNN–Bidirectional-GRU demonstrated best results trained on 50% data. [ABSTRACT FROM AUTHOR]

A good weight initialization is crucial to accelerate the convergence of the weights in a neural network. However, training a neural network is still time-consuming, despite recent advances in weight initialization approaches. In this paper, we propose a mathematical framework for the weight initialization in the last layer of a neural network. We first derive analytically a tight constraint on the weights that accelerates the convergence of the weights during the back-propagation algorithm. We then use linear regression and Lagrange multipliers to analytically derive the optimal initial weights and initial bias of the last layer, that minimize the initial training loss given the derived tight constraint. We also show that the restrictive assumption of traditional weight initialization algorithms that the expected value of the weights is zero is redundant for our approach. We first apply our proposed weight initialization approach to a Convolutional Neural Network that predicts the Remaining Useful Life of aircraft engines. The initial training and validation loss are relatively small, the weights do not get stuck in a local optimum, and the convergence of the weights is accelerated. We compare our approach with several benchmark strategies. Compared to the best performing state-of-the-art initialization strategy (Kaiming initialization), our approach needs 34% less epochs to reach the same validation loss. We also apply our approach to ResNets for the CIFAR-100 dataset, combined with transfer learning. Here, the initial accuracy is already at least 53%. This gives a faster weight convergence and a higher test accuracy than the benchmark strategies. [ABSTRACT FROM AUTHOR]

In order to achieve a fast and accurate prediction for remaining useful life (RUL) of proton exchange membrane fuel cells (PEMFCs), an improved degradation model named RCLMA is developed which integrates the residual convolutional blocks, the long short-term memory (LSTM) units and the multi-head self-attention layers. Firstly, the stack voltages of two PEMFCs are collected, reconstructed and smoothed. Secondly, the RCLMA model is constructed and validated based on the experimental data, and the effects of input size and output size on the prediction performance are studied. Finally, ablation experiments are conducted to investigate the role of each part of the RCLMA model in the prediction. The results show that the RCLMA model reaches the best prediction performance with a root mean square error (RMSE) of 0.01785 and a S c o r e R U L of 0.994589 when the input size is 200 and the output size is 30, and the prediction time does not exceed 5 s. All three parts of the RCLMA model make important contributions to the prediction performance, where the residual part is partial to finding the overall trend of the data, the recurrent part is inclined to obtain the time dependence and the self-attention part plays a good capture of both. [Display omitted] • Effects of input size and output size on prediction performance of RCLMA model are studied. • Ablation experiments are conducted to investigate the role of each part in prediction. • The best prediction performance is achieved with an RMSE of 0.01785 and a score RUL of 0.994589. • The self-attention part plays a good capture of both the overall trend and time dependence. [ABSTRACT FROM AUTHOR]

In recent years, fuel cells have attracted more and more attention in the energy field. However, accurate evaluation of durability is one of the key aspects of large-scale commercial application of fuel cells. The data-driven method has good fitting performance for predicting remaining useful life (RUL) at complex and nonlinear systems. In this paper, Cycle Reservoir with Jump with multi-inputs and multi-outputs (MIMO-CRJ) is proposed to further improve the fitting speed and reduce the complexity of the system which improved the structure of echo state network (ESN). Different from the previous method that only uses voltage to predict the voltage, this paper considers stack current, temperature and pressure, reactant flow rate in fuel cell operation as input variables, and extracts more degradation information to predict the voltage trend. Experimental results show that MIMO-CRJ has better performance than Cycle Reservoir with Jump with single -inputs and single-output (SISO-CRJ) and ESN networks under both static and quasi-dynamic current conditions. • The multi-input Cycle Reservoir with Jump network is proposed for prognostics. • The proposed model can be used for remaining useful life on the static current and dynamic current experiments. • The MIMO-CRJ is used for short-term prediction and long-term prediction of PEMFC. [ABSTRACT FROM AUTHOR]

A robust recursive zonotopic set-membership approach for remaining useful life forecasting with application to linear parameter-varying systems is proposed in this paper. The proposed approach addresses systems with degraded components formulated as a system-level prognostics problem. Thus, the critical degraded components of the system are augmented to the states resulting a nonlinear system that is reformulated as a linear parameter-varying model. Hence, joint estimation of states and parameters is adopted in a zonotopic set-membership scheme with an optimal linear matrix inequality-based tuning and assuming unknown-but-bounded noises and uncertainties. As a result, a recursive zonotopic set-membership approach is proposed for remaining useful life forecasting based on the prediction of the failure precursors of degraded systems. Finally, this approach is tested on a DC–DC converter case study with unknown degradation behaviors, and the obtained results show the estimation and the forecasting accuracy of this methodology. • Zonotopic set-membership for prognostics applications. • Joint estimation of states and parameters for degraded systems. • Bounded degradation prediction and remaining useful life forecasting. • LMI-based optimization for robust zonotopic observer tuning for LPV systems. [ABSTRACT FROM AUTHOR]

• A prognosis framework of reaction wheel motor in space satellites is proposed. • Real motor degradation data of 3 years accelerated-life tests are utilized. • System-level prognostics is employed to account the effects of component degradation. • Failure threshold is imposed on the system according to the design requirement. The reaction wheels actuated by motors are widely used for advanced attitude control of satellites. During the satellite operation, the performance of reaction wheel motor degrades and results in unexpected failures. To guarantee the reliability and safety of satellites, it is important to predict its remaining useful life while it is in operation. To address this issue, this study presents a system-level prognostics approach for the reaction wheel motor, by regarding it as a system composed of multiple components. The approach is demonstrated by using the motor operation data obtained during the accelerated-life tests on ground for 3 years. Health degradation of each components of the motor are estimated using the adaptive extended Kalman filter. Failure threshold of the motor performance is established by the design requirement on characteristic curve. The anomaly detection and failure prediction are performed using the shifting kernel particle filter. [ABSTRACT FROM AUTHOR]

A fuel cell is a power generation device that directly converts chemical energy into electrical energy through chemical reactions; fuel cells are widely used in aerospace, electric vehicle, and small-scale stationary engine applications. The complex phenomena including mass/heat transfer, electrochemical reactions, and ion/electron conduction, can significantly affect the energy efficiency and durability of fuel cells, but are difficult to determine completely. Machine learning (ML) performs well in solving complex problems in engineering applications and scientific research. In this paper, a systematic review is conducted to explore ML methods, including traditional ML and deep learning (DL) methods, applied to fuel cells for performance evaluation (material selection, chemical reaction modeling, and polarization curves), durability prediction (state of health, fault diagnostics, and remaining useful life), and application monitoring. Then comparisons of traditional ML and DL methods are discussed, while the similarities and differences between ML and integrated physics simulations are also concluded. Eventually, the scope of ML methods applied to fuel cells is presented, and outlooks of future researches on ML applications in fuel cells are identified. • A systematical review is conducted to study on tradition ML and DL methods applied to fuel cells. • Applications of ML in performance evaluation, durability prediction and application monitoring of fuel cells are analyzed. • The comparisons of tradition ML and DL methods are investigated. • Similarities and differences between ML and that integrated physics simulation are discussed. • Scope of ML methods applied to fuel cells for local and global problems is presented. [ABSTRACT FROM AUTHOR]

The remaining useful life prediction is significant for Lithium-ion batteries to ensure safety and reliability. Due to the advantages of handling time sequence data, recurrent neural network based methods have achieved impressive performance on RUL prediction. However, most of these methods develop the RUL prediction model without considering the operating time of the battery, which is an important factor on capacity degradation. Therefore, this paper proposed a Temporal Transformer Network (TTN) for RUL of Lithium-ion batteries. The proposed method combines the self-attention mechanism of the Transformer Network with Denoising Autoencoder to implement the noise of raw data. More importantly, the proposed method designs a temporal encoding layer to introduce the operating time to the input of RUL prediction model. The performance of the proposed method is evaluated on two frequently used battery datasets. The proposed method achieves the best result compared with other frequently used RUL prediction methods. [ABSTRACT FROM AUTHOR]

The search for sustainable but environmentally friendly medium of harnessing energy for the automotive industry has led to the evolution of various energy generating and converting devices. One of such energy converting device is fuel cells. Despite the merits associated to the performance of proton exchange membrane (PEM) fuel cells, issues relating to the cost and remaining useful life prediction still persist hence impeding their further commercialization especially in the automotive industry. In spite of the progress made by the research community in developing various predictive models in order to mitigate these challenges, the accuracy of these developed models has lately become active research direction. The current study explored the accuracy of recurrent neural network, bi recurrent neural network, combined convolutional neural network and bi recurrent neural network in predicting the remaining useful life of a PEM fuel cell. The presence of the convolutional neural network was mainly to ensure pre – processing of the bi recurrent neural network for the extraction of high level features. To reduce the possibility of overfitting, a dropout approach coupled with callback technique is adopted. Validation of the model was executed based on an experimental data. The outcome of the investigation highlighted the key role of the convolutional neural network in improving the accuracy of the recurrent neural network. Comparing the root mean square error (RMSE) and mean absolute percentage error (MAPE) of the present model with other models, the developed model yielded the least values indicating a higher accuracy. For instance, the relative error showed a least value of 0.12 for the combined convolutional neural network and bi recurrent neural network compared to the long short term memory with 2.61 reported in previous studies. • Evaluation of various existing predictive models for PEM fuel cells is studied. • Convolutional neural network accuracy in predicting the rate of degradation for PEM fuel cell is captured. • The developed predictive model have higher accuracy compared to other models. • Future studies should explore further combination of various predictive models and the effect on the accuracy. [ABSTRACT FROM AUTHOR]

Effective reliability management of lithium-ion batteries (LIBs) depends on accurately assessing the State of Health (SOH). Monitoring SOH requires characterizing health indicators (HIs) derived from LIB signals. However, the extraction of HIs using signal processing techniques is susceptible to interference from time-varying operating conditions. To address this issue, the Battery Phase Space Warping (BPSW) algorithm is proposed. This algorithm aims to distinguish slow aging from fast-varying signals by reconstructing discharge voltage signals into a high-dimensional phase space (PS), enabling a qualitative analysis of LIB dynamics. Slow battery aging manifests as trajectory warping in the reconstructed PS, and the BPSW algorithm applies a local polynomial model to describe these warped trajectories and predict reference states. Furthermore, the BPSW algorithm separates slow-varying HIs from fast-varying signals by extracting the warpage for different PS states, enabling accurate tracking of the hidden aging of LIBs. The effectiveness of the BPSW algorithm is validated through degradation experiments on LIBs under constant current and square wave loads, demonstrating its superior capability to track battery aging. Additionally, the BPSW algorithm proves robust in monitoring battery degradation under variable discharge conditions, significantly improving the monitoring of LIBs. • BPSW monitors battery aging by reconstructing voltage signals into phase space. • Trajectory warping in phase space reflects changes in battery parameters. • BPSW uses local polynomial regression for reference trajectories of health. • BPSW separates fast and slow variables, minimizing interference from discharge. • BPSW aids in battery health management and predicting remaining useful life. [ABSTRACT FROM AUTHOR]

A viable way to reduce carbon emissions and achieve sustainable development goals (SDGs) is through reliable and sustainable transportation, specifically through the utilization of electric vehicles (EVs) and integrated energy storage systems (ESSs). The operation and performance efficiency of EVs are based on accurate prediction of the remaining useful life (RUL), which improves the reliability, robustness, efficiency, and longevity of different ESSs like lithium-ion battery (LIB), supercapacitor (SC), and fuel cell (FC) in EV applications. Nonetheless, the RUL prediction of the LIB, SC and FC is considered difficult due to factors associated with capacity degradation and variable environmental conditions. In recent times, enormous attention to conduct the RUL prediction has been attained by expert deep learning (DL) techniques due to their supremacy with high volumes of data, fast computational performance, and large storage capacity. However, the DL techniques implemented for RUL prediction with ESS in EV applications are inadequate. Therefore, a novel review paper is presented which comprehensively reviews DL-based RUL prediction techniques for LIB, SC, and FC, encompassing DL models, their strength, weakness, accuracy, and research gaps. The objective of the presented review work is to demonstrate the superiority of the DL techniques for RUL prediction. Firstly, the review work examines different DL methods applied to conduct the RUL prediction of the LIB, SC and FC. Furthermore, the RUL prediction results with different factors like model implementation, critical hyperparameters, and data features/samples are analyzed. The review also outlines various execution features related to experiments, data, model features, and computational performance. Additionally, the current issues with the DL techniques for RUL prediction are identified. Lastly, future trends and outlooks are listed that would pave the way toward developing an accurate RUL prediction framework for ESSs technologies in reliable EV technology. [ABSTRACT FROM AUTHOR]

In order to accurately predict the remaining useful life (RUL) of mechanical equipment under conditions of missing data, this paper proposes a novel integrated framework for data repair and RUL prediction, named DRTCR. DRTCR possesses the dual functionality of missing data repair and RUL prediction. Initially, considering the underlying trends and periodic fluctuations in the signals, polynomial and trigonometric functions are incorporated into the implicit feature matrix (IFM) generated by Gaussian random functions. This approach involves constructing a data repair deep learning framework primarily based on LSTM and CNN. This framework facilitates the repair of missing data. Furthermore, this paper introduces the integration of trend, correlation, and robustness within the LSTM units. A predictive unit named TCRcell is devised, which enhances the feature extraction capability of the prediction model. Additionally, a corrective unit within TCRcell is constructed, predominantly centered around discrete Fourier transform (DFT) and discrete wavelet transform (DWT) as frequency domain feature extraction techniques. This incorporation introduces the physical mechanism information into DRTCR and refines the predictive outcomes of TCRcell. The global frequency perspective of DFT aids in capturing the overall trend in signal variations, while DWT precisely captures localized temporal details in the time series, resulting in a dynamically corrected TCRcell prediction outcome that combines both global trends and local intricacies. Finally, experimental validation was conducted to assess the predictive performance and robustness of DRTCR. [ABSTRACT FROM AUTHOR]

In order to improve the accuracy of predicting RUL of lithium-ion batteries, a lithium-ion battery RUL prediction method based on the DBO CNN-DSformer model is proposed. Firstly, the health characteristics of the battery are extracted and the local information of health features is mined using CNN. DSformer is utilized for global information, local information, and variable correlation learning of battery aging characteristics. The DBO is used to optimize the super-parameters of the CNN-DSformer model and build the DBO CNN-DSformer model. Finally, the battery aging data set was used for verification. The results show that DBO CNN-DSformer, which sets different prediction starting points, can extract sequence information from input data and establish long-term dependencies between sequences. The maximum average MRE error in the NASA data set was 0.05, the maximum average MAE was 0.018, and the maximum average AE error was within 5. The maximum average MRE error of the CALCE data set was 0.37, the maximum average MAE was 0.014, and the maximum average AE prediction error was within 10. Compared with LSTM, RNN, and Transformer models, it was found that DBO CNN-DSformer showed high prediction accuracy and good robustness. [Display omitted] [ABSTRACT FROM AUTHOR]

The development and progress in sensor, communication, and computing technologies have led to smart environments. In such environments, data can easily be acquired not only from the monitored entities but also from the surroundings where the entity is operating. The data from the operating environment, which is not useful for learning models in isolation, is referred to as explicit context here. The performance of the predictive models can potentially improve when explicit contexts are taken into account. Typically, the data from various sensors are present in the form of time series. Recurrent Neural Networks (RNNs) are used for such data as they implicitly can deal with temporal contexts. However, while using RNNs for various smart environment applications, the available explicit contexts are often ignored or incorporated as primary (or ordinary) features into the model. Further, the conventional RNN models such as Elman RNN, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) in their current form do not provide any mechanism to integrate explicit contexts. In this paper, we propose a Context Integrated RNN (CiRNN) that enables integrating explicit contexts represented in the form of contextual features. In CiRNN, the network weights are influenced by contextual features in such a way that the primary input features receive more weight if they are relevant in a given context. To show the efficacy of CiRNN, we selected three application domains, remaining useful life estimation for predictive maintenance, traffic forecasting for intelligent transportation systems, and appliance energy usage prediction for smart homes. These applications typically capture data from various sensors and provide scope for exploiting contextual information. For all three case studies, we used public datasets. The test results show that CiRNN performs better when compared to baseline models that incorporate contextual features as primary features, as well as in comparison to state-of-the-art models. The performance is measured in terms of RMSE, MAE, and score from an asymmetric scoring function. The latter measure is specific to the task of RUL estimation. • A method to effectively Integrate explicit contexts with Recurrent Neural Networks. • The primary features relevant in a given context receive more weight. • The method is tested across three smart environment applications. • The predictive performance is significantly improved through the proposed method. [ABSTRACT FROM AUTHOR]

As a key component in the transmission system of high-speed trains, bearings need to withstand certain loads while rotating at high speeds. Once a failure occurs, it can directly affect the safety of train operations. Therefore, it is of great significance to establish a reliable remaining useful life model to ensure train operation safety. Addressing the problems of redundant features in the existing multi-feature fusion process, which affects diagnostic performance, and the spurious fluctuations in the fusion features, which cause inaccurate determination of the start fault time and lead to low prediction accuracy of remaining useful life, we propose a method based on adaptive feature fusion and the autoregressive integrated moving average model for rolling bearing prediction. Firstly, features from the time domain, frequency domain, and entropy are extracted. A feature selection mechanism with minimum redundancy is constructed to screen the optimal sensitive feature set. Secondly, based on adaptive feature fusion, the optimal sensitive feature set is dynamically fused, and the spurious fluctuations of the health index are corrected using linear regression and the 3σ principle. Next, a bottom-up time series segmentation method is employed to divide the health status of the Improved Health Indicators. Finally, a remaining useful life prediction model based on the autoregressive integrated moving average model is established. This study demonstrates that the proposed method effectively identifies features that are most sensitive to degradation trends, accurately determines the initial fault moment of bearings, and achieves effective prediction of the remaining useful life of bearings. [ABSTRACT FROM AUTHOR]

With the rapid development of lithium-ion batteries in recent years, predicting their remaining useful life based on the early stages of cycling has become increasingly important. Accurate life prediction using early cycles (e.g., first several cycles) is crucial to rational design, optimal production, efficient management, and safe usage of advanced batteries in energy storage applications such as portable electronics, electric vehicles, and smart grids. In this review, the necessity and urgency of early-stage prediction of battery life are highlighted by systematically analyzing the primary aging mechanisms of lithium-ion batteries, and the latest fast progress on early-stage prediction is then comprehensively outlined into mechanism-guided, experience-based, data-driven, and fusion-combined approaches. The key models of each method and their typical research works are profoundly analyzed, and the pros and cons of each method are then evolved with an in-depth comparison of their prediction performances. The current challenges and future perspectives of early-stage prediction are finally addressed. This review is advantageous in fully and briefly understanding the principles, methods, development, and application of early-stage prediction of battery life and is directed to expedite research on novel, accurate, efficient, and simple theories and technologies for early-stage prediction. • The gap in the absence of a review on early life prediction is bridged. • The systematic definition and review on early life prediction methods are provided. • The aging mechanisms of lithium-ion batteries are systematically compiled and summarized. • The necessity and data source of lifetime prediction using early cycles are profoundly analyzed. • The pros and cons, and predictive ability of main prediction approaches are comparatively evaluated. • The current challenges and perspectives of early-stage prediction are comprehensively discussed. [ABSTRACT FROM AUTHOR]

Maintenance strategies play a crucial role in the development and success of wind farm management, as they have a significant impact on the profitability of wind generation. The current maintenance decision-making process focuses mostly on the wind farm owner, while ignoring the benefits from the perspective of maintenance contractor. To bridge this gap, a novel contractor-oriented maintenance strategy is developed for both onshore and offshore wind farms to maximize profits for the maintenance contractor. Compared to the reported works, two main contributions are made. First, it is the first time that the maintenance strategy of a wind farm is designed from the perspective of the maintenance contractor, with emphasis on an accurate profit assessment. For this purpose, by utilizing remaining useful life (RUL) information from monitorable and non-monitorable components, a degradation-related efficiency model is constructed to quantify the loss of turbine efficiency due to degraded components and to incorporate it into the maintenance decision-making process. Secondly, various identifiable failure modes of the turbine components are investigated and simulated, as the contractor is more concerned about the maintenance costs under different failure modes. A case study using the dataset collected from real wind farms is provided to demonstrate and validate the proposed maintenance optimization method. The results indicate the proposed method can contribute to the maximum profits of the contractor while satisfying the requirements of the wind farm owner. • Designing maintenance strategies for wind farms based on the profits of maintenance contractors. • Developing degradation-related efficiency models based on RUL to quantify turbine efficiency losses. • Various identifiable failure modes of turbine components are considered in the component-level repairs. • Considering RUL information for monitorable and non-monitorable turbine components separately. • Case study using the dataset collected from real wind farms is provided. [ABSTRACT FROM AUTHOR]

Rotating machinery plays a vital role in the industrial sector, and ensuring its health status is crucial for operational efficiency and safety. The construction of accurate health indicators (HIs) have gained significant attention as it facilitates fault detection, degradation assessment, and remaining useful life prediction. This paper provides a comprehensive review of HI construction methods, focusing on their underlying principles. The main objective of HI construction is to extract degradation information from monitoring signals. Firstly, the feature-based HIs methods are reviewed, which can be regard as the fault information-related HI construction methods. Secondly, the construction steps of multi-feature fusion-based HIs and distance-based HIs are then summarized, along with their practical applications. Thirdly, the deep learning-based HI construction methods are discussed from the perspective of unsupervised HIs and supervised HIs, which offers a comprehensive understanding of how the deep learning techniques can be applied to construct HIs. Finally, the current challenges and future research opportunities in this field are highlighted. By reading this paper, engineers and researchers can gain insights into the current research ideas and directions in HI construction. It is valuable for inspiring future research endeavors and fostering advancements in this area of study. [ABSTRACT FROM AUTHOR]

Data-driven Prognostics and Health Management (PHM) requires extensive and well-annotated datasets for developing algorithms that can estimate and predict the health state of systems. However, acquiring run-to-failure data is costly, time-consuming, and often lacks comprehensive sampling of failure states, limiting the effectiveness of PHM models. This paper explores the use of Self-Supervised Learning (SSL) in PHM, addressing key limitations and proposing a novel contrastive SSL approach using a nested siamese network structure to enhance degradation feature representation. The model's performance with sparse data improves by integrating downstream task information, particularly Remaining Useful Life (RUL) prediction, into the siamese structure during SSL pre-training. This approach enforces a consistency condition that failure times for two samples from the same monitoring sequence be identical. The proposed method demonstrates superior performance on the PRONOSTIA bearing dataset, outperforming state-of-the-art methods even with sparse labeling. Furthermore, the study delves into the impact of the upstream–downstream relationship in learning processes, asserting that fine-tuning significantly enhances RUL prediction by leveraging the foundational behaviors established during pre-training. Fine-tuning refines the model's ability to capture subtle degradation patterns by building on the initial feature representations learned in pre-training, thereby improving accuracy and robustness in RUL predictions. The generalizability of the proposed strategy is confirmed through an end-to-end tool wear prediction in a real industrial environment, illustrating the applicability of the proposed method across various datasets and models, and providing effective solutions for sparse data scenarios in prognostics. [ABSTRACT FROM AUTHOR]

Remaining useful life (RUL) prediction is critical to formulating appropriate maintenance strategies for machinery health management and is playing a vital role in the field of predictive maintenance. Limited by the time-varying operational conditions, traditional RUL prediction models trained on some run-to-failure (RTF) datasets are unlikely to be generalized to a new degradation process. To enhance the generalizability, recent studies have focused on the development of deep domain adaptation methods for RUL prediction, which mainly align the global temporal features across the source and target domains, resulting in imprecise predictions under time-varying operational conditions. In addition, existing RUL prediction methods are lacking in clear physical significance and interpretability. To address the above-mentioned issues, an operational condition attention (OCA) subnetwork is constructed to eliminate the entanglement between the time-varying operational conditions and monitoring data. Adversarial-based domain adaptation (ABDA) and distance-based domain adaptation (DBDA) methods were utilized respectively to reduce the distribution discrepancy of the temporal features. In this way, two novel domain adaption methods were proposed for RUL prediction with time-varying operational conditions. The comprehensive experiments were conducted on aero-engines to validate the proposed methods. Owing to the explicit modeling of the influence mechanism between the operational conditions and monitoring data, the proposed methods exhibit improved performance as well as higher prediction accuracy than traditional deep domain adaption methods while being certainly interpretable. • Interpretable OCA-informed domain adaptation network is proposed for RUL prediction. • The OCA subnetwork is designed for decoupling monitoring data from time-varying OCs. • The new physical-informed neural network helps elucidate the decision-making process. [ABSTRACT FROM AUTHOR]

The oxygen byproduct generated from water electrolysis can be utilized on-site in residential combined heat and power (CHP) systems to enhance economic efficiency. This study explores solutions for optimizing the use of enriched oxygen gas (EOG) and determining the ideal oxygen fraction to maximize economic viability. A novel EOG supply system is designed to regulate oxygen fraction, oxygen excess ratio (OER), and humidity. A dynamic model of a PEMFC integrated with the EOG supply system is developed using MATLAB/Simulink® software and partially validated. The dual effects of EOG on system efficiency and remaining useful lifetime (RUL) are analyzed using a comprehensive economic optimization algorithm, leading to the identification of optimal oxygen fractions for various hydrogen production modes. Additionally, a feedforward static decoupling controller is designed to independently control the OER and oxygen fraction. The results indicate that, under rated conditions (10 kW), system power increased by 15 %, along with a 10 % improvement in system efficiency, from 48 % to 58 %. In conclusion, the efficient use of enriched oxygen gas demonstrates significant economic feasibility and practical viability, making it highly suitable for distributed power generation applications. • On−site utilization of oxygen byproduct for the residential PEMFC-CHP system is presented in this paper. • The novel oxygen enrichment gas supply system model is designed in this paper. • System efficiency and remaining useful lifetime are investigated using the proposed economic optimization algorithm to obtain optimal oxygen fractions. • A decoupling controller is designed to achieve relative independent control. [ABSTRACT FROM AUTHOR]

In recent years, prognostics gained attention in various industries by optimizing maintenance, boosting operational efficiency, and preventing costly downtime. Central to prognostics is the Remaining Useful Life (RUL), representing the critical time before system failure. Deep learning advancements facilitate RUL forecasting by extracting features from diverse data formats such as time series, images, or sequences thereof, in one, two, or three dimensions, respectively. Yet, predicting RUL from image sequences often relies heavily on resource-intensive techniques like digital image correlation, complicating data acquisition. To address challenges with high-dimensional data and unreliable models, this study introduces ISTRUST, an innovative Transformer-based architecture. ISTRUST (Interpretable Spatiotemporal TRansformer for Understanding STructures) tackles the dual challenges posed by high-dimensional data and the black-box nature of existing models. Leveraging Transformers' attention mechanism, ISTRUST breaks down the spatiotemporal domain, effectively realizing interpretable RUL predictions under uncertainty using only sparse raw image sequences as input. Evaluated on fatigue-loaded composite samples showcasing crack propagation, ISTRUST interprets the relation between cracks and RUL via the attention mechanism. The results substantiate its capacity to interpret and clarify instances in which predictions may exhibit variability in accuracy. Through the attention mechanism, a strong correlation between the model's spatiotemporal focus and the RUL predictions is established, making it, to the best of our knowledge, the first model to provide interpretable stochastic RUL predictions directly from sequential images of this nature. • Predicting remaining useful life from raw image sequences in composite structures. • Interpretable AI via the attention mechanism of transformers. • Spatiotemporal decomposition for better accuracy and model interpretation. • Self-attention correlates crack propagation with remaining useful life. • Monte Carlo dropout for uncertainty estimation. [ABSTRACT FROM AUTHOR]

Current approaches for monitoring machine health (SOH) and effective prognostics depend on the extensive use of complete degradation data trajectories, implying the reliance on data generation techniques that involve functional degradation of the real system until the failure state is reached. These commonly adopted approaches that depend on labeled target data remain operationally and economically nonviable for most industries and safety critical systems. This paper presents novel approaches that alleviate the existing dependence of most prognostics procedures on Remaining Useful Life (RUL) labeled data for training. To this end, firstly, a hybrid data augmentation procedure is proposed that enables the integration of system knowledge available a priori as well as physics of failure, within the training data. Secondly, an unsupervised Health Index (HI) extraction approach is developed, followed by a long-term prediction of this same HI, that leads to an efficient prediction of RUL without labeled data. Finally, a reliability-based assessment is performed to validate the proposed approach. This comprehensive approach (i.e. integrating all the various stages involved in achieving a RUL prediction based on unlabeled data) is tested on a real industrial aircraft system demonstrating the effectiveness of the proposed approach in real industrial context. [Display omitted] • Comprehensive, unsupervised approach adapted to real-life industrial cases (i.e. where no degradation data are available). • Use of physics-based data augmentation to overcome the lack of real degradation data. • Use of the reconstruction error of an autoencoder to extract the Health Index. • Validation of AI-based predictions using established laws of reliability. [ABSTRACT FROM AUTHOR]

In recent years, 3D point clouds (PCs) have gained significant attention due to their diverse applications across various fields, such as computer vision (CV), condition monitoring (CM), virtual reality, robotics, autonomous driving, etc. Deep learning (DL) has proven effective in leveraging 3D PCs to address various challenges encountered in 2D vision. However, applying deep neural networks (DNNs) to process 3D PCs presents unique challenges. This paper provides an in-depth review of recent advancements in DL-based industrial CM using 3D PCs, with a specific focus on defect shape classification and segmentation within industrial applications. Recognizing the crucial role of these aspects in industrial maintenance, the paper offers insightful observations on the strengths and limitations of the reviewed DL-based PC processing methods. This knowledge synthesis aims to contribute to understanding and enhancing CM processes, particularly within the framework of remaining useful life (RUL), in industrial systems. • Survey 3D PC defect classification and segmentation in industrial systems. • Discusses taxonomies categorizing existing models based on data handling designs. • Comparison analysis of the models for industrial damage identification. • Discusses strengths and limitations of DL-based processing methods for 3D PC. • Identifies research gaps and suggests potential avenues for future advancements. [ABSTRACT FROM AUTHOR]

Accurate remaining useful life (RUL) prediction holds significant importance for health management of aero-engines, ensuring safety and reducing the maintenance cost. The coupling between variable operating conditions and diverse sensor signals makes it hard to construct an accurate and stable model, which predicts precise results and simultaneously obtains condition-independent features reflecting the degradation trend. To address the problem, this paper proposes a novel approach named residual-based adversarial feature decoupling (RAFD) for RUL prediction of aero-engine which achieves decoupling both explicitly and implicitly. The paper formally defines the coupling relationships within the signals, consisting of information from normal pattern, degradation pattern, and operating conditions which are decoupled explicitly and implicitly. An operation-condition-mapping model (OCMM) is developed to explicitly decouple normal pattern, establishing the mapping between operation conditions and sensor signals in the healthy stage. Residuals serve as inputs of the prediction model, a continuous adversarial neural network specifically designed for implicit decoupling which extracts condition-independent features and predicts RUL accurately. The effectiveness of our method is validated on NASA's N-CMAPSS dataset, resulting in superior prediction performance and feature visualization when compared with other existing methods. • Signal coupling is defined and decoupled by RAFD explicitly and implicitly. • OCMM is proposed to extract residuals for explicit degradation indication. • CANN is proposed to capture implicit decoupling of operation conditions. • The superiority of the method is validated on N-CMAPSS dataset. [ABSTRACT FROM AUTHOR]

Lithium-ion batteries (LIBs) degradation can be classified into capacity reduction and increase of the internal resistance. Accurate estimation of degradation is important to determine the economic value and the remaining useful life of the battery. This paper adopts dynamic electrochemical impedance spectroscopy (EIS) to evaluate degradation while charging the battery, without requiring additional time for the analysis. A new method is proposed to remove the artifacts arising from the time-variance of the system during the measurement. The validity of the method is confirmed by comparing it with the results obtained by static EIS measurements at equilibrium. The spectra are fitted to an equivalent circuit containing three resistances (R 0 , R 1 , R 2). The method is applied to data obtained before and after aging cycles performed at 1.4C charge and 2.0C discharge rates, at 25 °C, 10 °C and −5 °C. The resistance of the films (R 1), which is obtained from the semicircle at high frequencies in the EIS spectrum, has proved to be the key indicator of aging, and linearly depends on the SoC. The increase of R 1 dominates among all the resistances, particularly at higher temperatures. Dynamic EIS is considered a powerful tool to diagnose LIBs in practical applications. [Display omitted] • Analysis of degradation based on dynamic electrochemical impedance spectroscopy. • The proposed model corrects the artifact caused by the variation of SoC while measuring. • The method is applied on cells aged with fast charging at 25 °C, 10 °C and −5 °C. • Films and charge transfer resistances dominate the battery aging at higher temperatures. • The resistance associated with the films, R 1 , is a key indicator of degradation and SoC. [ABSTRACT FROM AUTHOR]