Real-time Remaining Useful Life Research Articles

Transfer learning-based deep learning models for proton exchange membrane fuel remaining useful life prediction

Proton exchange membrane fuel cells (PEMFCs) offer power generation capabilities for diverse applications including commercial enterprises, industrial sectors, and residential technologies. Nevertheless, the comprehensive integration of PEMFC applications could be improved by challenges related to degradation and durability. The imperative development of efficient performance prognostic models assumes a pivotal role in the prognosis of remaining useful life (RUL), health monitoring, and effective utilization of PEMFCs. This paper centers on the prognostication of critical components within PEMFCs and introduces a transfer learning approach based on variational autoencoder and bi-directional long short-term memory with an attention mechanism (Bi-LSTM-AM) model. This approach combines feature fusion, knee-point detection, and a sophisticated deep-learning-based predictive model. Notably, incorporating the variational autoencoder as the framework for feature fusion introduces a novel perspective previously unexplored. Identifying the knee point and knowing the start point on the training data, facilitates optimized parameter computation. The application of transfer learning facilitates the transfer of optimal model parameters and weights from a source to a target dataset. Conclusively, the estimation of stack voltage degradation and real-time RUL prediction based on the test dataset is executed by implementing our proposed method. The stack voltage prediction findings showcase the Bi-LSTM-AM model’s superior performance relative to comparison models. The proposed online rolling prediction model, utilizing a sliding window technique for RUL prediction, yields significantly enhanced accuracy, culminating in a relative error margin ranging from approximately 1.69% to 5.04%.

Residual life prediction of lithium-ion batteries based on data preprocessing and a priori knowledge-assisted CNN-LSTM

Lithium-ion batteries have become widely used in many industries due to their outstanding performance, making it vital to accurately predict the remaining useful life (RUL) of these batteries. This will aid in developing energy allocation strategies and ensure the safe use of lithium batteries. To overcome the issue of inaccurate RUL prediction, a new method is proposed that leverages data preprocessing and a prior knowledge-assisted convolutional neural network-long short-term memory neural network (CNN-LSTM). This method utilizes capacity as the health factor and employs complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to decompose the capacity sequence, eliminating noise components through data reconstruction. The reconstructed capacity sequence data are then used to pretrain the CNN-LSTM neural network, forming a priori knowledge. Finally, real-time battery capacity data are used to train the prior knowledge-aided CNN-LSTM neural network for real-time RUL prediction of Lithium-ion batteries. The results show that this method significantly improves the RUL prediction accuracy and reduces the prediction error while being more robust than existing methods.

Availability Analysis of the Critical Production System in SMEs Using the Markov Decision Model

Small and medium-sized enterprises (SMEs) in today’s world have numerous issues in ensuring the availability and safety of critical machines and their components in the working environment. As a result, SMEs considered planning and scheduling maintenance tasks to be a significant threat. The goal of this research is to identify the critical subsystem of the automobile spare parts production plant in the southern region of Tamil Nadu, India, and then to prioritize the maintenance activity and set up an architecture for autonomous preventive maintenance (PM) in SMEs that includes an optimal decision support system. The transition state diagram of the individual and simultaneous production system has been developed with the application of the Markov decision model approach. It was used to analyze the present variables of the production systems and forecast the optimal maintenance parameters such as the failure rate and the repair rate, through the production system’s availability analysis. This availability analysis reveals that system B (Piercing) is classified as the most critical because of the abrupt availability variation compared to all other production systems, concerning the corresponding maintenance parameters such as a failure rate of 0.0371, a repair rate of 0.7094, and the availability of the piercing system of 0.5056. Finally, the use of an autonomous PM management system and the most effective maintenance workforce has enhanced productivity and customer satisfaction in SMEs. The predictive maintenance management system has been further investigated to determine the real-time remaining useful life (RUL) of critical systems in the automobile spare parts manufacturing plant in the southern region of Tamil Nadu, India.

Real-time remaining useful life prediction based on adaptive kernel window width density

Remaining useful life (RUL) prediction plays an important role in improving the availability and productivity of systems. To improve the accuracy of real-time RUL prediction during system operation, we propose a modeling method for real-time RUL prediction based on adaptive kernel window width density. First, a non-parametric kernel density estimation (KDE) real-time RUL prediction model is proposed, and a window width model with adaptive kernel window width density is established by introducing a local density factor in the window width selection. The local density of sample points is calculated by the k-nearest neighbor distance, and the KDE is performed by adaptively selecting the window width value according to the local density of sample points in the region of non-uniform distribution of sample points. As the monitoring data changes in real time, the kernel density estimates of known samples are used to recursively update the kernel density estimates of new samples. Moreover, the logarithmic transformation of random variables and space mapping are used in the establishment of the RUL prediction model. A model of logarithmic kernel diffeomorphism transformation is established to solve the boundary shift problem of kernel estimation in the prediction to improve the prediction accuracy. Finally, the validity of the method is verified through case studies, and the accuracy of the model is judged using evaluation quasi-measures.

Data-driven dynamic adaptive replacement policy for units subject to heterogeneous degradation

This paper proposes a data-driven dynamic adaptive replacement policy for units subject to heterogeneous degradation. A Wiener process with random drift parameter is used to model the heterogeneous degradation. For each individual unit, the real-time degradation data contain the useful information about its unique degradation pattern. Thus the real-time degradation data of an operating unit is used to dynamically estimate the posteriori distribution of its drift parameter and further update its remaining useful life (RUL) distribution. With the real-time RUL distribution, a cost model is developed based on the one-cycle criterion, aiming at minimum expected cost per unit of time for the cycle of the operating unit. The failure rate from the RUL distribution is proven to non-monotonic, first increasing and then decreasing. The monotonicity of the cost function under such a failure rate is analyzed, based on which the optimal preventive replacement (PR) interval is derived analytically. The cost function changes dynamically with the updating of RUL predictions and thus enables the replacement plan to be adaptively revised. The effectiveness and advantages of the proposed policy are illustrated through a simulation study and a practical case study.

Remaining Useful Life Prediction of Rolling Bearings Based on Segmented Relative Phase Space Warping and Particle Filter

Predictive maintenance plays a crucial role in the field of intelligent machinery fault diagnosis, which improves the efficiency of maintenance. This paper focuses on the extraction of real-time damage feature and the prediction of remaining useful life (RUL) in predictive maintenance of rolling bearings. Some RUL prediction approaches lack dynamic foundations and require large amounts of data and prior knowledge. This paper proposes the algorithm of segmented relative phase space warping (SRPSW) and a strategy combining double exponential model (DEM) and particle filter (PF) to predict the RUL. SRPSW provides a dynamic basis for real-time RUL prediction in different stages. The DEM-based PF reduces the need for prior knowledge and improves accuracy. The analysis results from normal and accelerated degradation experiments show that the proposed SRPSW overcomes the failure of the original PSW in depicting the later operating stage of bearings. Further, the relative damage indicators (RDIs) extracted by SRPSW are more accurately than commonly used indicators. The predicted results show that the DEM-based PF does not require professional and prior information whilst ensuring the accuracy of RUL prediction. The proposed approach in this paper provides a new avenue for predictive maintenance of bearings.

A hybrid prognostic method based on gated recurrent unit network and an adaptive Wiener process model considering measurement errors

Remaining useful life (RUL) prediction is fundamental to prognostics and health management (PHM). Considering the advantages of both model-based and data-driven prognostic approaches, this paper develops a hybrid prognostic method for machinery degradation. First, a 3σ criterion-based algorithm is introduced to detect the initial timepoint of degradation. Second, gated recurrent unit (GRU) network is utilized to learn the degradation characteristics based on the available data and thereby predict the long-term degradation trend by a multi-prediction procedure. Then, an adaptive Wiener process model considering measurement errors is constructed. The states of this model consisting of the drift rate and the underlying degradation value are updated adaptively based on the monitored observations and the predictions by GRU using a Kalman filtering algorithm. The predicted values of the RUL can be determined according to the underlying degradation and the failure threshold. Finally, to account for the drift adaptivity in the future degradation, exponentially weighted average method is adopted to aggregate the estimated drift sequence from the current time until failure for the derivation of real-time RUL distributions. The effectiveness and superiority are illustrated by a simulation study and an application to rolling element bearings.

Real-Time Prediction of Remaining Useful Life and Preventive Maintenance Strategy Based on Digital Twin

AbstractThe functional parts of a machine tool determine its reliability level to a great extent. The failure prediction of the functional part is helpful to prepare the maintenance scheme in time, in order to ensure a stable manufacturing process and the required production quality. Due to the rise of digital twin (DT), which has the characteristics of virtual reality interaction and real-time mapping, a DT-based real-time prediction method of the remaining useful life (RUL) and preventive maintenance scheme is proposed in this study. In this method, a DT model of the manufacturing workshop is established based on real-time perceptual information obtained by the proposed acquisition method. Subsequently, the real-time RUL of the functional part is predicted by establishing an RUL prediction model based on the nonlinear-drifted Brownian motion, which takes the working conditions and measurement errors into consideration. On this basis, the optimal preventive maintenance scheme can be determined and fed back to the manufacturing workshop, in order to guide the maintenance of relevant parts. Finally, an example case study is presented to illustrate the feasibility and effectiveness of the proposed method.

Similarity-based information fusion grey model for remaining useful life prediction of aircraft engines

PurposeOnline health monitoring of large complex equipment has become a trend in the field of equipment diagnostics and prognostics due to the rapid development of sensing and computing technologies. The purpose of this paper is to construct a more accurate and stable grey model based on similar information fusion to predict the real-time remaining useful life (RUL) of aircraft engines.Design/methodology/approachFirst, a referential database is created by applying multiple linear regressions on historical samples. Then similarity matching is conducted between the monitored engine and historical samples. After that, an information fusion grey model is applied to predict the future degradation trajectory of the monitored engine considering the latest trend of monitored sensory data and long-term trends of several similar referential samples, and the real-time RUL is obtained correspondingly.FindingsThe results of comparative analysis reveal that the proposed model, which is called similarity-based information fusion grey model (SIFGM), could provide better RUL prediction from the early degradation stage. Furthermore, SIFGM is still able to predict system failures relatively accurately when only partial information of the referential samples is available, making the method a viable choice when the historical whole life cycle data are scarce.Research limitations/implicationsThe prediction of SIFGM method is based on a single monotonically changing health indicator (HI) synthesized from monitoring sensory signals, which is assumed to be highly relevant to the degradation processes of the engine.Practical implicationsThe SIFGM can be used to predict the degradation trajectories and RULs of those online condition monitoring systems with similar irreversible degradation behaviors before failure occurs, such as aircraft engines and centrifugal pumps.Originality/valueThis paper introduces the similarity information into traditional GM(1,1) model to make it more suitable for long-term RUL prediction and also provide a solution of similarity-based RUL prediction with limited historical whole life cycle data.

Improved trajectory similarity-based approach for turbofan engine prognostics

Trajectory similarity-based prediction (TSBP) is an emerging real-time remaining useful life (RUL) prediction method that has drawn considerable attention in the field of data-driven prognostics. TSBP is fast, and the corresponding model is easy to train. However, TBSP only provides a point estimation of RUL, which is insufficient for some specific prognostic applications. Hence, this study introduces an improved TSBP method to handle the issue of prognostic uncertainty. On the basis of an adaptive kernel density estimation technique and β-criterion, the improved TSBP method not only provides an accurate and precise point prediction of RUL but also specifies the confidence interval of RUL prediction. The capability of obtaining the confidence interval of RUL can enhance the TSBP method for uncertainty management. The effectiveness of the proposed method is validated through two cases studies, which are related to turbofan engine prognostics.

A service-oriented dynamic multi-level maintenance grouping strategy based on prediction information of multi-component systems

The development of product-service-system (PSS) urges the emergence of service-oriented maintenance mode (SOMA), which leads to the shift of maintenance manners from traditional high-cost in-house maintenance to the performance-based proactive service offerings for complex multi-component systems. In this case, a novel and adaptive maintenance grouping strategy that considers service responses, service interactions and dynamic prediction requirements is desired to implement the predictive maintenance planning. Thus, this paper presents a service-oriented dynamic multi-level predictive maintenance grouping strategy. It involves component level, dynamic grouping level and system level to fulfill the individual optimization and grouping optimization. Firstly, the individual service time based on real-time remaining useful life (RUL) distribution information is optimized and obtained with the minimum average cost. Secondly, considering the existing economic and resource dependence in predictive and corrective services, this paper dynamically groups the predicted optimal individual services from rolling planning horizon at each calculated service planning time. The penalty costs and grouping service costs are both constructed. Then, a modified k-means method is designed to find out the optimal short-term grouping execution strategy from medium-term prediction information where the average cost savings and relative availability improvement degree are regarded as the optimization objectives. Finally, some numerical examples with three predictive grouping scenarios show that our grouping strategy could provide a feasible and effective long-term dynamic maintenance planning for proactive OEM or PSS providers. It is adapted to the SOMA on complex multi-component systems.

Developing ladder network for intelligent evaluation system: Case of remaining useful life prediction for centrifugal pumps

Intelligent evaluation system has been widely used in industries to estimate essential indexes which are unable to be measured directly through physical devices. Due to the complexity of labeling samples, common data-driven techniques such as supervised learning are developed on a small number of labeled data, while a large amount of unlabeled data is discarded. The amount of labeled information greatly limits the improvement of prediction accuracies. Furthermore, conventional evaluation approaches have only static structures, which makes the dynamic characteristics of parameters difficult to be presented. This paper proposes a ladder network (LN) based semi-supervised learning model to evaluate parameter dynamics, and a case of remaining useful life (RUL) prediction for centrifugal pumps is illustrated. LN datasets comprise a small part of labeled data and a large amount of unlabeled data. We exploited fluid-structure interaction (FSI) numerical simulation to replace actual monitoring, as well as built a RUL prediction model to annotate useful life for offline datasets. After that, the RUL was performed in the online stage by substituting real-time monitored variables into the network. The case study indicates that the LN-based intelligent evaluation system identifies the real-time RUL profile and achieves better predictive outcomes than supervised learning approaches.

Average Accumulative Based Time Variant Model for Early Diagnosis and Prognosis of Slowly Varying Faults.

Early detection of slowly varying small faults is an essential step for fault prognosis. In this paper, we first propose an average accumulative (AA) based time varying principal component analysis (PCA) model for early detection of slowly varying faults. The AA based method can increase the fault size as well as decrease the noise energy. Then, designated component analysis (DCA) is introduced for developing an AA-DCA method to diagnose the root cause of the fault, which is helpful for the operator to make maintenance decisions. Combining the advantage of the cumulative sum (CUSUM) based method and the AA based method, a CUSUM-AA based method is developed to detect faults at earlier times. Finally, the remaining useful life (RUL) prediction model with error correction is established by nonlinear fitting. Once online fault size defined by detection statistics is obtained by an early diagnosis algorithm, real-time RUL prediction can be directly estimated without extra recursive regression.

A prediction method for the real-time remaining useful life of wind turbine bearings based on the Wiener process

Abstract A performance degradation model and a real-time remaining useful life (RUL) prediction method are proposed on the basis of temperature characteristic parameters to determine the RUL of wind turbine bearings. First, using the moving average method, the relative temperature data of wind turbine bearings are smoothed, and the temperature trend data are obtained on the basis of the uncertainty of wind speed and wind direction that causes the temperature of wind turbine bearings to vary widely. Second, given that the degradation speed of bearings changes with operational time and uncertain external factors, the performance degradation model is established with the Wiener process. The parameters of this model are obtained through the maximum likelihood estimation method. Third, according to the failure principle of the first temperature monitoring value beyond the first warning threshold, the RUL prediction model for wind turbine bearings is established on the basis of an inverse Gaussian distribution. Finally, the performance degradation process and real-time RUL prediction are demonstrated by predicting the RUL of a practical rear bearing of a wind turbine generator. The comparison of the predicted RUL and actual RUL shows that the proposed model and prediction method are correct and effective.

Remaining useful life estimation based on Wiener degradation processes with random failure threshold

Remaining useful life (RUL) estimation based on condition monitoring data is central to condition based maintenance (CBM). In the current methods about the Wiener process based RUL estimation, the randomness of the failure threshold has not been studied thoroughly. In this work, by using the truncated normal distribution to model random failure threshold (RFT), an analytical and closed-form RUL distribution based on the current observed data was derived considering the posterior distribution of the drift parameter. Then, the Bayesian method was used to update the prior estimation of failure threshold. To solve the uncertainty of the censored in situ data of failure threshold, the expectation maximization (EM) algorithm is used to calculate the posteriori estimation of failure threshold. Numerical examples show that considering the randomness of the failure threshold and updating the prior information of RFT could improve the accuracy of real time RUL estimation.

A Run-Time Dynamic Reconfigurable Computing System for Lithium-Ion Battery Prognosis

As safety and reliability critical components, lithium-ion batteries always require real-time diagnosis and prognosis. This often involves a large amount of computation, which makes diagnosis and prognosis difficult to implement, especially in embedded or mobile applications. To address this issue, this paper proposes a run-time Reconfigurable Computing (RC) system on Field Programmable Gate Array (FPGA) for Relevance Vector Machine (RVM) to realize real-time Remaining Useful Life (RUL) estimation. The system leverages state-of-the-art run-time dynamic partial reconfiguration technology and customized computing circuits to balance the hardware occupation and computing efficiency. Optimal hardware resource consumption is achieved by partitioning the RVM algorithm according to a multi-objective optimization. Moreover, pipelined and parallel computation circuits for kernel function and matrix inverse are proposed on FPGA to further accelerate the computation. Experimental results with two different battery data sets show that, without sacrificing the RUL prediction performance, the embedded RC platform significantly reduces the computation time and the requirement of hardware resources. This demonstrates that complex prognostic tasks can be implemented and deployed on the proposed system, and it can be extended to the embedded computation of other machine learning algorithms.

A multi-time scale approach to remaining useful life prediction in rolling bearing

This paper presents a novel multi-time scale approach to bearing defect tracking and remaining useful life (RUL) prediction, which integrates enhanced phase space warping (PSW) with a modified Paris crack growth model. As a data-driven method, PSW describes the dynamical behavior of the bearing being tested on a fast-time scale, whereas the Paris crack growth model, as a physics-based model, characterizes the bearing's defect propagation on a slow-time scale. Theoretically, PSW constructs a tracking metric by evaluating the phase space trajectory warping of the bearing vibration data, and establishes a correlation between measurement on a fast-time scale and defect growth variables on a slow-time scale. Furthermore, PSW is enhanced by a multi-dimensional auto-regression (AR) model for improved accuracy in defect tracking. Also, the Paris crack growth model is modified by a time-piecewise algorithm for real-time RUL prediction. Case studies performed on two run-to-failure experiments indicate that the developed technique is effective in tracking the evolution of bearing defects and accurately predict the bearing RUL, thus contributing to the literature of bearing prognosis .

Tool life predictions in milling using spindle power with the neural network technique

Tool wear is an important limitation to machining productivity and part quality. In this paper, remaining useful life (RUL) prediction of tools is demonstrated based on the machine spindle power values using the neural network (NN) technique. End milling tests were performed on a stainless steel workpiece at different spindle speeds and spindle power was recorded. The NN curve fitting approach with different MATLAB™ training functions was applied to the root mean square power (Prms) values. Sample Prms growth curves were generated to take into account uncertainty. The Prms value in the time domain was found to be sensitive to tool wear. Results show a good agreement between the predicted and true RUL of tools. The proposed method takes into account the uncertainty in tool life and the percentage increase in nominal Prms value during the RUL prediction. Using MATLAB™ on an Intel i7 processor, the computation takes 0.5s Thus, the method is computationally inexpensive and can be incorporated for real time RUL predictions during machining.

A novel multistage Support Vector Machine based approach for Li ion battery remaining useful life estimation

Real-time prediction of remaining useful life (RUL) is an essential feature of a robust battery management system (BMS). In this work, a novel method for real-time RUL estimation of Li ion batteries is proposed that integrates classification and regression attributes of Support Vector (SV) based machine learning technique. Cycling data of Li-ion batteries under different operating conditions are analyzed, and the critical features are extracted from the voltage and temperature profiles. The classification and regression models for RUL are built based on the critical features using Support Vector Machine (SVM). The classification model provides a gross estimation, and the Support Vector Regression (SVR) is used to predict the accurate RUL if the battery is close to the end of life (EOL). By the critical feature extraction and the multistage approach, accurate RUL prediction of multiple batteries is accomplished simultaneously, making the proposed method generic in nature. In addition to accuracy, the multistage approach results in faster computations, and hence a trained model can potentially be used for real-time onboard RUL estimation for electric vehicle battery packs.

Real time remaining useful life prediction based on nonlinear Wiener based degradation processes with measurement errors

Real time remaining useful life (RUL) prediction based on condition monitoring is an essential part in condition based maintenance (CBM). In the current methods about the real time RUL prediction of the nonlinear degradation process, the measurement error is not considered and forecasting uncertainty is large. Therefore, an approximate analytical RUL distribution in a closed-form of a nonlinear Wiener based degradation process with measurement errors was proposed. The maximum likelihood estimation approach was used to estimate the unknown fixed parameters in the proposed model. When the newly observed data are available, the random parameter is updated by the Bayesian method to make the estimation adapt to the item’s individual characteristic and reduce the uncertainty of the estimation. The simulation results show that considering measurement errors in the degradation process can significantly improve the accuracy of real time RUL prediction.