System Failure Data Research Articles

A grouping module assessment method for ocean engineering systems: Subsea tree system as a case

In the ocean engineering environment, the quality failure data of complex systems are difficult to obtain due to the high experiment cost. In addition, using a single model to analyze risk, reliability, availability, and maintainability is a big challenge. Based on the fault tree, Dynamic Bayes and Markov models, a state assessment method for ocean engineering systems with multiple maintenance modes is proposed in this paper. This method gives full play to the advantages of the three methods, and comprehensively analyzes the risk, reliability, availability and maintainability. This method uses fault tree and Markov model to pre-process fault data, and then inputs the pre-processed fault data into the multi-state degradation model based on dynamic Bayesian theory. Considering the maintenance strategies of no repair, perfect repair, imperfect repair and preventive repair, the model is iterated and adjusted until the model has processed all the event data and the updated model can best reflect the state of the system. The method is verified by taking the subsea tree of the subsea production system as an example. The obtained tree reliability index (mean time without failure) is basically consistent with the failure statistics of offshore and onshore reliability databases, which verifies the accuracy of the proposed method.

Degradation index construction and learning-based prognostics for stochastically deteriorating feedback control systems

Degradation-based prognostics is crucial for the health management of technological systems. In this work, we are interested in the degradation index construction and remaining useful life prognostics for stochastically deteriorating feedback control systems. The main challenges reside in the lack of knowledge about the system structure and the latent internal damage, as well as in the fault tolerance nature of feedback control systems. Our solution is to consider the whole system as a black-box, and use its easy-to-observe reference input/time response output to estimate the system transfer function. The associated H∞ norm, also called maximum energy gain, is defined as a system degradation index. Since the system fault tolerance does not allow to efficiently model the index evolution by common stochastic processes, traditional prognostics based on degradation processes are no longer applicable. To remedy, we propose to fit the system remaining useful life population to the versatile Birnbaum–Saunders distribution, and adopt a segmenting piecewise polynomials algorithm to learn the mapping between the distribution parameters and degradation index from degradation and failure data of similar systems. By this way, the remaining useful life distribution of deteriorating feedback control systems can be predicted in real-time given the system input/output at an inspection time. We numerically experiment our method on a stabilization loop control device driven by proportional–integral–differential controller in an inertial platform. Numerous sensitivity results under various configurations of system characteristics and training data corroborate the outperformance of proposed degradation index and the learning-based prognostics method.

Bayesian geoadditive water pipe failure forecasting model by optimizingthe updating period

Abstract Municipal water managers rely on pipe deterioration models to plan maintenance, repair, and replacement. Although efforts have been made to increase their accuracy, these models are subject to uncertainties in the predictions. In this paper, an optimization procedure of the Bayesian updating period of the parameters of an existing deterioration model is proposed to sequentially reduce the uncertainty in the prediction of the water pipe breakage rate variable. This latter is modeled using a structured geoadditive regression technique where covariates are allowed to have linear (e.g., categorical) and nonlinear (e.g., continuous) relationships with the response variable. Unknown and unobserved covariates are included in the model through a geospatial component that captures spatial auto-correlations and local heterogeneities. The optimization procedure searches through the time series data to identify the optimal updating period horizon that corresponds to the minimum error between the predicted coefficient of determination between predictions and observations using the unupdated and updated models. The process is repeated until the entire time series data is covered. The application of this approach to failure data of large Canadian urban water systems shows a significant reduction in the uncertainty of the parameters and increases the accuracy in the prediction of the output response variable.

A Two-Stage Data-Driven Spatiotemporal Analysis to Predict Failure Risk of Urban Sewer Systems Leveraging Machine Learning Algorithms.

Risk-informed asset management is key to maintaining optimal performance and efficiency of urban sewer systems. Although sewer system failures are spatiotemporal in nature, previous studies analyzed failure risk from a unidimensional aspect (either spatial or temporal), not accounting for bidimensional spatiotemporal complexities. This is owing to the insufficiency of good-quality data, which ultimately leads to under-/overestimation of failure risk. Here, we propose a generalized methodology/framework to facilitate a robust spatiotemporal analysis of urban sewer system failure risk, overcoming the intrinsic challenges of data imperfections-e.g., missing data, outliers, and imbalanced information. The framework includes a two-stage data-driven modeling technique that efficiently models the highly right-skewed sewer system failure data to predict the failure risk, leveraging a bidimensional space-time approach. We implemented our analysis for Bogotá, the capital city of Colombia. We train, test, and validate a battery of machine learning algorithms-logistic regression, decision trees, random forests, and XGBoost-and select the best model in terms of goodness-of-fit and predictive accuracy. Finally, we illustrate the applicability of the framework in planning/scheduling sewer system maintenance operations using state-of-the-art optimization techniques. Our proposed framework can help stakeholders to analyze the failure-risk models' performance under different discrimination thresholds, and provide managerial insights on the model's adequate spatial resolution and appropriateness of decentralized management for sewer systemmaintenance.

EM Algorithm for Estimating Reliability of Multi-Release Open Source Software Based on General Masked Data

Multi-release is critical for modern open source software product in order to satisfy more customer requirements. Masked data, a kind of missing data, is the system failure data when the exact cause of the failures might be unknown. That is, the cause of the system failures may be any one of the objects. However, due to the influence of the test strategy in real project, the cause of the system failures may be a subset of the system objects, not any one of the objects. In this paper, the mathematical description of general masked data is presented based on the traditional masked data. Furthermore, a novel multi-release open source software (OSS) reliability model based on general masked data is proposed. Different from traditional multi-release OSS reliability model, the proposed approach is based on additive model with general masked data other than change point model. And then, the maximum likelihood estimation (MLE) process of the model parameters is derived in detail, and expectation maximization (EM) algorithm is used to solve the extremely complicated problem of the log-likelihood function. Finally, two data sets from real open source software project are applied to the proposed approach, and the results show that the proposed reliability model is useful and powerful.

ELS algorithm for estimating open source software reliability with masked data considering both fault detection and correction processes

Masked data are the system failure data when the exact cause of the failures might be unknown. That is, the cause of the system failures may be any one of the components. Additionally, to incorporate more information and provide more accurate analysis, modeling software fault detection and correction processes have attracted widespread research attention recently. However, stochastic fault correction time and masked data brings more difficulties in parameter estimation. In this paper, a framework of open source software growth reliability model with masked data considering both fault detection and correction processes is proposed. Furthermore, a novel Expectation Least Squares (ELS) method, an EM-like (Expectation Maximization) algorithm, is used to solve the problem of parameter estimation, because of its mathematical convenience and computational efficiency. It is note that the ELS procedure is easy to use and useful for practical applications, and it just needs more relaxed hidden assumptions. Finally, three data sets from real open source software project are applied to the proposed framework, and the results show that the proposed reliability model is useful and powerful.

Availability analysis of safety-critical and control systems of NPP using stochastic modeling

Non-functional requirements are essentially important and play vital role for applications ranging from safety-critical systems (SCS) to simple gaming applications to ensure their quality. SCS demands not only for safe and reliable systems but systems those remain safe and available while under attacks. Availability analysis approaches include, but are not limited to cluster technique, Markov based chain models, Reliability Block Diagrams (RBD), Fault Tree Analysis (FTA) and Flow Network. The classical approaches fail to account for the comprehensive and accurate analysis of the diverse characteristics such as temporal behavior of systems, uncertainty in system behavior and failure data, functional dependencies among components and multiple failure modes for components or systems. This paper presents a novel approach for the availability analysis of a Digital Feed Water Control System (DFWCS) of nuclear power plant, which considers the maintenance and repair of the main-steam safety valves. The approach will be useful when no operational profile data is available for that. The system has been modeled using Stochastic Petri Net capturing all the system requirements along with the partial failures of its subsystems and common-cause failures and analyzed using TimeNet tool. The proposed methodology proves to be efficient and overcomes the limitations of the traditional approaches and the Markov model approach as it computes the state-transition probabilities, rather than assuming or qualitatively assessing it.

Deep Learning with Dynamically Weighted Loss Function for Sensor-Based Prognostics and Health Management.

Deep learning has been employed to prognostic and health management of automotive and aerospace with promising results. Literature in this area has revealed that most contributions regarding deep learning is largely focused on the model’s architecture. However, contributions regarding improvement of different aspects in deep learning, such as custom loss function for prognostic and health management are scarce. There is therefore an opportunity to improve upon the effectiveness of deep learning for the system’s prognostics and diagnostics without modifying the models’ architecture. To address this gap, the use of two different dynamically weighted loss functions, a newly proposed weighting mechanism and a focal loss function for prognostics and diagnostics task are investigated. A dynamically weighted loss function is expected to modify the learning process by augmenting the loss function with a weight value corresponding to the learning error of each data instance. The objective is to force deep learning models to focus on those instances where larger learning errors occur in order to improve their performance. The two loss functions used are evaluated using four popular deep learning architectures, namely, deep feedforward neural network, one-dimensional convolutional neural network, bidirectional gated recurrent unit and bidirectional long short-term memory on the commercial modular aero-propulsion system simulation data from NASA and air pressure system failure data for Scania trucks. Experimental results show that dynamically-weighted loss functions helps us achieve significant improvement for remaining useful life prediction and fault detection rate over non-weighted loss function predictions.

Performance enhancement of production systems using fuzzy based availability analysis and simulation method

Due to the complexity of today's engineering systems, accurate system performance analysis is important to obtain a more productive system design. This study introduces a new approach to obtain a new system design considering system availability to increase its productivity in a more consistent and logical manner. To obtain more productive system design, both fuzzy availability analysis and simulation are used together. The conventional simulation modelling can be used to analyse the system behaviour considering failure and repair data to predict system throughput. This approach solves the problems of scarce data for failures and repairs in the system. The fuzzy availability analysis is used to consider system failures and repairs based on fuzzy set theory. In other words, the integration of simulation results and fuzzy system availability allows for analysing system performance to improve system productivity with new system design even though system failure and repair data are scarce. This new approach has been applied to improve system productivity in a battery production company in Turkey. The analysis results show that this new approach is able to analyse system performance accurately and improve system productivity with a new system design. [Received: 26 November 2018; Accepted: 17 December 2019]

Prediction of restoration factor for various maintenance types of railway systems using analytical hierarchy process

Purpose For repairable systems (RSs), reliability estimation is generally performed using virtual age models. Virtual age models consider the effect of maintenance actions by reducing system age using restoration factor (RF). RF is generally estimated from system failure data using various statistical methods. However, RSs such as railway systems experience various types of maintenance actions at different times during their life cycle. To consider all these different types of actions, we need multiple RFs in the virtual age model. As failure data are limited, the estimation of so many parameters becomes a complex problem and it can lead to erroneous inferences. These RFs are representative of effects of maintenance activities on the system. Therefore, these can be predicted from the information about the maintenance actions performed on the system. The paper aims to discuss these issues. Design/methodology/approach The paper considers different types of maintenance actions to predict RF of the system. These maintenance actions involve the replacement of components at some level of assembly. Each component in an assembly has its own impact on assembly restoration. RF for assembly/systems can be obtained by aggregating effects of multiple component replacement using analytical hierarchy process . The RF values obtained for different types of maintenance actions are then used to calculate the virtual age of the system at different failure points. Using these virtual age failure points, suitable distribution is fitted and parameters are estimated. The distribution and parameters provide information about reliability of the system at any point of time. Findings This paper provides an easier approach that gives different RFs for different types of PM and CM. To calculate RFs, it considers the impact of maintenance actions performed as well as the impact of the component on which they are performed. It is simpler and gives more consistent results than other approaches, which estimate RF using different statistical methods. Originality/value This paper provides an alternative approach to predict RF parameters instead of estimating these parameters using statistical methods. Estimation of parameters using different statistical methods is complex in nature and gives erroneous and inconsistent results. The approach given in this paper is simpler and gives more reliable results. This approach can be useful in estimating parameters for RSs when failure data are limited.

Statistical evaluation of PV system performance and failure data among different climate zones

The research project INFINITY focuses on design improvements for PV system components operating in different climatic regions. This topic gained pertinence in recent years due to increased installation rates worldwide with the highest growth rates in non-moderate climates. As PV components are not optimized to the respective climate zone, the target of the project is to gather failure data from PV installations in various climate zones. A standardized data input was achieved by using an extended format of the IEA PVPS Task 13 failure survey, allowing to enter PV system and failure information in rich detail. The components' and failure structure were elaborated by combining input from reliability workshop proceedings, PVPS Task 13 and project partners' experience in PV reliability testing. These structures were mapped into a database to collect and prepare the input for further statistical analysis. Collected data was filtered, offering a tidy data set of 1048 samples from 340 sub-systems from 45 countries worldwide. After data filtering, a power loss evaluation, a cluster analysis and a trend analysis were performed. This paper presents results of comparing PV module failures in different climate zones starting with a range of PV systems installed in the 1970s, to recently installed systems.

Multiaxis Loading Device for Reliability Tests of Machine Tools

Reliability of machine tools is critical to high- efficiency manufacturing, and the failure data of feeding systems are requisite to analyze the reliability. This paper presents a multiaxis loading device (MLD) to simulate loading condition of an operating feeding system of a machine tool in a real machining process; this could cause breakdowns of the system and collect failure data. The MLD was developed based on a six-degree-of-freedom (6-DOF) parallel mechanism, and its prototype was presented. Its kinematics was analyzed, and dynamic equations were established based on the principle of virtual work. A fuzzy proportional-integral-derivative controller was designed to build an explicit force control system, whose proportional and integral gain can be adjusted according to the errors between reference and actual forces; and a force and a velocity feedforward were integrated to improve the performance of the control system. Loading experiments were conducted on an experimental platform to validate the loading capacity of the MLD; the spindle of the platform was set to be stationary and driven to perform feeding motions with different dimensions. Results of the experiment with the stationary spindle indicate that the MLD can simultaneously provide a 3-DOF force and a 3-DOF moment with acceptable errors. Experiments with moving spindle illustrate that the MLD accomplished tracking a spindle that is executing a 3-DOF trajectory-unknown feeding motion and adding 3-DOF forces on the feeding system through the spindle. The actual output multiaxis forces have little hysteresis and errors less than 10%. Compared with the long-term machining to load the feeding system in traditional reliability tests of machine tools, the proposed MLD provides an efficient method to simulate the loading condition of the system with reduced testing costs.

Unsupervised Fault Diagnosis of Journal Bearing Rotor System with Heterogeneous Data

In general, it is extremely difficult to obtain failure data of real systems in the field such as power plant rotor systems. To accommodate the dearth of field failure data, conventional approaches employ data collected from a testbed that emulates the normal and faulty conditions of the real systems. Nevertheless, it is obvious that approaches developed with failure data solely collected from a testbed may not be ideal to diagnose the real systems. To this end, this paper proposes a unsupervised fault diagnostic approach for journal bearing systems that incorporates heterogeneous data from the testbed and real field systems. To demonstrate the validity of the proposed approach, a case study is conducted with the RK4 rotor kit and the power plant journal bearing system. The combination of vibration image generation with deep learning helps us use data from systems with an identical working principle but different scales. We anticipate that, by incorporating the heterogeneous data, the proposed approach can diagnose the conditions of actual journal bearing systems in the field more accurately.

Wireless sensor network reliability modelling based on masked data

This paper studies the reliability modelling of wireless sensor networks WSNs with the masked data that are often observed in practice. The masked data are the system failure data when exact subsystems or components causing system failures cannot be identified. When the masked data are observed, however, it is difficult to estimate the WSN reliability since the failure processes of the subnets cannot be decomposed into simple subsystem processes. In this paper, an additive non-homogeneous poisson process NHPP model is proposed to describe the failure process of the WSN with subnets. The maximum likelihood estimation MLE procedure is developed to estimate the parameters in the proposed model. By applying the given procedure, the WSN reliability estimate can be relatively easy to obtain. A numerical example based on simulation data with random masking is also provided to illustrate the applicability of the methodology.

Analytical method for optimization of maintenance policy based on available system failure data

An analytical optimization method for preventive maintenance (PM) policy with minimal repair at failure, periodic maintenance, and replacement is proposed for systems with historical failure time data influenced by a current PM policy. The method includes a new imperfect PM model based on Weibull distribution and incorporates the current maintenance interval T0 and the optimal maintenance interval T to be found. The Weibull parameters are analytically estimated using maximum likelihood estimation. Based on this model, the optimal number of PM and the optimal maintenance interval for minimizing the expected cost over an infinite time horizon are also analytically determined. A number of examples are presented involving different failure time data and current maintenance intervals to analyze how the proposed analytical optimization method for periodic PM policy performances in response to changes in the distribution of the failure data and the current maintenance interval.

A Full Bayesian Approach for Masked Data in Step-Stress Accelerated Life Testing

Bayesian analysis of the series system failure data under step-stress accelerating life testing is proposed when the cause of failure may not have been identified but has only been narrowed down to a subset of all potential risks. A general Bayesian formulation is investigated for the log-location-scale distribution family that includes most commonly used parametric lifetime distributions. Reparameterization is introduced for estimating the lifetime under the use condition stress and other parameters directly. The posterior analysis is done by Markov chain Monte Carlo sampling. The methodology is illustrated through the Weibull distributions, and a numerical example.

A Model-Driven Approach for the Failure Data Analysis of Multiple Repairable Systems Without Information on Individual Sequences

This paper proposes a model-driven approach which is able to analyze the failure data of multiple repairable systems when no information on the individual sequences is available, thus overcoming the limitations of the previous models proposed in the literature. The proposed approach can analyze the failure data of systems subject to minimal, imperfect, or worse repairs; and is developed under a general form of the baseline failure intensity. Both failure- and time-truncation are considered, provided that the truncation time is the same for all the systems. Due to its physical interpretation, the proposed approach allows the reliability characteristics of the failure process to be estimated, and it allows the effect of the repairs on the system reliability to be checked via testing procedures. The proposed approach has been applied to a well-known real data set, which inspired the work in this paper, showing its flexibility and great potential.

Applying the generalized autoregressive conditional Heteroskedastic model to analyze and forecast the field failure data of repairable systems

Purpose – Analyzing and forecasting reliability is increasingly important for enterprises. An accurate product reliability forecasting model cannot only learn and track a product's reliability and operational performance, but can also offer useful information that allows managers to take follow‐up actions to improve the product's quality and cost. The Generalized Autoregressive Conditional Heteroskedastic (GARCH) model is already extensively used to analyze and forecast time series data. However, the GARCH model has not been used to analyze and forecast the failure data of repairable systems. Based on these concerns, this study proposes the GARCH model to analyze and forecast the field failure data of repairable systems.Design/methodology/approach – This paper proposes the GARCH model to analyze and forecast the field failure data of repairable systems. Empirical results from electronic systems designed and manufactured by suppliers of the Chrysler Corporation are presented and discussed.Findings – The pr...

야전운용제원을 활용한 RAM분석 사례: 해군함정을 중심으로

야전운용제원을 이용한 RAM분석은 무기체계 도입과 운용에 유용한 중요 분석이다. 하지만 해군함정의 경우 분석에 핵심적인 고장정보가 없어 분석이 제대로 수행되지 못하는 실정이다. 본 연구에서는 정비실적과 자재소비실적을 사용하여 고장을 추적하는 프로세스를 새롭게 제안하고, 제안한 방법으로 분석이 가능하다는 것을 입증하였다. 제안한 분석방법으로 함정 9척에 대해 분석한 결과, 실제 장비운용자들의 경험치와 상당히 부합하는 것으로 확인되어 그 타당성이 확보되었다. 국내 운용환경과 다양한 변수들의 영향이 반영된 연구결과는 기존의 RAM예측값 보정과 검증에 유용하게 활용될 수 있을 것이며, 자료가 충분히 누적된다면 한국 고유의 RAM분석DB로 활용이 가능할 것으로 예상한다. RAM analysis using field data is important and useful for acquisition and operation of weapon systems. But this kind of analysis didn't conducted due to the absence of weapon system's failure data in case of warship. This study suggest failure tracking process based on maintenance and repair parts consumption data and proves suggested method enables conducting RAM analysis. This method obtains validity through field operator's reviews. The results on 9 warships in this study significantly correspond with weapon system operator's experiences. Because the result of this study contains effects of various operational environments in field, it will useful for correction and verification on existing another weapon system's RAM estimates. If this kind of RAM data are sufficiently accumulated, it will be used to DB of korean unique RAM analysis.

Bayesian reliability when system and subsystem failure data are obtained in the same time period

Previously, Bayesian anomaly was reported for estimating reliability when subsystem failure data and system failure data were obtained from the same time period. As a result, a practical method for mitigating Bayesian anomaly was developed. In the first part of this paper, however, we show that the Bayesian anomaly can be avoided as long as the same failure information is incorporated in the model. In the second part of this paper, we consider a problem of estimating the Bayesian reliability when the failure count data on subsystems and systems are obtained from the same time period. We show that Bayesian anomaly does not exist when using the multinomial distribution with the Dirichlet prior distribution. A numerical example is given to compare the proposed method with the previous methods.