Reliability analysis is essential to guide maintenance strategies in structural health monitoring of complex systems that have multiple variables and suffer from multiple failure modes. However, most existing prognostic approaches are based on sufficient data from a class of systems, and focus on univariate systems with one failure mode. This paper proposes a novel approach to address this problem. Individual monitoring data is combined with physical information to recursively estimate the performance states of the system using the extended Kalman filter. The correlation among multiple variables is characterized by constructing multivariate distributions derived from estimation results. In addition, the impact of the correlation between multiple failure modes on system reliability is comprehensively investigated through the formulation of multidimensional functional variables and receptive fields. The failure rate is then derived to achieve real-time evaluation of system reliability. Furthermore, the distribution of future performance states is predicted considering multi-source uncertainty propagation, and system reliability is predicted using Bayes' theorem. Finally, a comparative case study concerning a liquid-level control system is presented to demonstrate the effectiveness of the proposed technique in reliability evaluation and prediction.

Condition-based maintenance (CBM) powered by inspection/monitoring technology is crucial to guarantee safety and economical operations of various industrial assets. The implementation of prevailing CBM procedures for large-scale heterogeneous systems, however, is increasingly challenged by model intractability and computational cost stemming from the synergistic effect of information completeness and structure complexity. In this article, we innovatively devises a tractable CBM model for multicomponent continuously degrading systems under nonperfect inspection information, which is applicable to heterogeneous system structure and arbitrary hierarchical maintenance actions. The maintenance optimization problem of interest constitutes a continuous-state partially observable Markov-decision process applicable to heterogeneous system structures. A series of structure properties associated with systematic conditional reliability and accessibility of optimal solution are established, following which a multiagent reinforcement learning model governed by partial-independent parameter-sharing mechanism is employed to allow for solution search under continuous state–action space. A customized proximal policy algorithm is then leveraged to facilitate efficient agent training by diminishing the cure of dimension. Comparative experiments conducted on train wheel treads verify the superior model performance over cost control and computational efficiency improvement.

Traditional degradation-based reliability evaluation methods are typically based on rich data from a population of similar products, providing an average description of product performance. To capture individual characteristics for personalized maintenance, a dynamic reliability evaluation framework is proposed based on the individual monitoring data, which integrates a two-stage scheme and incorporates the physical model. The state-space model is first constructed based on Paris' Law to accurately describe bearing degradation, combining both physical mechanisms and secondary random factors. Then, an online stage division strategy based on an expanding time window is proposed, which implements change point detection and performs parameter estimation to serve as a priori information. Next, degradation state distributions and model parameters are adaptively estimated in the second stage using the extended Kalman filter, and the reliability is evaluated in real time based on the interval failure rate. Finally, to demonstrate the efficacy of the proposed framework, a comparative practical case study on bearing vibration data is presented.

Inspection errors are extensively reported in equipment health management due to multisource noises and technical limitations, particularly in hidden defect diagnosis of the multistage failure process. This article proposes a state-age-dependent maintenance and spare control strategy to compensate inspection-error-induced risk (attributed to both false positive and false negative) during defect identification. Specifically, a dual-phase adaptive inspection accommodating health variation is scheduled, following which both spare ordering and replacement are postponed to compensate implication of false-positive error. In addition, age-based replacement supported by preponed standard ordering is implemented promptly to alleviate false-negative error impact. To mitigate downtime losses, a dynamic selection mechanism upon failure occurrence between urgent and standard orderings is executed. The long-run operational cost rate is minimized by the joint optimization of postponed intervals of ordering and replacement, as well as the second-phase inspection interval. The model applicability is demonstrated through numerical experiments conducted on high-speed train bogie bearings.

With the advancements in sensor technology, it is now possible to measure and record a multitude of features that reflect the health condition of complex systems. These measurements are stored in a sizable data matrix, enabling the detection of anomalies. Nevertheless, the presence of this large data matrix poses a significant computational burden. The dimension-reduction methods, such as non-negative matrix factorization (NMF), can efficiently reduce computational burden. However, their pure data-driven nature can lead to overfitting and biases in anomaly detection results. To address this shortcoming, we propose a physics-enhanced NMF (PNMF) method by incorporating physical knowledge into NMF with the help of graph technique. The graph technique organizes measurements and features into two graph objects, respectively, and the physical knowledge guides the formation of edges between nodes in the graph. This allows the PNMF to capture not only the data-driven patterns but also the physical structure inherent in the system. The closed-form update algorithm is developed for the PNMF model, which can guarantee the convergence of parameters estimation. The superior performance of the PNMF model in detecting anomalies is demonstrated by comparing prevailing methods in both public datasets and real-world applications.

Supervisory control and data acquisition (SCADA) data from a complex mechanical system, such as a high-speed train power bogie, nonpower bogie, and wind turbine, are widely used for anomaly detection and fault isolation. The SCADA data include measurements of process variables and exogenous covariates for key components in the system. The process variables refer to the performance characteristics of the key component while the exogenous covariates are working loads or working conditions of the complex mechanical system. Dominated by such physical mechanisms as dynamic motion laws of the system, there are complex relationships between the process variables and covariates, that complicate anomaly detection and fault isolation. To solve this problem, we propose a framework that integrates physical knowledge and statistical learning. We first build a spline model to capture the relationship between process variables and exogenous covariates. To make the model interpretable, we use physical knowledge to impose constraints on the model parameters. We then conduct anomaly detection at a system level based on the physical-statistical regression model. Once an anomaly is detected, we propose a Lasso-based method to isolate the faulty components. Our fault isolation method does not require historical failure data or knowing the true number of faulty components. Real-world case studies on power bogies from high-speed trains illustrate the advantages of our framework: the best benchmark achieves at least 2.50% lower F1-score in anomaly detection and 6.01% lower F1-score in fault isolation compared to our method.

Condition-based maintenance (CBM), as a key component of asset health management, is crucial to enhance the operational safety and availability of diverse mechatronic systems, such as railway vehicles, wind power equipment, nuclear devices, etc. A common phenomenon observed in CBM is the existence of dispersibility regarding degradation-induced failure threshold, which affects the precision of maintenance decisions. This article addresses such challenges by scheduling a prognosis-centered intelligent CBM policy, which harnesses dynamic lifetime information to support both scheduled and opportunistic maintenance decision-making. The degradation is characterized by a generalized-form stochastic process, and the lifetime distribution is assessed through the fusion of multiple uncertainties. A dynamic reliability criterion is set to determine whether and when to postpone maintenance, whose interval is controlled by the remaining lifetime as well as an optimizable safety coefficient. The postponement interval, in turn, enables the planning of opportunistic maintenance to mitigate system downtime. The operational cost rate is minimized through the joint optimization of the inspection interval, conditional reliability threshold, and safety coefficient. The superiorities of the proposed policy over some conventional/heuristic maintenance policies are demonstrated by a case study on filed maintenance planning of high-speed train bearing.

Intelligent maintenance powered by advanced sensor technology is crucial to ensure the safe and reliable operation of wind turbines. Most maintenance models are scheduled solely based on age/degradation conditions while ignoring the dynamics of wind conditions and residual lifetime that significantly affect maintenance executions. This article addresses such challenges by constructing a dynamic age-state-dependent intelligent opportunistic maintenance framework that is capable of integrating 1) degradation and age state, 2) estimation of remaining lifetime, and 3) both the positive (extra maintenance opportunities) and negative impacts (maintenance delays) of wind conditions. Specially, component-level maintenance is allowed to be postponed to balance lifetime extension and resource allocation, whose implementation interval is controlled by real-time estimations of lifetime and dynamic wind velocities. Moreover, both wind- and health-centered opportunistic maintenance are incorporated to mitigate power generation losses. The applicability and superiority of the proposed framework are validated by a case study on an Ontario wind farm.

Load sharing is a common mechanism in redundant systems, which possesses significant impacts on operational safety. Although failure analyses of such systems are abundant, the risk mitigation methodology through elaborate maintenance scheduling is still insufficient. To address such deficiency, we innovatively designed a two-threshold group maintenance policy for$k$-out-of-$n$load-sharing systems. Such policy aims to save costs by mitigating the failure risks and minimizing the disturbance of maintenance activities to ensure continuous operation. Compared with existing studies, the proposed policy and modeling approach have two prominent superiorities. First, they are not limited to basic two-component systems which are addressed most. Second, arbitrary lifetime distribution is allowed in maintenance decision-making, which effectively expands the realistic application scope. To relieve the computation burden arising therefrom, a surrogate-aided approach is proposed to enhance the practicability for large-scale systems. We demonstrate the generality and superior performance of the approach through numerical experiments.

Remaining useful life (RUL) prediction under time-varying operating conditions is critical to the prognostics and health management of rotating machinery. In the literature, both the degradation rate and variation of a machinery component are often assumed to be solely dependent on operating conditions. However, this strong assumption is usually violated in many industrial applications. In this article, a systematic method for RUL prediction for a rotating machinery component is developed by considering the joint dependency of degradation rate and variation on time-varying operating conditions. In particular, a system state function and an observation function are utilized to characterize the component's degradation process. A quantitative relationship between the drift and diffusion parameters is established to reflect their joint dependency on the operating conditions. A two-stage hybrid approach that jointly implements maximum likelihood estimation and least squares estimation methods is proposed to facilitate parameter estimation in model development based on offline degradation data, and a Bayesian algorithm based on online condition monitoring data is utilized for RUL prediction in online implementation. A simulation study and a real application to rolling element bearings are provided to illustrate the effectiveness of the proposed method in practice.

This paper proposes a stress-strength time-varying correlation interference model for structural reliability analysis using Copulas. First, the stochastic stress is developed by incorporating the interaction between basic variables and time variable into the quadratic response surface method, and the stochastic strength is characterized by the linear or exponential degradation model. Second, the Copula selection method is given, and we propose time-varying stable (unstable) model for Kendall's tau to describe the time-varying correlation characteristic. Third, the structural reliability estimation method is developed, especially the method for the nondifferentiable Copulas can be used to calculate the probability in any area for any type of Copula when the marginal distribution is continuous. Finally, the lower confidence limit of structural reliability is given based on the survival coefficient. The comparison results of the high-temperature structural reliability estimation from different situations are illustrated in the simulation example to demonstrate the availability of the proposed model.