Life Prediction Research Articles

Multi-stage degradation feature with dynamic feedback mechanism for remaining useful life prediction

Multi-stage degradation feature with dynamic feedback mechanism for remaining useful life prediction

Multi-Condition Remaining Useful Life Prediction Based on Mixture of Encoders

Accurate Remaining Useful Life (RUL) prediction is vital for effective prognostics in and the health management of industrial equipment, particularly under varying operational conditions. Existing approaches to multi-condition RUL prediction often treat each working condition independently, failing to effectively exploit cross-condition knowledge. To address this limitation, this paper introduces MoEFormer, a novel framework that combines a Mixture of Encoders (MoE) with a Transformer-based architecture to achieve precise multi-condition RUL prediction. The core innovation lies in the MoE architecture, where each encoder is designed to specialize in feature extraction for a specific operational condition. These features are then dynamically integrated through a gated mixture module, enabling the model to effectively leverage cross-condition knowledge. A Transformer layer is subsequently employed to capture temporal dependencies within the input sequence, followed by a fully connected layer to produce the final prediction. Additionally, we provide a theoretical performance guarantee for MoEFormer by deriving a lower bound for its error rate. Extensive experiments on the widely used C-MAPSS dataset demonstrate that MoEFormer outperforms several state-of-the-art methods for multi-condition RUL prediction.

A Deep-Learning Method for Remaining Useful Life Prediction of Power Machinery via Dual-Attention Mechanism

Remaining useful life (RUL) prediction is a cornerstone of Prognostic and Health Management (PHM) for power machinery, playing a crucial role in ensuring the reliability and safety of these critical systems. In recent years, deep learning techniques have shown great promise in RUL prediction, providing more reliable and accurate outcomes. However, existing models often struggle with comprehensive feature extraction, especially in capturing the complex behavior of power machinery, where non-linear degradation patterns arise under varying operational conditions. To tackle this limitation, we propose a multi-feature fusion model leveraging a dual-attention mechanism. Initially, convolutional neural networks (CNNs) and channel attention mechanisms are employed to preliminarily extract spatial features. Subsequently, a layer combining a Gate Recurrent Unit (GRU) and self-attention mechanisms is used to further extract and integrate temporal features. Finally, RUL values are predicted via regression. The effectiveness of the proposed method was validated on C-MAPSS datasets, and its superior performance in RUL prediction was demonstrated through comparative analysis with other methods.

Remaining useful life prediction integrating working conditions and uncertainty quantification based on multilayer graph neural networks

Remaining useful life prediction integrating working conditions and uncertainty quantification based on multilayer graph neural networks

A novel multi-sensor data fusion enabled health indicator construction and remaining useful life prediction of aero-engine

Remaining useful life (RUL) prediction is vital to formulate a suitable maintenance strategy in manufacturing systems health management. Multisensor data fusion of complex engineering systems has attracted substantial attention due to the fact that a single sensor can only collect partial information. Health indicator (HI) construction plays a crucial role in multisensor data fusion and machinery prognostic, mainly because it attempts to quantify a history and ongoing degradation process by fusing the advantages of multiple sensors. However, large numbers of coefficients are involved for most of the existing HIs. Additionally, simplifications during modeling may inhibit the wide application of the constructed HI. To address these two challenges, a new multisensor data fusion method is proposed in this paper by constructing a HI for the characterization of the degradation process. Firstly, the sensors that collect invalid data or conflicting data are removed through a correlation coefficient operation. Then, principal component analysis (PCA) is adopted to reduce the number of coefficient before constructing the HI. Furthermore, the objective function is constructed under the comprehensive consideration of the three factors of the HI, that is, monotonicity, trendability, and fitting errors. The effectiveness of the proposed method is verified using the C-MAPSS dataset. Multiple comparison results show that the HI possesses excellent performance in both degradation characterization and remaining useful life prediction.

A generalized diffusion model for remaining useful life prediction with uncertainty

Forecasting the remaining useful life (RUL) is a crucial aspect of prognostics and health management (PHM), which has garnered significant attention in academic and industrial domains in recent decades. The accurate prediction of RUL relies on the creation of an appropriate degradation model for the system. In this paper, a general representation of diffusion process models with three sources of uncertainty for RUL estimation is constructed. According to time-space transformation, the analytic equations that approximate the RUL probability distribution function (PDF) are inferred. The results demonstrate that the proposed model is more general, covering several existing simplified cases. The parameters of the model are then calculated utilizing an adaptive technique based on the Kalman filter and expectation maximization with Rauch-Tung-Striebel (KF-EM-RTS). KF-EM-RTS can adaptively estimate and update unknown parameters, overcoming the limits of strong Markovian nature of diffusion model. Linear and nonlinear degradation datasets from real working environments are used to validate the proposed model. The experiments indicate that the proposed model can achieve accurate RUL estimation results.

Lube oil life prediction for heavy earth moving machinery (HEMM): A machine learning approach

Evaluating lubricant life is crucial for maintaining equipment reliability and preventing failures. Conventional methods often depend on original equipment manufacturer recommendations for lubricant changes, which may result in the premature disposal of operationally effective lubricants, leading to economic costs and degrading overall efficiency. The chemical properties of these samples are evaluated by calculating multiple parameters such as total acid number, total base number, oxidation index, soot level, and water contamination. In addition, rheological properties through viscosity index analysis and the tribological properties via friction-wear analysis are determined. In this study, an artificial neural network (ANN) (a four-layer perceptron) and an adaptive neuro-fuzzy inference system (ANFIS) are applied to predict oil conditions based on multiple calculated parameters. Performance and comparison of these advanced mathematical models are evaluated using statistical indices. Overall, the artificial intelligence (AI)-powered approach proved effective in predicting lubricant life for HEMM. Among the AI models, the ANN model demonstrated particularly strong performance, with a correlation coefficient of 0.99 compared to 0.98 for the ANFIS model. Implementing the ANN model could lead to a potential 19% reduction in current engine oil expenses, which would lower operating costs and decrease environmental impact by reducing the frequency of oil disposal.

Multi-dimensional failure risk load prediction of electric motor based on physical model and data-driven approach

Dynamic prediction of failure risk loads is a critical prerequisite for system damage monitoring and service life evaluation. As the failure mechanism of electric vehicle drive motors under multi-source load excitations is complex, and the risk load poses challenges in rapid acquisition, a physics-based and data-driven method for multi-dimensional failure risk load prediction of motors is proposed. Based on actual operational data collected from users, we analyze the failure risk loads of various components of a motor. A physical simulation model of the motor is established to extract multi-dimensional failure risk load time history, and the effectiveness of the simulation model is verified using data from bench tests. Moreover, using the physical model simulation data as a foundation, a two-stage data-driven model is constructed. In the first stage, the genetic algorithm optimized back propagation neural network (GA-BPNN) model is implemented for multi-dimensional loss prediction of the motor, and serves as the primary input for the second stage model, which integrates prediction accuracy and training efficiency. In the second stage, by comprehensively considering the spatial and temporal correlation features with the input layer reconstructed using correlation matrix weights, an improved convolutional neural network with bidirectional long short-term memory (CNN-BiLSTM) model is employed to predict multi-scale loads. Compared with traditional models, the proposed method exhibits significantly improved prediction accuracy, with R2 greater than 0.95. Furthermore, the damage error between the predicted load and the expected load is within 10%, which provides technical support for reliability evaluation and life prediction of electric motor.

Remaining Useful Life Prediction of Lithium-ion Battery based on AConvST-LSTM-Net-TL

Abstract The accurate prediction of the Remaining Useful Life (RUL) of lithium-ion batteries can significantly enhance the safety and reliability of these batteries, thereby reducing operational risks. However, numerous existing methodologies operate under the assumption that both training and testing data adhere to the same distribution pattern, hindering the application of successful laboratory-based models to different target batteries. Hence, this study introduces a transfer learning model for predicting lithium-ion battery life, utilizing an attention mechanism-driven convolutional neural network (ACNN) in conjunction with a spatiotemporal Long Short-Term Memory network (ST-LSTM). Initially, the deep charging process data is leveraged for battery capacity estimation, where a Generalized Additive Model (GAM) is employed to delineate the nonlinear trend of battery capacity decay and characterize the capacity degradation. Following feature extraction, the Maximum Mean Discrepancy (MMD) value is computed to assess the distribution disparity between the two domains. Subsequently, the AConvST-LSTM-Net computes capacity estimations, loss functions for both source and target domains, merging these with the MMD value to formulate a comprehensive loss function. In conclusion, by employing transfer learning, the refined model is adapted to the target domain dataset, enabling the accurate prediction of the RUL of lithium-ion batteries. This approach is validated using the NASA lithium-ion battery dataset and the National Big Data Alliance for New Energy Vehicles (NDANEV), demonstrating the superior accuracy and robustness of the AConvST-LSTM-Net-TL model in comparisn to other prediction methods.

Reusable Turbine Material GH4586: Stochastic Analysis and Reliability Assessment of Low‐Cycle Fatigue Life Parameters at Multiple Temperatures

ABSTRACTTurbines, crucial in reusable rocket engines, benefit significantly from precise fatigue life forecasting in their force‐thermal cyclic loading conditions. Further, low‐cycle fatigue life predictions can be random due to uncertainties in material properties. This study aims to analyze the probabilistic low‐cycle fatigue life of GH4586 under different temperature. Mechanical and thermal tests were carried out for the temperatures between 87 K and 1173 K. Random distributions of low‐cycle fatigue parameters were obtained using various methods. The cyclic life reliability of a reusable rocket engine turbine rotor blisk was analyzed using stochastic low‐cycle fatigue parameters. Mitchell's method had a higher prediction accuracy when the reliability was greater than 0.8. Sensitivity analysis shows the significant impact of GH4586 fatigue parameters on cyclic life reliability. Sensitivity coefficients of strength coefficient and ductility coefficient are 32.73% and 27.06%, respectively. These findings provide valuable insights that inform the design of turbines and enhance their reliability.

Remaining useful life prediction of rolling bearings using a residual attention network with multi-scale feature extraction and temporal dependency enhancement

ABSTRACT Predicting the remaining useful life (RUL) of rolling bearings is crucial for industrial machinery maintenance, non-destructive testing and evaluation (NDT). To address the challenges posed by noise interference and redundant information, this paper proposes a novel approach utilising residual attention networks and multi-scale feature extraction. The method enhances feature extraction by combining shallow and deep convolutional layers while employing bidirectional LSTM to capture both short-term and long-term dependencies in time series data. The incorporation of a residual attention fusion module further enhances the model’s ability to focus on important features, ensuring more stable training and better prediction performance. After validation on PHM2012, IMS, and laboratory self-constructed datasets, the proposed model demonstrates superior performance compared to existing methods, significantly reducing prediction errors. The practical application of this work lies in its potential integration into industrial predictive maintenance systems, providing a solid foundation for expanding predictive maintenance strategies in complex real-world industrial environments.

Fatigue Life Prediction of 2024-T3 Clad Al Alloy Based on an Improved SWT Equation and Machine Learning

The multi-parameter and nonlinear characteristics of the Smith Watson Topper (SWT) equation present considerable challenges for predicting the fatigue life of 2024-T3 clad Al alloy. To overcome these challenges, a novel model integrating traditional fatigue analysis methods with machine learning algorithms is introduced. An improved SWT fatigue life prediction equation is developed by incorporating key factors such as the mean stress effect, stress concentration factor, and surface roughness coefficient. Extreme gradient boosting, Random Forest, and their derived models are used to construct the fatigue life prediction model. The L-BFGS algorithm was then integrated with the established machine learning model to solve for the multi-parameter of the improved SWT equation. Thus, an accurate modified SWT prediction equation for 2024-T3 clad Al alloy was obtained. To further optimize the solution, the deep deterministic policy gradient and deep reinforcement learning algorithms are introduced to dynamically optimize the nonlinear equation, achieving a more efficient and accurate solution. The improved SWT fatigue life prediction equation and its solution method proposed in this study provide new insights for fatigue life prediction of clad metallic materials.

Numerical Prediction of Fatigue Life for Landing Gear Considering the Shock Absorber Travel

Due to the complexity of the landing gear’s (LG) structural integrity and its loads under various static or dynamic working conditions, the fatigue life assessment for LG is a highly challenging task. On the basis of the whole geometric model of a large passenger aircraft’s main landing gear (MLG), the quasi-static finite element model (FEM) of the whole MLG is established, and the high-cycle fatigue issue of the Main Fitting (MF) is studied by considering the variation in shock absorber travel (SAT). Firstly, the ground loads under actual fatigue conditions are equivalently converted into the forces acting on the center of the left and right axles of the MLG, and based on these spatial force decompositions, the magnitude and direction of the load for 12 different basic unit load cases (ULC) are obtained. That is, the stress of the MLG under actual fatigue conditions can be obtained by superimposing these ULCs. Then, considering that the SAT of the MLG varies under different fatigue conditions, and to reduce the number of finite element (FE) simulations, this article simplifies all the SAT experienced by the MLG into seven specific values, so as to establish seven quasi-static FEMs of the MLG with the specified stroke of the shock absorber. In this way, the fatigue stress of the MLG with any actual SAT can be obtained by interpolating the stress components of the seven FEMs. Only 84 FE simulations are needed to efficiently obtain the fatigue stress spectra from the ground load spectra. Finally, according to the material S-N curve and Miner’s damage accumulation criterion, evaluate the fatigue life of the Main Fitting. The results of the stress component interpolation and superposition method show that at least five different SATs of the whole MLG’s FEM are needed to effectively convert the fatigue loads into a stress spectrum. The fatigue life prediction results indicate that the minimum lifespan of the MF is 53164 landings, which means that the fatigue life meets the requirement design.

A new dual-channel transformer-based network for remaining useful life prediction

A new dual-channel transformer-based network for remaining useful life prediction

Stage-Based Remaining Useful Life Prediction for Bearings Using GNN and Correlation-Driven Feature Extraction

Stage-Based Remaining Useful Life Prediction for Bearings Using GNN and Correlation-Driven Feature Extraction

A Novel Mechanism-Equivalence-Based Tweedie Exponential Dispersion Process for Adaptive Degradation Modeling and Life Prediction

A Novel Mechanism-Equivalence-Based Tweedie Exponential Dispersion Process for Adaptive Degradation Modeling and Life Prediction

Temporal convolution long short-term memory network with multiple attention for remaining useful life prediction of rolling bearings

Temporal convolution long short-term memory network with multiple attention for remaining useful life prediction of rolling bearings

Characterization of Fatigue Properties of Fiber-Reinforced Polymer Composites Based on a Multiscale Approach

This study presents a methodology for characterizing the constituent properties of composite materials by back-calculating from the laminate behavior under fatigue loading. Composite materials consist of fiber reinforcements and a polymer matrix, with the fatigue performance of the laminate governed by the interaction between these constituents. Due to the challenges in directly measuring the properties of individual fibers and the polymer matrix, a reverse-engineering approach was employed. Using the micro-mechanics of fatigue (MMFatigue), we predicted the laminate’s fatigue behavior based on assumed constituent properties and compared these predictions with experimental data from fatigue tests. The properties of the fiber and polymer matrix were iteratively adjusted to minimize the differences between predictions and experimental results, enabling accurate fatigue characterization. To ensure robustness, three laminate angles—0°, 30°, and 60°—were evaluated at three temperatures: low temperature (LT: −40 °C), room temperature (RT: 25 °C), and high temperature (HT: 85 °C). The error, defined as the fatigue life difference between the prediction and the experimental results, were obtained as 2.48% at LT, 7.18% at RT, and 1.25% at HT for a laminate angle of 45°. Finally, the applicability of the multiscale-based fatigue life prediction method was demonstrated through studies on laminates with various angles under tension–compression, and compression–compression cyclic loads, as well as composite pressure vessels under cyclic loading.

Uncertainty-corrected fractional generalized Pareto motion for lithium-ion battery life prediction and value-at-risk-based maintenance framework

Uncertainty-corrected fractional generalized Pareto motion for lithium-ion battery life prediction and value-at-risk-based maintenance framework

Finite-Element-Based Time-Dependent Service Life Prediction for Carbonated Reinforced Concrete Aqueducts

This study proposes a time-dependent reliability analysis method for aqueduct structures based on concrete carbonation and finite element analysis. The primary goal of this study is to improve the reliability assessment of reinforced concrete aqueducts by incorporating environmental factors such as carbonation over time. First, a three-dimensional finite element model of a reinforced concrete aqueduct is established using the Midas 2022 Civil software, incorporating a time-varying function derived from a predictive model of concrete carbonation depth. Point estimation is then integrated with structural finite element analysis to calculate the first four moments of random variables as functions of concrete carbonation. Additionally, the original performance function is transformed into a normal distribution using dual power transformation and the Jarque–Bera test. The high-order unscented transformation (HUT) is subsequently employed to estimate the first four moments of the transformed performance function, facilitating the calculation of time-varying reliability indices for the carbonated concrete aqueduct. Based on the time-varying reliability index data, a reliability function corresponding to different time points is fitted and applied to service life prediction. The results demonstrate that the proposed method effectively reduces large errors associated with the fourth-moment method in calculating large reliability indices. Furthermore, the comparison with Monte Carlo simulation (MCS) results validates the high efficiency and accuracy of the proposed method, offering a valuable tool for addressing the reliability challenges of aqueducts exposed to carbonation and other environmental factors over time.