Search

The remaining useful life (RUL) prediction based on deep learning obtains certain prediction results by training model with deterministic parameters. However, this method ignores the error range of parameter estimation, resulting in the inability to fully characterise the uncertainty. Therefore, a RUL prediction method based on anomaly correction is proposed. Firstly, to address the problem of RMS deviating from the degradation trajectory caused by external random factors, this paper proposes an adaptive time window anomaly correction strategy with sliding windows are used to construct degenerate slope. Secondly, to solve the problem of prediction result errors caused by changing random factors, this paper considers heteroscedasticity in prognosis and constructs a probability density function library to dynamically match the probability distribution. Then, aiming at the mean value as the prediction result may lead to overconfidence, a probability-based weighted enhanced method is proposed to improve the prediction accuracy. Finally, two datasets are used to verify the effectiveness and superiority of the proposed method. [ABSTRACT FROM AUTHOR]

Advanced reactor designers are looking to maximize the system capacity factor to make advanced reactors more economically competitive and meet the projected energy demand. To achieve this goal, we propose a Dynamic Operation and Maintenance Optimization (DyOMO) framework to perform system-level predictive maintenance (PdM) using a dynamic Bayesian network and component-specific PdM using deep neural networks. At the system level, DyOMO detects the presence of anomalous phenomena, determines the most influential degradation mode, and estimates the remaining useful life (RUL) distribution for the system. At the component level, DyOMO summarizes the health state of key system components, determines the presence of an anomaly using a feedforward neural network, and predicts component RUL using a Bayesian neural network. To evaluate the overall performance of DyOMO, normal operations of a Pebble-Bed High-Temperature Gas-cooled Reactor (PB-HTGR) were simulated with realistic component degradation for the steam turbine and steam generator. Across the 20 independent reactor life simulations, it was found that maintenance was always performed before any safety limits were violated and before a component failed. Specifically, the system-level PdM suggested maintenance on the steam generator once the steam pressure approached its safety limit, and the component-specific PdM suggested maintenance on the steam turbine once the turbine blade hardness degraded. The results indicate that through the continuous monitoring of the system and individual components, the DyOMO framework improves safety and increases the availability of the reactor when compared to traditional maintenance philosophies. [ABSTRACT FROM AUTHOR]

The remaining useful life prediction of rolling bearing is vital in safety and reliability guarantee. In engineering scenarios, only a small amount of bearing performance degradation data can be obtained through accelerated life testing. In the absence of lifetime data, the hidden long-term correlation between performance degradation data is challenging to mine effectively, which is the main factor that restricts the prediction precision and engineering application of the residual life prediction method. To address this problem, a novel method based on the multi-layer perception neural network and bidirectional long short-term memory network is proposed. Firstly, a nonlinear health indicator (HI) calculation method based on kernel principal component analysis (KPCA) and exponential weighted moving average (EWMA) is designed. Then, using the raw vibration data and HI, a multi-layer perceptron (MLP) neural network is trained to further calculate the HI of the online bearing in real time. Furthermore, The bidirectional long short-term memory model (BiLSTM) optimized by particle swarm optimization (PSO) is used to mine the time series features of HI and predict the remaining service life. Performance verification experiments and comparative experiments are carried out on the XJTU-SY bearing open dataset. The research results indicate that this method has an excellent ability to predict future HI and remaining life. [ABSTRACT FROM AUTHOR]

In the field of oil and gas equipment management, frequent maintenance is conducted, resulting in unnecessary costs. Relying solely on a single artificial intelligence model often leads to low predictive accuracy and inadequate robustness because of the poor data quality. Therefore, a method based on ensemble learning modeling is proposed to accurately assess the health status of industrial equipment and predict its remaining useful life. By conducting big data analysis on the operational history data of all oil and gas equipment, various fault instances are extracted and grouped accordingly. The integration of meta-learning convolutional shrinkage neural networks (ML-CSNN), domain expert rules, and support vector machine (SVM) models forms a hybrid model aimed at constructing a robust classification model. The effectiveness of the proposed method is validated using operational data from shengli oilfield production wells in China. The proposed method achieves 0.98 in accuracy, 0.93 in precision, 0.94 in recall, and 0.93 in F1 score, which is an improvement of 9–25% compared to the popular integrated learning models, such as GBDT and XGBoost. By designing the ablation study, results demonstrate the method's ability to accurately predict potential faults of oil and gas equipment, thereby facilitating the enhancement of predictive maintenance strategies. [ABSTRACT FROM AUTHOR]

With the development of deep learning, the potential for its use in remaining useful life (RUL) has substantially increased in recent years due to the powerful data processing capabilities. However, the relationships and interdependencies of operation parameters in non-Euclidean space are ignored utilizing the current deep learning-based methods during the degradation process for engine. To address this challenge, an improved sand cat swarm optimization-assisted Graph SAmple and aggregate and gate recurrent unit (ISCSO-GraphSage-GRU) is proposed to achieve RUL prediction in this work. Firstly, the maximum information coefficient (MIC) is utilized for describing the interdependent relations of measured parameters. Building on this foundation, the constructed graph data is used as input to GraphSage-GRU so as to overcoming the shortcomings of existing deep learning methods. Additionally, this work proposed an improved sand cat swarm optimization (ISCSO) to improve the predicted performance of GraphSage-GRU, including tent mapping in population initialization and a novel adaptive approach enhance the exploration and exploitation of sand cat swarm optimization. The CMAPSS dataset is used to validate the effectiveness and advancedness of ISCSO-GraphSage-GRU, and the experimental results show that the R2 of the ISCSO-GraphSage-GRU is greater than 0.99, RMSE is less than 6. [ABSTRACT FROM AUTHOR]

The development of a smart predictive monitoring system for modeling tool wear during the milling process has gained attention in the field of non-destructive testing for different material structures. The use of the Acoustic-Emission (AE) technique has been beneficial in inspecting damages in material structures. This study focuses on inspecting tool materials during the milling process and predicting the remaining useful life (RUL) of this tool under a high-speed milling machine cutter. The dataset used is available online and includes six series of cutting processes, with seven variables data signals present in each series (three for forces, three for vibrations, and one AE). The main aim of this study is to solely use the AE signals from the dataset to investigate the RUL of tool materials. The steps in processing and pre-processing the dataset involve detecting AE signals from the six cutting series, normalizing, and feature extraction. Relevant features are selected using the technique of Infinite Feature Selection (InfFS), followed by modeling of tool wear using the selected features through the online support vector regression (SVR) method, trained with three series (C1, C4, C6), and prediction of RUL based on the obtained models. The predicted models for the three tested series (C2, C3, C5) at each level of the cutting process have shown good performances, directly influencing the RUL of the tool. This developed processing approach enables the prediction of models for tool wear using only AE signals, marking a notable advancement in the field of predictive monitoring systems of materials under milling processes. [ABSTRACT FROM AUTHOR]

Predicting the remaining useful life of equipment can help organizations improve efficiency, reduce costs and enhance safety by enabling them to plan maintenance and repairs, and optimize their use of equipment. A major challenge in this field is the development of accurate deep learning models, particularly when data is limited to a few run-to-failure trajectories. Traditional methods, such as pre-training on a larger dataset followed by fine-tuning on the main data, often fall short under these constraints. To address this challenge, we propose an auxiliary training approach that integrates auxiliary objectives from related but distinct datasets. This approach enriches the learning process, utilizing knowledge from a broader data range and acting as a regularization mechanism to improve generalization from limited data. The effectiveness of the proposed method is demonstrated by experiments on two well-known public datasets, CMAPSS and N-CMAPSS, across eight distinct settings, and is shown to outperform state-of-the-art approaches such as single-task learning and pre-training followed by fine-tuning. [ABSTRACT FROM AUTHOR]

Predictive maintenance, recognized as an effective health management strategy for extending the lifetime of devices, has emerged as a hot research topic in recent years. A general method is to execute two separate steps: data-driven remaining useful life (RUL) prediction and a maintenance strategy. However, among the numerous studies that conducted maintenance and replacement activities based on the results of RUL prediction, little attention has been paid to the impact of preventive maintenance on sensor-based monitoring data, which further affects the RUL for repairable degrading devices. In this paper, an adaptive RUL prediction method is proposed for repairable degrading devices in order to improve the accuracy of prediction results and achieve adaptability to future degradation processes. Firstly, a phased degradation model based on an adaptive Wiener process is established, taking into account the impact of imperfect maintenance. Meanwhile, integrating the impact of maintenance activities on the degradation rate and state, the probability distribution of RUL can be derived based on the concept of first hitting time (FHT). Secondly, a method is proposed for model parameter identification and updating that incorporates the individual variation among devices, integrating maximum likelihood estimation and Bayesian inference. Finally, the effectiveness of the RUL prediction method is ultimately validated through numerical simulation and its application to repairable gyroscope degradation data. [ABSTRACT FROM AUTHOR]

Remaining useful life (RUL) prediction techniques based on degradation modeling to ensure the safety and reliability of stochastic degraded devices are of critical value. However, although the performance degradation of various devices in engineering practice is affected by multi-timescales, there are few RUL prediction methods for stochastic degraded devices with dual time scales in the existing literature. The research of prediction methods considering continuous and discrete time scales is blank. Toward this end, this paper proposes a novel prognostic method for nonlinear degradation devices based on the continuous-discrete dual time scales. First, the performance degradation with a discrete timescale is determined with a compound Poisson process and integrated into the diffusion process model to establish a nonlinear dual-timescale degradation model. Subsequently, the joint expression of the dual-timescale RUL distribution of the device in the first hitting time sense solves with threshold transformation. Finally, the effectiveness and superiority of the proposed method are verified through a numerical simulation and a gyroscope example. The experimental results show that the proposed method can provide more comprehensive information about the life distribution and effectively improve the accuracy of the RUL prediction. [ABSTRACT FROM AUTHOR]