Showing papers in "IEEE Transactions on Reliability in 2012"
TL;DR: A non-linear model to estimate the remaining useful life of a system based on monitored degradation signals is presented and it is revealed that considering nonlinearity in the degradation process can significantly improve the accuracy of remaining usefulLife estimation.
Abstract: Remaining useful life estimation is central to the prognostics and health management of systems, particularly for safety-critical systems, and systems that are very expensive. We present a non-linear model to estimate the remaining useful life of a system based on monitored degradation signals. A diffusion process with a nonlinear drift coefficient with a constant threshold was transformed to a linear model with a variable threshold to characterize the dynamics and nonlinearity of the degradation process. This new diffusion process contrasts sharply with existing models that use a linear drift, and also with models that use a linear drift based on transformed data that were originally nonlinear. Both existing models are based on a constant threshold. To estimate the remaining useful life, an analytical approximation to the distribution of the first hitting time of the diffusion process crossing a threshold level is obtained in a closed form by a time-space transformation under a mild assumption. The unknown parameters in the established model are estimated using the maximum likelihood estimation approach, and goodness of fit measures are applied. The usefulness of the proposed model is demonstrated by several real-world examples. The results reveal that considering nonlinearity in the degradation process can significantly improve the accuracy of remaining useful life estimation.
423 citations
TL;DR: The results of the developed prognostics method, particularly the estimation of the RUL, can help improving the availability, reliability, and security while reducing the maintenance costs.
Abstract: This paper addresses a data-driven prognostics method for the estimation of the Remaining Useful Life (RUL) and the associated confidence value of bearings The proposed method is based on the utilization of the Wavelet Packet Decomposition (WPD) technique, and the Mixture of Gaussians Hidden Markov Models (MoG-HMM) The method relies on two phases: an off-line phase, and an on-line phase During the first phase, the raw data provided by the sensors are first processed to extract features in the form of WPD coefficients The extracted features are then fed to dedicated learning algorithms to estimate the parameters of a corresponding MoG-HMM, which best fits the degradation phenomenon The generated model is exploited during the second phase to continuously assess the current health state of the physical component, and to estimate its RUL value with the associated confidence The developed method is tested on benchmark data taken from the “NASA prognostics data repository” related to several experiments of failures on bearings done under different operating conditions Furthermore, the method is compared to traditional time-feature prognostics and simulation results are given at the end of the paper The results of the developed prognostics method, particularly the estimation of the RUL, can help improving the availability, reliability, and security while reducing the maintenance costs Indeed, the RUL and associated confidence value are relevant information which can be used to take appropriate maintenance and exploitation decisions In practice, this information may help the maintainers to prepare the necessary material and human resources before the occurrence of a failure Thus, the traditional maintenance policies involving corrective and preventive maintenance can be replaced by condition based maintenance
310 citations
TL;DR: A data-driven prognostics method, where the RUL of the physical system is assessed depending on its critical component, and the proposed method is applied to real data corresponding to the accelerated life of bearings, and experimental results are discussed.
Abstract: Prognostics activity deals with the estimation of the Remaining Useful Life (RUL) of physical systems based on their current health state and their future operating conditions. RUL estimation can be done by using two main approaches, namely model-based and data-driven approaches. The first approach is based on the utilization of physics of failure models of the degradation, while the second approach is based on the transformation of the data provided by the sensors into models that represent the behavior of the degradation. This paper deals with a data-driven prognostics method, where the RUL of the physical system is assessed depending on its critical component. Once the critical component is identified, and the appropriate sensors installed, the data provided by these sensors are exploited to model the degradation's behavior. For this purpose, Mixture of Gaussians Hidden Markov Models (MoG-HMMs), represented by Dynamic Bayesian Networks (DBNs), are used as a modeling tool. MoG-HMMs allow us to represent the evolution of the component's health condition by hidden states by using temporal or frequency features extracted from the raw signals provided by the sensors. The prognostics process is then done in two phases: a learning phase to generate the behavior model, and an exploitation phase to estimate the current health state and calculate the RUL. Furthermore, the performance of the proposed method is verified by implementing prognostics performance metrics, such as accuracy, precision, and prediction horizon. Finally, the proposed method is applied to real data corresponding to the accelerated life of bearings, and experimental results are discussed.
268 citations
TL;DR: The modified joint distribution bounds in terms of Kendall's tau and Spearman's rho provide an improvement to Frechet-Hoeffding bounds for estimating the possible system reliability range.
Abstract: We develop s-dependent competing risk model for systems subject to multiple degradation processes and random shocks using time-varying copulas. The proposed model allows for a more flexible dependence structure between risks in which (a) the dependent relationship between random shocks and degradation processes is modulated by a time-scaled covariate factor, and (b) the dependent relationship among various degradation processes is fitted using the copula method. Two types of random shocks are considered in the model: fatal shocks, which fails the system immediately; and nonfatal shocks, which does not. In a nonfatal shock situation there are two impacts towards the degradation processes: sudden increment jumps, and degradation rate accelerations. The comparison results of the system reliability estimation from both constant and time-varying copulas are illustrated in the numerical examples to demonstrate the application of the proposed model. The modified joint distribution bounds in terms of Kendall's tau and Spearman's rho provide an improvement to Frechet-Hoeffding bounds for estimating the possible system reliability range.
229 citations
TL;DR: An in-line coding algorithm which uses a stack-based implementation, and a recursive algorithm to pick out the equivalent full and half cycles of the irregular load profile to provide a more efficient cycle counting method using less memory storage, and making more efficient use of computational resources within the real-time environment.
Abstract: In many reliability design and model-based health management applications where load profiles are variable and unpredictable, it is desirable to have efficient cycle counting methods to identify equivalent full and half cycles within the irregular load profile. Conventional cycle-based lifetime models can then be applied directly to provide information about the life consumption of the products. The use of an off-line rainflow algorithm is a common solution for arbitrary loads, but it cannot be applied in real time in its original form. This paper presents an in-line coding algorithm which uses a stack-based implementation, and a recursive algorithm to pick out the equivalent full and half cycles of the irregular load profile. The method can be integrated easily within time-domain or serial data applications to generate equivalent full and half cycles as they occur. Thus it is of particular significance for life estimation in real-time applications where use of the traditional implementations of the counting algorithm is impractical. In comparison with the off-line traditional rainflow method, the on-line method doesn't require any knowledge of the time history of the load profile because it processes each minimum or maximum when it occurs. Therefore, it provides a more efficient cycle counting method using less memory storage, and making more efficient use of computational resources within the real-time environment.
204 citations
TL;DR: Reliability models are developed for systems subject to multiple s-dependent competing failure processes with a changing, dependent failure threshold based on degradation and random shock modeling.
Abstract: We present reliability and maintenance models for systems subject to multiple s-dependent competing failure processes with a changing, dependent failure threshold. In our model, two failure processes are considered: soft failure caused by continuous degradation together with additional abrupt degradation due to a shock process, and hard failure caused by the instantaneous stress from the same shock process. These two failure processes are correlated or s-dependent in two respects: 1) the arrival of each shock load affects both failure processes, and 2) the shock process impacts the hard failure threshold level. In previous research, the failure thresholds are fixed constants, which is appropriate for most design and reliability problems. However, the nature of the failure threshold has become a critical issue for certain classes of complex devices. When withstanding shocks, the system is deteriorating, and its resistance to failure is weakening. In this case, it becomes more sensitive to hard failure. In this paper, three cases of dependency between the shock process and the hard failure threshold level are studied. The first case is that the hard failure threshold value changes to a lower level when the first shock is recorded above a critical value, or a generalized extreme shock model. The second case is that the hard failure threshold value decreases to a lower level when the time lag between two sequential shocks is less than a threshold δ, or a generalized δ-shock model. The third case is that the hard failure threshold value reduces to a lower level right after m shocks whose magnitudes are larger than a critical value, or a generalized m-shock model. Based on degradation and random shock modeling, reliability models are developed for these two s-dependent failure processes with a shifting failure threshold. Two preventive maintenance policies are also applied and compared to decide which one is more beneficial. Then a Micro-Electro-Mechanical System example is given to demonstrate the reliability models and maintenance polices.
159 citations
TL;DR: The benefits of prognostics in terms of system life-cycle processes, such as design and development, production, operations, logistics support, and maintenance, are discussed.
Abstract: Prognostics is an engineering discipline utilizing in-situ monitoring and analysis to assess system degradation trends, and determine remaining useful life. This paper discusses the benefits of prognostics in terms of system life-cycle processes, such as design and development, production, operations, logistics support, and maintenance. Challenges for prognostics technologies from the viewpoint of both system designers and users will be addressed. These challenges include implementing optimum sensor systems and settings, selecting applicable prognostics methods, addressing prognostic uncertainties, and estimating the cost-benefit implications of prognostics implementation. The research opportunities are summarized as well.
149 citations
TL;DR: The results clearly show that the proposed technique is more effective than several other popular, state of the art fault localization techniques, and can easily be applied to programs with multiple bugs as well.
Abstract: We propose the application of a modified radial basis function neural network in the context of software fault localization, to assist programmers in locating bugs effectively This neural network is trained to learn the relationship between the statement coverage information of a test case and its corresponding execution result, success or failure The trained network is then given as input a set of virtual test cases, each covering a single statement The output of the network, for each virtual test case, is considered to be the suspiciousness of the corresponding covered statement A statement with a higher suspiciousness has a higher likelihood of containing a bug, and thus statements can be ranked in descending order of their suspiciousness The ranking can then be examined one by one, starting from the top, until a bug is located Case studies on 15 different programs were conducted, and the results clearly show that our proposed technique is more effective than several other popular, state of the art fault localization techniques Further studies investigate the robustness of the proposed technique, and illustrate how it can easily be applied to programs with multiple bugs as well
144 citations
TL;DR: The problem of optimal design for degradation tests based on a gamma degradation process with random effects, and the effects of model mis-specification that occur when the random effects are not taken into consideration in the gamma degradation model are discussed.
Abstract: Degradation models are usually used to provide information about the reliability of highly reliable products that are not likely to fail within a reasonable period of time under traditional life tests, or even accelerated life tests. The gamma process is a natural model for describing degradation paths, which exhibit a monotone increasing pattern, while the commonly used Wiener process is not appropriate in such a case. We discuss the problem of optimal design for degradation tests based on a gamma degradation process with random effects. To conduct a degradation experiment efficiently, several decision variables (such as the sample size, inspection frequency, and measurement numbers) need to be determined carefully. These decision variables affect not only the experimental cost, but also the precision of the estimates of lifetime parameters of interest. Under the constraint that the total experimental cost does not exceed a pre-specified budget, the optimal decision variables are found by minimizing the asymptotic variance of the estimate of the 100p-th percentile of the lifetime distribution of the product. Laser data are used to illustrate the proposed method. Moreover, we assess analytically the effects of model mis-specification that occur when the random effects are not taken into consideration in the gamma degradation model. The numerical results of these effects reveal that the impact of model mis-specification on the accuracy and precision of the prediction of percentiles of the lifetimes of products are somewhat serious for the tail probabilities. A simulation study also shows that the simulated values are quite close to the asymptotic values.
142 citations
TL;DR: This paper studies importance measures in reliability, including their definitions, probabilistic interpretations, properties, computations, and comparability, and categorizes importance measures into the structure, reliability, and lifetime types based on the knowledge for determining them.
Abstract: Many importance measures have been proposed with respect to the diverse considerations of system performance, reflecting different probabilistic interpretations and potential applications. This paper studies importance measures in reliability, including their definitions, probabilistic interpretations, properties, computations, and comparability. It categorizes importance measures into the structure, reliability, and lifetime types based on the knowledge for determining them. It covers importance measures of individual components, and ones of pairs and groups of components. It also investigates importance measures in consecutive-k-out-of-n systems.
135 citations
TL;DR: This paper presents an overview of reliability testing, reliability estimation, and prediction models and approaches for the design of test plans which result in providing failure data and/or degradation data in a limited test duration.
Abstract: This paper presents an overview of reliability testing, reliability estimation, and prediction models. It also presents approaches for the design of test plans which result in providing failure data and/or degradation data in a limited test duration. Equivalence of different test plans which result in similar reliability estimates is also discussed. Use of degradation data for reliability estimates and maintenance decisions are presented.
TL;DR: An optimal release planning problem is formulated which minimizes the cost of testing of the release that is to be brought into market under the constraint of removing a desired proportion of faults from the current release.
Abstract: Long-lived software systems evolve through new product releases, which involve up-gradation of previous released versions of the software in the market. But, upgrades in software lead to an increase in the fault content. Thus, for modeling the reliability growth of software with multiple releases, we must consider the failures of the upcoming upgraded release, and the failures that were not debugged in the previous release. Based on this idea, this paper proposes a mathematical modeling framework for multiple releases of software products. The proposed model takes into consideration the combined effect of schedule pressure and resource limitations using a Cobb Douglas production function in modeling the failure process using a software reliability growth model. The model developed is validated on a four release failure data set. Another major concern for the software development firms is to plan the release of the upgraded version. When different versions of the software are to be released, then the firm plans the release on the basis of testing progress of the new code, as well as the bugs reported during the operational phase of the previous version. In this paper, we formulate an optimal release planning problem which minimizes the cost of testing of the release that is to be brought into market under the constraint of removing a desired proportion of faults from the current release. The problem is illustrated using a numerical example, and is solved using a genetic algorithm.
TL;DR: It is shown that the system reliability can be evaluated using robust algorithms within O(n-k+1) computational time and that the reliability of the warm standby system can be calculated using well-established numerical procedures that are available for the beta distribution.
Abstract: We study reliability characteristics of the k-out-of- n warm standby system with identical components subject to exponential lifetime distributions. We derive state probabilities of the warm standby system in a form that is similar to the state probabilities of the active redundancy system. Subsequently, the system reliability is expressed in several forms that can provide new insights into the system reliability characteristics. We also show that all properties and computational procedures that are applicable for active redundancy are also applicable for the warm standby redundancy. As a result, it is shown that the system reliability can be evaluated using robust algorithms within O(n-k+1) computational time. In addition, we provide closed-form expressions for the hazard rate, probability density function, and mean residual life function. We show that the time-to-failure distribution of the k-out-of-n warm standby system is equal to the beta exponential distribution. Subsequently, we derive closed-form expressions for the higher order moments of the system failure time. Further, we show that the reliability of the warm standby system can be calculated using well-established numerical procedures that are available for the beta distribution. We prove that the improvement in system reliability with an additional redundant component follows a negative binomial (Polya) distribution, and it is log-concave in n. Similarly, we prove that the system reliability function is log-concave in n. Because the k -out-of-n system with active redundancy can be considered as a special case of the k-out-of-n warm standby system, we indirectly provide some new results for the active redundancy case as well.
TL;DR: A prognostic method which predicts the Remaining Useful Life (RUL) of a degrading system by means of an ensemble of empirical models is proposed.
Abstract: The safety of nuclear power plants can be enhanced, and the costs of operation and maintenance reduced, by means of prognostic and health management systems which enable detecting, diagnosing, predicting, and proactively managing the equipment degradation toward failure. We propose a prognostic method which predicts the Remaining Useful Life (RUL) of a degrading system by means of an ensemble of empirical models. The RUL predictions of the individual models are aggregated through a Kalman Filter (KF)-based algorithm. The method is applied to the prediction of the RUL of turbine blades affected by a developing creep.
TL;DR: In this paper, the authors present a new concept in failure diagnosis, the modeling of intermittent failure dynamics, which is relevant for productive processes which sustain a high level of maintenance caused by failures.
Abstract: Diagnosis of intermittent failures is relevant for productive processes which sustain a high level of maintenance caused by failures. Systems including electrical contacts suffer a considerable amount of intermittent failures which are usually not diagnosed. A characterization of the failure dynamics may save unnecessary repairs, and enable a better planning component substitution. This paper presents a new concept in failure diagnosis, the modeling of intermittent failure dynamics. It develops two methodologies for modeling intermittent failure dynamics, and generating information for maintenance scheduling.
TL;DR: It is observed that the proposed tests are quite powerful when compared to an existing goodness-of-fit test proposed for progressively Type-II censored data due to Balakrishnan et al.
Abstract: We propose a general purpose approximate goodness-of-fit test that covers several families of distributions under progressive Type-II censored data. The test procedure is based on the empirical distribution function (EDF), and generalizes the goodness-of-fit test proposed by Chen and Balakrishnan [11] to progressively Type-II censored data. The new method requires some tables for critical values, which are constructed by Monte Carlo simulation. The power of the proposed tests are then assessed for several alternative distributions, while testing for normal, Gumbel, and log-normal distributions, through Monte Carlo simulations. It is observed that the proposed tests are quite powerful when compared to an existing goodness-of-fit test proposed for progressively Type-II censored data due to Balakrishnan et al. . The proposed goodness-of-fit test is then illustrated with two real data sets.
TL;DR: The results show that IIM can be used to identify the key state of a component that affects the system performance most and the IIM of component states concerns not only the probability distributions and transition intensities of the states of the object component, but also the change in theSystem performance under the change of the state distribution of theobject component.
Abstract: This paper mainly focuses on the integrated importance measure (IIM) of component states based on loss of system performance. To describe the impact of each component state, we first introduce the performance function of the multi-state system. Then, we present the definition of IIM of component states. We demonstrate its corresponding physical meaning, and then analyze the relationships between IIM and Griffith importance, Wu importance, and Natvig importance. Secondly, we present the evaluation method of IIM for multi-state systems. Thirdly, the characteristics of IIM of component states are discussed. Finally, we demonstrate a numerical example, and an application to an offshore oil and gas production system for IIM to verify the proposed method. The results show that 1) the IIM of component states concerns not only the probability distributions and transition intensities of the states of the object component, but also the change in the system performance under the change of the state distribution of the object component; and 2) IIM can be used to identify the key state of a component that affects the system performance most.
TL;DR: The extensions of importance measures in reliability, focusing on their interrelations to the B-importance, the dominant relations among them, the dual relations, their performances in typical systems, and their computations are investigated.
Abstract: To identify the critical components or sets of components in a system, various importance measures have been proposed with different probabilistic perspectives and applications. Many of these importance measures are actually related to each other in some ways. This paper summarizes the importance measures in reliability, and presents relations and comparisons among them, focusing on their interrelations to the B-importance, the dominant relations among them, the dual relations, their performances in typical systems, and their computations. The early versions of the importance measures are for binary coherent systems, while the recent research is not limited to this type of system. This paper investigates the extensions of importance measures in noncoherent systems, multistate systems, continuum systems, and repairable systems.
TL;DR: The paper proposes a general methodology of how to perform rolling element bearing prognostics, and presents the results using a robust regression curve fitting approach.
Abstract: The enhanced ability to predict the remaining useful life of helicopter drive train components offers potential improvement in the safety, maintainability, and reliability of a helicopter fleet. Current existing helicopter health and usage monitoring systems provide diagnostic information that indicates when the condition of a drive train component is degraded; however, prediction techniques are not currently used. Although various algorithms exist for providing remaining life predictions, prognostic techniques have not fully matured. This particular study addresses remaining useful life predictions for the helicopter oil-cooler bearings. The paper proposes a general methodology of how to perform rolling element bearing prognostics, and presents the results using a robust regression curve fitting approach. The proposed methodology includes a series of processing steps prior to the prediction routine, including feature extraction, feature selection, and health assessment. This approach provides a framework for including prediction algorithms into existing health and usage monitoring systems. A case study with the data collected by Impact Technology, LLC. is analysed using the proposed methodology. Future work would consider using the same methodology, but comparing the accuracy of this prediction method with Bayesian filtering techniques, usage based methods, and other time series prediction methods.
TL;DR: Two new maintenance policies using prognostic information are introduced based on a degradation and measurement model of crack growth propagation, and their maintenance cost models are evaluated jointly by analytical and simulation approaches and compared with two more classical benchmark models.
Abstract: This paper deals with maintenance decision-making for single-unit deteriorating systems operating under indirect condition monitoring. Based on a degradation and measurement model of crack growth propagation, two new maintenance policies using prognostic information are introduced. Their maintenance cost models are evaluated jointly by analytical and simulation approaches, and are compared with two more classical benchmark models. Such complete models integrating degradation phenomenon, monitoring characteristics, state estimation, prognostics, and maintenance assessment can give rise to fruitful numerical analyses and discussions. The main contributions of the paper are to i) analyze jointly the condition-based and dynamic structure of the considered maintenance policies; ii) propose some effective methods to reduce the effect of measurement uncertainty in condition-based maintenance decision-making; and iii) show the relevance of quantification methods when deciding to resort to prognostic approaches, and to invest in condition monitoring devices.
TL;DR: Bayesian methods for ADDT planning under a class of nonlinear degradation models with one accelerating variable using a Bayesian criterion based on the estimation precision of a specified failure-time distribution quantile at use conditions to find optimum test plans are described.
Abstract: Accelerated Destructive Degradation Tests (ADDTs) provide timely product reliability information in practical applications. This paper describes Bayesian methods for ADDT planning under a class of nonlinear degradation models with one accelerating variable. We use a Bayesian criterion based on the estimation precision of a specified failure-time distribution quantile at use conditions to find optimum test plans. A large-sample approximation for the posterior distribution provides a useful simplification to the planning criterion. The general equivalence theorem (GET) is used to verify the global optimality of the numerically optimized test plans. Optimum plans usually provide insight for constructing compromise plans which tend to be more robust, and practically useful. We present a numerical example with a log-location-scale distribution to illustrate the Bayesian test planning methods, and to investigate the effects of the prior distribution and sample size on test planning results.
TL;DR: This article proposes a novel approach to search for all MP in a general flow network using Minimal Paths or Minimal Cuts of the network to evaluate the reliability of a complex network.
Abstract: An active research field is the evaluation of the reliability of a complex network. The most popular methods for such evaluation often use Minimal Paths (MP) or Minimal Cuts (MC) of the network. Although there are many algorithms developed to search for MP or MC, most of them are inefficient for searching a large network due to the combinatorial explosion problem. Another disadvantage is that the existing algorithms are applicable to specific counts of source and sink nodes (e.g., one-to-one, one-to-many, and so on). This article proposes a novel approach to search for all MP in a general flow network. The term “general” means that the approach can be used to search for all MP with multi-sources, multi-sinks in the network. The edges can be directed, undirected, or hybrid (mixed with directed and undirected arcs). Some benchmarks from the well-known algorithms in the literature are examined and compared. Moreover, the comprehensive tests are also performed with the grid networks, as well as the well-known networks in the literature to show the efficiency of the approach. A sample code is provided in the article for quick validation.
TL;DR: Two new importance measures are proposed as functions of time that can provide timely feedback on the critical components prior to failure based on the measured or observed degradation of degrading components.
Abstract: This paper proposes two new importance measures: one new importance measure for systems with -independent degrading components, and another one for systems with -correlated degrading components. Importance measures in previous research are inadequate for systems with degrading components because they are only applicable to steady-state cases and problems with discrete states without considering the continuously changing status of the degrading components. Our new importance measures are proposed as functions of time that can provide timely feedback on the critical components prior to failure based on the measured or observed degradation. Furthermore, the correlation between components is considered for developing these importance measures through a multivariate distribution. To evaluate the criticality of components, we analysed reliability models for multi-component systems with degrading components, which can also be utilized for studying maintenance models. Numerical examples show that the proposed importance measures can be used as an effective tool to assess component criticality for systems with degrading components.
TL;DR: This analysis is based on Markov reward models, and suggests that host failure rate is the most important parameter when the measure of interest is the system mean time to failure.
Abstract: Server virtualization is a technology used in many enterprise systems to reduce operation and acquisition costs, and increase the availability of their critical services. Virtualized systems may be even more complex than traditional nonvirtualized systems; thus, the quantitative assessment of system availability is even more difficult. In this paper, we propose a sensitivity analysis approach to find the parameters that deserve more attention for improving the availability of systems. Our analysis is based on Markov reward models, and suggests that host failure rate is the most important parameter when the measure of interest is the system mean time to failure. For capacity oriented availability, the failure rate of applications was found to be another major concern. The results of both analyses were cross-validated by varying each parameter in isolation, and checking the corresponding change in the measure of interest. A cost-based optimization method helps to highlight the parameter that should have higher priority in system enhancement.
TL;DR: This paper proposes some useful sufficient conditions to determine strong diagnosability, and the conditional diagnosable of a system, and applies them to show that an n-dimensional augmented cube AQn is strongly (2n -1)-diagnosable for n ≥ 5, andThe result demonstrates that the conditional Diagnosability of AQ n is about three times larger than the classical diagnosis.
Abstract: The problem of fault diagnosis has been discussed widely, and the diagnosability of many well-known networks has been explored. Strong diagnosability, and conditional diagnosability are both novel measurements for evaluating reliability and fault tolerance of a system. In this paper, some useful sufficient conditions are proposed to determine strong diagnosability, and the conditional diagnosability of a system. We then apply them to show that an n-dimensional augmented cube AQn is strongly (2n -1)-diagnosable for n ≥ 5, and the conditional diagnosability of AQn is 6n - 17 for n ≥ 6. Our result demonstrates that the conditional diagnosability of AQn is about three times larger than the classical diagnosability.
TL;DR: A technical framework and roadmap of embedded diagnostics and prognostics (ED/EP) for complex mechanical systems is presented based on the methodology of system integration and parallel design, which includes six key elements (embedded sensors, embedded sensing design, embedded sensors placement, embedded signals transmission, ED/EP algorithms, and embedded self-power).
Abstract: Prognostics and Health Management (PHM) technologies have emerged as a key enabler to provide early indications of system faults and perform predictive maintenance actions. Implementation of a PHM system depends on accurately acquiring in real time the present and estimated future health states of a system. For electronic systems, built-in-test (BIT) makes it not difficult to achieve these goals. However, reliable prognostics capability is still a bottle-neck problem for mechanical systems due to a lack of proper on-line sensors. Recent advancements in sensors and micro- electronics technologies have brought about a novel way out for complex mechanical systems, which is called embedded diagnostics and prognostics (ED/EP). ED/EP can provide real-time present condition information and future health states by integrating micro-sensors into mechanical structures when designing and manufacturing, so ED/EP has a revolutionary progress compared to traditional mechanical fault diagnostic and prognostic ways. But how to study ED/EP for complex mechanical systems has not been focused so far. This paper explores the challenges and needs of efforts to implement ED/EP technologies. In particular, this paper presents a technical framework and roadmap of ED/EP for complex mechanical systems. The framework is based on the methodology of system integration and parallel design, which includes six key elements (embedded sensors, embedded sensing design, embedded sensors placement, embedded signals transmission, ED/EP algorithms, and embedded self-power). Relationships among these key elements are outlined, and they should be considered simultaneously when designing a complex mechanical system. Technical challenges of each key element are emphasized, and the corresponding existed or potential solutions are summarized in detail. Then a suggested roadmap of ED/EP for complex mechanical systems is brought forward according to potential advancements in related areas, which can be divided into three different stages: embedded diagnostics, embedded prognostics, and system integration. In the end, the presented framework is exemplified with a gearbox.
TL;DR: A decision support model based on options theory, a financial derivative tool extended to real assets, to valuate maintenance decisions after a remaining useful life prediction is developed and demonstrated.
Abstract: Safety, mission, and infrastructure critical systems are adopting prognostics and health management, a discipline consisting of technologies and methods to assess the reliability of a product in its actual life-cycle conditions to determine the advent of failure and mitigate system risks. The output from a prognostic system is the remaining useful life of the system; it gives the decision-maker lead-time and flexibility to manage the health of the system. This paper develops a decision support model based on options theory, a financial derivative tool extended to real assets, to valuate maintenance decisions after a remaining useful life prediction. We introduce maintenance options, and develop a hybrid methodology based on Monte Carlo simulations and decision trees for a cost-benefit-risk analysis of prognostics and health management. We extend the model, and combine it with least squares Monte Carlo methods to valuate one type of maintenance options, the waiting options; their value represents the cost avoidance opportunities and revenue obtained from running the system through its remaining useful life. The methodologies in this paper address the fundamental objective of system maintenance with prognostics: to maximize the use of the remaining useful life while concurrently minimizing the risk of failure. We demonstrate the methodologies on decision support for sustaining wind turbines by showing the value of having a prognostics system for gearboxes, and determining the value of waiting to perform maintenance. The value of the waiting option indicates that having the system available throughout the predicted remaining useful life is more beneficial than having downtime for maintenance, even if there is a high risk of failure.
TL;DR: The aim of this paper is to formalize and discuss the connexionist-systems-based approaches to ensure multi-step ahead predictions for prognostics and pointed out five approaches: the Iterative, Direct, DirRec, Parallel, and MISMO approaches.
Abstract: Prognostics and Health Management aims at estimating the remaining useful life of a system (RUL) , i.e. the remaining time before a failure occurs. It benefits thereby from an increasing interest: prognostic estimates (and related decision-making processes) enable increasing availability and safety of industrial equipment while reducing costs. However, prognostics is generally based on a prediction step which, in the context of data-driven approaches as considered in this paper, can be hard to achieve because future outcomes are in essence difficult to estimate. Also, a prognostic system must perform sufficient long term estimates, whereas many works focus on short term predictions. Following that, the aim of this paper is to formalize and discuss the connexionist-systems-based approaches to ensure multi-step ahead predictions for prognostics. Five approaches are pointed out: the Iterative, Direct, DirRec, Parallel, and MISMO approaches. Conclusions of the paper are based, on one side, on a literature review; and on the other side, on simulations among 111 time series prediction problems, and among a real engine fault prognostics application. These experiments are performed using the exTS (evolving extended Takagi-Sugeno system). As for comparison purpose, three types of performances measures are used: prediction accuracy, complexity (computational time), and implementation requirements. Results show that all three criteria are never optimized at the same time (same experiment), and best practices for prognostics application are finally pointed out.
TL;DR: An approach to conducting a goodness-of-flt test is proposed, a Bayesian approach to selecting the most adequate model among several competitive candidates is introduced, and a framework that incorporates the model selection results into the preventive maintenance decision making is developed.
Abstract: Many imperfect maintenance models have been developed to mathematically characterize the efficiency of maintenance activity from various points of view. However, the adequacy of an imperfect maintenance model must be validated before it is used in decision making. The most adequate imperfect maintenance model among the candidates to facilitate decision making is also desired. The contributions of this paper lie in three aspects: 1 it proposes an approach to conducting a goodness-of-flt test, 2 it introduces a Bayesian approach to selecting the most adequate model among several competitive candidates, and 3 it develops a framework that incorporates the model selection results into the preventive maintenance decision making. The effectiveness of the proposed methods is demonstrated by three designed numerical studies. The case studies show that the proposed methods are able to identify the most adequate model from the competitive candidates, and incorporating the model selection results into the maintenance decision model achieves better estimation for applications with limited data.
TL;DR: This article investigates two major stratification alternatives for software defect prediction using Analysis of Variance, and finds that the main effect of under-sampling is significant at α = 0.05, as is the interaction between under- and over-sampled.
Abstract: Numerous studies have applied machine learning to the software defect prediction problem, i.e. predicting which modules will experience a failure during operation based on software metrics. However, skewness in defect-prediction datasets can mean that the resulting classifiers often predict the faulty (minority) class less accurately. This problem is well known in machine learning, and is often referred to as “learning from imbalanced datasets.” One common approach for mitigating skewness is to use stratification to homogenize class distributions; however, it is unclear what stratification techniques are most effective, both generally and specifically in software defect prediction. In this article, we investigate two major stratification alternatives (under-, and over-sampling) for software defect prediction using Analysis of Variance. Our analysis covers several modern software defect prediction datasets using a factorial design. We find that the main effect of under-sampling is significant at α = 0.05, as is the interaction between under- and over-sampling. However, the main effect of over-sampling is not significant.