Remaining Useful Life Prediction Method Research Articles

A multi-constrained domain adaptation network for remaining useful life prediction of bearings

The prediction of the remaining useful life (RUL) of bearings could reduce maintenance costs and prevent serious consequences. Due to the variable working conditions of bearings, domain adaptation methods for cross-domain RUL prediction have received extensive attention. However, existing research often focuses solely on aligning marginal distributions while overlooking the alignment of conditional distributions and fails to retain domain-specific information during the alignment process. To address these issues, we propose a multi-constrained domain adaptation (MCDA) method for cross-domain RUL prediction. MCDA method is based on an adversarial domain adaptation architecture, where the improved domain discriminator aligns the data features of different degradation stages separately to achieve the alignment of marginal distribution and simultaneously aligns the conditional distribution. The state discrimination module is introduced to align the marginal and conditional distributions while retaining the target domain-specific information. To confirm the effectiveness of the proposed method, we conducted RUL prediction tasks for bearings in various cross-domain scenarios using both publicly and self-built datasets. The experimental results showed that the MCDA method could better handle cross-domain RUL prediction tasks compared to existing methods.

Remaining useful life prediction for lithium-ion batteries based on sliding window technique and Box-Cox transformation

Accurate remaining useful life (RUL) prediction technique matters in lithium-ion battery use, optimization, and replacement. This article presents an RUL prediction method combining the sliding window (SW) technique and Box-Cox transformation (BCT). This method achieves online RUL prediction with acceptable accuracy, is independent of offline training data, and only brings a low computational burden. The SW technique is employed for gathering a certain amount of capacity data, which is subsequently transformed using BCT-related techniques to construct a capacity degradation model. The identified model is then extrapolated to predict battery RUL, and the prediction uncertainties are calculated through Monte Carlo (MC) simulation. A simple implementation shows that this hybrid method outperforms the history-based polynomial and BCT methods in battery RUL prediction. Given the segmented capacity degradation trend, a constraint can be imposed on the model parameter for optimization purposes. Experimental results demonstrate that the optimized method obtains lower root-mean-square errors (RMSEs) of RUL predictions during the last 20 % of the battery lifetime than the original one, and the precise RULs are predicted with standard deviations mainly within [1, 10] cycles.

Remaining Useful Life Estimation of Turbofan Engines with Deep Learning Using Change-Point Detection Based Labeling and Feature Engineering

Accurate remaining useful life (RUL) prediction is one of the most challenging problems in the prognostics of turbofan engines. Recently, RUL prediction methods for turbofan engines mainly involve data-driven models. Preprocessing the sensor data is essential for the performance of the prognostic models. Most studies on turbofan engines use piecewise linear (PwL) labeling, which starts with a constant initial RUL value in normal/healthy operating time. In this study, we designed a prognostic procedure that includes difference-based feature construction, change-point-detection-based PwL labeling, and a 1D-CNN-LSTM (one-dimensional–convolutional neural network–long short-term memory) hybrid neural network model for RUL prediction. The procedure was evaluated on the subset FD001 of the C-MAPSS dataset. The proposed procedure was compared with machine learning and deep learning models with and without the new difference feature. Also, the results were compared with the studies that used similar labeling approaches. Our analysis of the numerical results underscores the clear superiority of the proposed 1D-CNN-LSTM model with the difference feature in RUL prediction, with a score of 437.2 and an RMSE value of 16.1. This result illustrates the superior predictive capability of the 1D-CNN-LSTM model, which outperformed traditional machine learning methods and one of the earliest deep learning methods. These findings emphasize the superior predictive capability of the 1D-CNN-LSTM model and underline the potential of the feature engineering process for more accurate and robust RUL prediction in the context of turbofan engine prognostics.

Knowledge enhanced ensemble method for remaining useful life prediction under variable working conditions

Remaining useful life (RUL) prediction is essential in enhancing the safety and reliability of rotating machinery. Deep learning techniques have been extensively researched and demonstrated promising results in RUL prediction tasks. But most existing models are designed for machinery equipment in a specific condition. In this case, a novel prediction method, knowledge-enhanced convolutional Transformer ensemble model (KE-CTEM), is proposed in this study. First, a feature extraction neural network (FENN) is introduced to extract features and transfer the working conditions information of existing datasets as knowledge to downstream RUL prediction tasks. Then, a convolutional Transformer model is leveraged to capture the input data degradation patterns and predict RUL values. Finally, knowledge-enhanced strategy and ensemble strategy are proposed to enhance the robustness of the model and improve the prediction accuracy.To verify the practicality and effectiveness of the proposed method, run-to-failure data of bearings from PRONOSTIA platform are utilized for RUL prognostics. Compared with several representative and state-of-the-art methods, the experimental results demonstrate the superiority and feasibility of the proposed method. And ablation study indicates the high efficiency and robustness of each module within the proposed model. Compared with representative RUL prediction methods, the proposed KE-CTEM demonstrates superior performance in terms of RMSE and MAPE with a reduction of 32.0% and 16.2%, respectively.

An attention-based temporal convolutional network method for predicting remaining useful life of aero-engine

Researches on Remaining Useful Life (RUL) prediction of aero-engine could help to make maintenance plans, improve operation reliabilities and reduce maintenance costs. While deep learning methods have been widely used in RUL prediction research, most deep learning-based RUL prediction methods tend to treat input features as equally important. Contributions of different channels and time steps from input features are not considered simultaneously, which will inevitably affect efficiencies and accuracies of RUL prediction. Therefore, a novel deep learning-based RUL prediction method named attention-based temporal convolutional network (ATCN) is proposed in this article. First, an improved self-attention mechanism is used to weight contributions of different time steps from input features. Input features of time steps closely related to RUL are enhanced by the improved self-attention mechanism, which could improve efficiencies of feature extraction in a network. Then, a temporal convolutional network is constructed to capture long-term dependent information and extract feature representations from weighted features of the improved self-attention mechanism. Next, a squeeze-and-excitation mechanism is adopted to weight contributions of different channels from feature representations, which could help to improve prediction accuracies of the network. Finally, a fully connected layer is constructed to fuse weighted features to output RUL values. A commercial modular aero-propulsion system simulation (C-MAPSS) dataset from NASA is applied to verify effects of the proposed method. Performances of the proposed method are compared with those based on different neural network architectures, such as CNN, RNN, LSTM, DCNN, TCN, BiGRU-TSAM, AGCNN and channel attention plus Transformer. Results show that the proposed method could yield results with higher accuracy for RUL prediction of aero-engine than other methods.

A sliding sequence importance resample filtering method for rolling bearings remaining useful life prediction based on two Wiener-process models

The remaining useful life (RUL) prediction of rolling bearings is an important part of prognostic and health management of mechanical systems. The model based on Wiener process can describe the time variability in the degradation process of bearings. However, in practical engineering, the degradation trends of bearings are often inconsistent, and it is difficult to fit the actual degradation trends of bearings with a single Wiener process model-based filtering method. Therefore, to improve the generalization ability, this paper uses linear model and exponential model based on Wiener process to predict bearing RUL. A sliding sequence importance resample filtering algorithm is proposed to track the degradation state of the bearing and reduce the prediction error by combining the two degradation models. Last, the superiority and effectiveness of the proposed method are illustrated by comparing with other commonly used RUL prediction methods on the basis of PRONOSTIA dataset.

Physics-informed multi-state temporal frequency network for RUL prediction of rolling bearings

Accurate prediction of remaining useful life (RUL) has been a key issue in the field of Prognostic and Health Management (PHM), which aims at predictive maintenance to improve equipment reliability and safety. Aiming at the problem that the existing RUL prediction methods are weak in perceiving long-term features and poor in capturing periodic dependence, which leads to inaccurate and seriously lagging RUL prediction results of rolling bearings, this article proposes a physics-informed bearing RUL prediction approach, namely multi-state temporal frequency network (MSTFN). Firstly, a physics-informed dynamic adaptive inverse discrete Fourier transform (IDFT) frequency domain block is constructed, which maps the spectral information from the known domain to the spectral interval of the unknown domain for extracting the periodic features of the sequence. Secondly, a residual self-attention multi-state gated control unit (RSA-MSGCU) is proposed, which incorporates a novel multi-state hierarchical division mechanism in memory cells to enhance the medium- and long-term feature perception capability. Based on RSA-MSGCU, a trend prediction time domain block is built to extract the trend features of the sequence. Finally, the periodic and trend features are fused to achieve compatibility between the IDFT physical model and the data-driven model and then make the final RUL prediction.

Remaining useful life prediction of lithium-ion batteries using EM-PF-SSA-SVR with gamma stochastic process

Due to the complex changes in physicochemical properties of lithium-ion batteries during the process from degradation to failure, it is difficult for methods based on physical or data-driven models to fully characterize this nonlinear process, and existing methods that hybridize physical and data-driven models suffer from ambiguous hybridization, which results in the vast majority of existing methods for predicting the remaining useful life (RUL) of lithium-ion batteries suffering from a lack of accuracy and robustness. In this study, a novel hybrid approach based on empirical modeling and data-driven techniques is proposed for predicting the RUL of lithium-ion batteries. To better capture its complexity, stochasticity, and state transition, and improve the modeling accuracy and RUL prediction precision, Gamma stochasticity and state-space modeling are used to empirically model the complex Li-ion battery degradation process. Moreover, the expectation maximization (EM) method of particle filtering (PF) was used to estimate the hidden parameters of the empirical model, and the estimated parameters were corrected using an optimized support vector regression (SVR) method to enhance the generalization performance and robustness of the data-driven model. The results show that the gamma state-space model is effective in capturing the inherent stochastic properties of the battery degradation and the proposed hybrid method outperforms the existing prediction methods in RUL prediction. The experiments show that the sparrow search algorithm (SSA) optimized SVR is considered to be the most effective correction method for the estimated parameters, while the new EM-PF-SSA-SVR hybrid method provides better performance for state assessment and RUL prediction of lithium-ion batteries. It is indicated that the proposed EM-PF-SSA-SVR method with Gamma stochastic process has hybrid validity and superior performance with equal performance and less parameter computation relative to the existing state-of-the-art deep learning RUL prediction methods.

Intelligent prediction of rolling bearing remaining useful life based on probabilistic DeepAR-Transformer model

Remaining useful life (RUL) prediction for rolling bearings requires highly accurate and stable long-term prediction capabilities in equipment health management, which demands that the prediction model has strong data reasoning and regression performance. However, it is difficult to accurately capture long-term dependencies via traditional convolutional neural network because the information loss and insufficient analysis are unavoidable during the feature extraction process. An end-to-end time series forecasting method called D-former for RUL prediction of rolling bearings is proposed in this paper. The method mainly consists of DeepAR and a multi-layer encoder, so it is able to extract degradation features directly from the original signal. This method has the following salient features: (1) the designed multi-head attention mechanism can highlight important feature information and realize parallel computing, so the method is extremely suitable for processing long-term time series; (2) the important time feature information is rearranged through DeepAR, so the method has the excellent domain adaptability, and it can achieve accurate prediction of rolling bearing RUL under different working conditions. The verification experiment was implemented on the IEEE PHM 2012 dataset and the XJTU-SY bearing dataset. The experimental results show that the proposed D-former method is actually superior to the existing mainstream RUL prediction methods.

Multiscale global and local self-attention-based network for remaining useful life prediction

Remaining useful life (RUL) prediction plays an important role in prognostics and health management (PHM) and can significantly enhance equipment reliability and safety in various engineering applications. Accurate RUL prediction enables proactive maintenance planning, helping prevent potential hazards and economic losses caused by equipment failures. Recently, while deep learning-based methods have swept the RUL prediction field, most existing methods still have difficulties in simultaneously extracting multiscale global and local degradation feature information from raw multi-sensor monitoring data. To address these issues, we propose a novel multiscale global and local self-attention-based network (MGLSN) for RUL prediction. MGLSN consists of global and local feature extraction subnetworks based on self-attention, which work in parallel to simultaneously extract the global and local degradation features of equipment and can adaptively focus on more important parts. While the global network captures long-term dependencies between time steps, the local network focuses on modeling local temporal dynamics. The design of parallel feature extraction can avoid the mutual influence of information from global and local aspects. Moreover, MGLSN adopts a multiscale feature extraction design (multiscale self-attention and convolution) to capture the global and local degradation patterns at different scales, which can be combined to better reflect the degradation trend. Experiments on the widely used Commercial Modular Aero-Propulsion System Simulation (CMAPSS), New CMAPSS (N-CMAPSS), and International Conference on Prognostics and Health Management 2008 challenge datasets provided by NASA show that MGLSN significantly outperforms state-of-the-art RUL prediction methods and has great application prospects in the field of PHM.

Remaining useful life prediction for space bearing with cage friction fault based on data pre-screening and gated recurrent unit

The remaining useful life (RUL) of bearings in space inertia actuators is crucial for performance maintenance requirements. But it is quite difficult to accurately predict the RUL of space bearings due to the significant intermittency and nonstationary properties caused by cage friction faults commonly occurring during the operation of the actuator. This paper proposes a data-driven method for RUL prediction of space bearings by incorporating the gated recurrent unit network with a novel data pre-screening approach. In the proposed method, a stacked autoencoder and clustering approach are introduced into the data pre-processing method, and a health index called Overrun-Distance is constructed for lifetime assessment. To verify the proposed method, a series of vibration tests on flywheels equipped with space bearings are conducted and used for RUL evaluation. The results show that the proposed RUL prediction method is applicable to space bearings for RUL prediction with high accuracy and effectiveness.

Remaining useful life prediction for rolling bearings based on RVM-Hausdorff distance

To address the shortcomings of existing bearing remaining useful life (RUL) prediction process such as low accuracy and reliance on expert experience for parameter estimation, this paper proposes a bearing RUL prediction method combining relevance vector (RV) machine (RVM) and hybrid degradation model. The bearing degradation characteristics are extracted from the acquired vibration acceleration signals, the time-varying 3σ criterion is then used to determine the bearing first predicting time, and the sequence from initial failure time point to the inspection time is regressed by differential kernel parameter RVM to obtain the different sparse RVs. A mixed degenerate model combined single exponential, weighted double exponential, and polynomial is used to fit the sparse RVs to obtain the fitted curve clusters. The similarity based on bidirectional Hausdorff distance is used to select the best degradation curve, and to extrapolate the best degradation curve to the failure threshold. The experimental results indicate that the proposed method overcomes the widespread drawbacks of monotonicity and trend bias in model-based methods, and has better prediction efficiency than the conventional exponential models.

A novel transfer gear life prediction method by the cross-condition health indicator and nested hierarchical binary-valued network

As one of the crucial techniques in prognostics and health management, the long-term remaining useful life (RUL) prediction of gears is essential for maintenance decisions in engineering applications. However, most previous RUL prediction approaches may have unacceptable prediction performance due to insufficient prior life-cycle data and variable operating conditions. In this paper, a novel transfer life prediction method is proposed for gear RUL estimation under different working conditions. First, a health indicator (HI) transfer construction framework, QFMDCAET, is developed for generating cross-condition HIs under various working conditions based on a quadratic function-based multi-scale deep convolutional auto-encoder network and multi-kernel maximum mean discrepancy. With the acquired gear HIs, the nested hierarchical binary-valued network (NHBN) is designed to estimate the gear RUL. In NHBN, a Gumbel function is applied to improve the memory ability of NHBN. Moreover, a nested hierarchical mechanism is also used to fully utilize the sequence information of the cell state in NHBN. The experimental results illustrate the effectiveness of the proposed NHBN in gear RUL prediction, especially for long-term RUL prediction. Lastly, the comparison results indicate that the proposed NHBN-based RUL prediction method is superior to other conventional and advanced RUL prediction methods.

Adaptive staged RUL prediction of rolling bearing

Remaining useful life (RUL) prediction is an important prognostic task in mechanical maintenance. However, a single health indicator cannot provide complete degradation information and clearly divide the degradation process, which reduces the accuracy of predicting the RUL of key components. Therefore, an adaptive staged RUL prediction method based on feature fusion is proposed. Firstly, a proportion-weighted method is proposed, which integrates various degradation information of rolling bearings to construct an evaluative health indicator (EHI). Secondly, aiming at the problem of clustering boundary ambiguity in different degradation stages, a two-step outlier detection strategy is proposed to solve the boundary point state interweaving problem between stages. Then, a staged RUL prediction model is proposed to match the model of the corresponding stage adaptively and dynamically to accurately predict RUL. Finally, two different sets of experiments are carried out to verify the accuracy and superiority of the proposed method in the RUL prediction.

Remaining Useful Life Prediction for Lithium-Ion Batteries Based on Iterative Transfer Learning and Mogrifier LSTM

Lithium-ion battery health and remaining useful life (RUL) are essential indicators for reliable operation. Currently, most of the RUL prediction methods proposed for lithium-ion batteries use data-driven methods, but the length of training data limits data-driven strategies. To solve this problem and improve the safety and reliability of lithium-ion batteries, a Li-ion battery RUL prediction method based on iterative transfer learning (ITL) and Mogrifier long and short-term memory network (Mogrifier LSTM) is proposed. Firstly, the capacity degradation data in the source and target domain lithium battery historical lifetime experimental data are extracted, the sparrow search algorithm (SSA) optimizes the variational modal decomposition (VMD) parameters, and several intrinsic mode function (IMF) components are obtained by decomposing the historical capacity degradation data using the optimization-seeking parameters. The highly correlated IMF components are selected using the maximum information factor. Capacity sequence reconstruction is performed as the capacity degradation information of the characterized lithium battery, and the reconstructed capacity degradation information of the source domain battery is iteratively input into the Mogrifier LSTM to obtain the pre-training model; finally, the pre-training model is transferred to the target domain to construct the lithium battery RUL prediction model. The method’s effectiveness is verified using CALCE and NASA Li-ion battery datasets, and the results show that the ITL-Mogrifier LSTM model has higher accuracy and better robustness and stability than other prediction methods.

Federated learning with network pruning and rebirth for remaining useful life prediction of engineering systems

Remaining useful life (RUL) prediction has achieved considerable success through centralized learning methods. However, traditional data aggregation may cause privacy disclosure, and existing prediction models are often too large to be trained efficiently. This paper proposes a RUL prediction method in the federated learning (FL) framework, which aims to develop a lightweight model using network pruning and rebirth strategies. First, a deep convolutional neural network (DCNN) is designed as the prediction model. Next, the Taylor expansion and the l2 norm pruning criteria are executed on the convolutional and fully-connected layers of DCNN to prune some unimportant feature maps and neurons, respectively. After each pruning operation, the network rebirth strategies, including model relocation, federated averaging (FedAvg), and selective retraining, are used to fine-tune the pruned model in the FL. Finally, the network pruning and rebirth occur alternately to produce a compact RUL prediction model with fewer parameters, which can achieve the same good performance as the original one. Experiments study on the C-MAPSS dataset demonstrates the effectiveness of the proposed method.

A similarity-based remaining useful life prediction method using multimodal degradation features and adjusted cosine similarity

When a large amount of full life-cycle data are available, similarity-based methods are the preferred method for remaining useful life (RUL) prediction due to their reliability and accuracy. Traditional similarity-based RUL prediction methods use a single model and single-scale degradation features, which are incapable of fully capturing the degradation behavior of the system. Additionally, the similarity of spatial orientation is neglected in the similarity-matching process. To fill these research gaps, a novel method is developed based on multimodal degradation features and adjusted cosine similarity (ACS) to tackle complex-system RUL prediction in this paper. Complete ensemble empirical mode decomposition with adaptive noise is employed to decouple global degradation and random fluctuations in run-to-failure sensor data. Slow feature analysis is utilized to obtain local degradation features, and residual terms are used as global degradation features. Then, multimodal degradation features are transformed into one-dimensional health degradation indicators by bidirectional gated recurrent unit autoencoder. An ACS is developed to estimate the matching similarity between the test degradation curve and the training degradation curve. The proposed scheme captures the time-varying multimodal degradation behavior and provides libraries of health curves with multiple degradation patterns. The designed scheme is evaluated on the C-MAPSS dataset and the results illustrate the competitiveness and effectiveness of the proposed method.

Dual-Thread Gated Recurrent Unit for Gear Remaining Useful Life Prediction

Remaining useful life (RUL) prediction can provide a foundation for the operation and maintenance of industrial equipment. In order to improve the predictive ability for the complex degradation trajectory, a new dual-thread gated recurrent unit (DTGRU) is explored. It uses a dual-thread learning strategy to mine the stationary and non-stationary information from the input data and the difference of hidden states at two adjacent time steps. Then the state transition updating formulas of DTGRU are derived. Using the collected gear vibration signals and degradation-trend-constrained variational autoencoder, the gear health indicator (HI) is constructed. Based on the constructed HI and DTGRU, a novel RUL prediction method is developed. Via multiple gear life-cycle datasets, the effectiveness of the DTGRU-based RUL prediction approach is verified. Furthermore, compared with the existing typical prediction methods, the experimental results show that DTGRU has higher predictive ability in terms of HI fitting precision and RUL prediction performance.

A new method for remaining useful life prediction by implementing joint learning of sensor dynamic graph and spatio-temporal features

Accurate implementation of remaining useful life (RUL) prediction of equipment is essential for health management and maintenance decisions. Advances in sensing and big data technologies have provided the possibility for research on data-driven life prediction methods. However, the current RUL prediction methods still need to improve in utilizing multi-source information. The present techniques consider more the relationship between the temporal information and RUL prediction among the multi-source information and rarely target the research and exploitation of the potential connection between sensor networks and RUL. Therefore, this paper proposes a spatio-temporal feature extraction network based on the sensor dynamic graph: DST-GT model. The method extracts the unidirectional relationship between sensors from monitoring data to construct a dynamic spatio-temporal graph reflecting the sensor relationship. The DST-GT model uses graph convolution based on message selection (MSGCN) to model the spatial dependencies of sensors, and uses multi-scale gated temporal convolution module (MGTCN) to model the temporal dependencies in sensor state monitoring data. In this paper, graph learning, graph convolution and temporal convolution modules are jointly learned in an end-to-end framework. The results on two widely used datasets and comparisons with other methods demonstrate the accuracy and advancement of DST-GT networks for RUL prediction.

Res-HSA: Residual hybrid network with self-attention mechanism for RUL prediction of rotating machinery

Remaining useful life (RUL) prediction is of great significance for improving maintenance efficiency and ensuring the reliability of rotating machinery. In recent years, there are a large number of deep-learning-based methods for RUL prediction of rotating machinery. However, the effect of conventional end-to-end RUL prediction methods relies on the distribution consistency of training data and test data, and conventional health indicator extrapolation RUL prediction methods are susceptible to interference from abnormal fluctuations in the health curve. To overcome the problem, this paper proposes a new RUL prediction method for rotating machinery using health indicators constructed by the residual hybrid network with self-attention mechanism (Res-HSA). First of all, we propose the residual hybrid network combined with self-attention mechanism to extract the high-level degenerate feature. Then, the health assessment model based on Res-HSA is proposed to generate the health indicators of the equipment. To assist in network training, the segmented data labels based on the degradation rule are applied to optimize the labels of training sets. Finally, to address the problem of abnormal fluctuations in the health curve, a fitting interval selection method is used to optimize conventional curve fitting schemes to calculate RUL. Two public datasets, IEEE-PHM-2012-challenge datasets and C-MAPSS datasets, are used to verify the effectiveness of the proposed method. The experiment results on two public datasets show that the RUL prediction method proposed in this paper has good prediction performance. Compared to the state-of-the-art method, the method proposed in this article reaches the most advanced level in some test projects, while the rest of the projects can be very close to the most advanced method.