The accuracy of traditional anomaly detection techniques implemented on full-dimensional spaces degrades significantly as dimensionality increases, thereby hampering many real-world applications. This work proposes an approach to selecting meaningful feature subspace and conducting anomaly detection in the corresponding subspace projection. The aim is to maintain the detection accuracy in high-dimensional circumstances. The suggested approach assesses the angle between all pairs of two lines for one specific anomaly candidate: the first line is connected by the relevant data point and the center of its adjacent points; the other line is one of the axis-parallel lines. Those dimensions which have a relatively small angle with the first line are then chosen to constitute the axis-parallel subspace for the candidate. Next, a normalized Mahalanobis distance is introduced to measure the local outlier-ness of an object in the subspace projection. To comprehensively compare the proposed algorithm with several existing anomaly detection techniques, we constructed artificial datasets with various high-dimensional settings and found the algorithm displayed superior accuracy. A further experiment on an industrial dataset demonstrated the applicability of the proposed algorithm in fault detection tasks and highlighted another of its merits, namely, to provide preliminary interpretation of abnormality through feature ordering in relevant subspaces.

A reliability study based on a Bayesian semi-parametric framework is performed in order to explore the impact of the position of a locomotive wheel on its service lifetime and to predict its other reliability characteristics. A piecewise constant hazard regression model is used to analyse the lifetime of locomotive wheels using degradation data and taking into account the bogie on which the wheel is located. Gamma frailties are included in this study to explore unobserved covariates within the same group. The goal is to flexibly determine reliability for the wheel. A case study is performed using Markov chain Monte Carlo methods and the following conclusions are drawn. First, a polynomial degradation path is a better choice for the studied locomotive wheels; second, under given operational conditions, the position of the locomotive wheel, i.e. on which bogie it is mounted, can influence its reliability; third, a piecewise constant hazard regression model can be used to undertake reliability studies; fourth, considering gamma frailties is useful for exploring the influence of unobserved covariates; and fifth, the wheels have a higher failure risk after running a threshold distance, a finding which could be applied in optimisation of maintenance activities

In this paper, the wheel-sets’ running surface wear data based on re-profiling measurement from 16 bogies of heavy haul locomotives at Malmbanan (Sweden) are studied. The case study undertakes: reliability and degradation analysis, wear rate analysis and their comparison (including total wear rate, natural wear rate, re-profiling wear rate, the ratio of re-profiling and natural wear). The results show that: 1) for the studied group, a linear degradation path is more suitable; 2) following the linear degradation, the best life distribution is a 3-parameter Weibull distribution; 3) comparing the wearing data of the wheel-sets’ running surfaces is an effective way to optimize maintenance strategies; 4) more natural wear occurs for the wheels installed in axle 1 and axle 3, supportive evidence for other related studies at Malmbanan.

This paper undertakes a general reliability study using both classical and Bayesian semi-parametric degradation approaches. The goal is to illustrate how degradation data can be modelled and analysed to flexibly determine reliability to support preventive maintenance strategy making, based on a general data-driven framework. With the proposed classical approach, both Accelerated Life Tests (ALT) and Design of Experiments (DOE) technology are used to determine how each critical factor affects the prediction of performance. With the Bayesian semi-parametric approach, a piecewise constant hazard regression model is used to establish the lifetime using degradation data. Gamma frailties are included to explore the influence of unobserved covariates within the same group. Ideally, results from the classical and Bayesian approaches will complement each other. To demonstrate these approaches, this paper considers a case study of locomotive wheel-set reliability. The degradation data are prepared by considering an Exponential and a Power degradation path separately. The results show that both classical and Bayesian semi-parametric approaches are useful tools to analyse degradation data and can, therefore, support a company in decision making for preventive maintenance. The approach can be applied to other technical problems (e.g. other industries, other components).

The recent proliferation of Markov chain Monte Carlo (MCMC) approaches has led to the use of the Bayesian inference in a wide variety of fields. To facilitate MCMC applications, this paper proposes an integrated procedure for Bayesian inference using MCMC methods, from a reliability perspective. The goal is to build a framework for related academic research and engineering applications to implement modern computational-based Bayesian approaches, especially for reliability inferences. The procedure developed here is a continuous improvement process with four stages (Plan, Do, Study, and Action) and 11 steps, including: (1) data preparation; (2) prior inspection and integration; (3) prior selection; (4) model selection; (5) posterior sampling; (6) MCMC convergence diagnostic; (7) Monte Carlo error diagnostic; (8) model improvement; (9) model comparison; (10) inference making; (11) data updating and inference improvement. The paper illustrates the proposed procedure using a case study.

This paper undertakes a reliability study using both classical and Bayesian semi-parametric frameworks to explore the impact of a locomotive wheel's position on its service lifetime and to predict its other reliability characteristics. The goal is to illustrate how degradation data can be modeled and analyzed by using classical and Bayesian approaches. The adopted data in the case study have been collected from the Swedish company. The results show that: 1) an exponential degradation path is a better choice for the studied locomotive wheels; 2) both classical and Bayesian semi-parametric approaches are useful tools to analysis degradation data; 3) under given operation conditions, the position of the locomotive wheel could influence its reliability

The service life of railway wheels can differ significantly depending on their installed position, operating conditions, re-profiling characteristics, etc. This paper compares the wheels on two selected locomotives on the Iron Ore Line in northern Sweden to explore some of these differences. It proposes integrating reliability assessment data with both degradation data and re-profiling performance data. The following conclusions are drawn. First, by considering an exponential degradation path and given operation condition, the Weibull frailty model can be used to undertake reliability studies; second, among re-profiling work orders, rolling contact fatigue (RCF) is the principal reason; and third, by analysing re-profiling parameters, both the wear rate and the re-profiling loss can be monitored and investigated, a finding which could be applied in optimisation of maintenance activities.

This paper undertakes a reliability study using a Bayesian survival analysis framework to explore the impact of a locomotive wheel's installed position on its service lifetime and to predict its reliability characteristics. The Bayesian Exponential Regression Model, Bayesian Weibull Regression Model and Bayesian Log-normal Regression Model are used to analyze the lifetime of locomotive wheels using degradation data and taking into account the position of the wheel. This position is described by three different discrete covariates: the bogie, the axle and the side of the locomotive where the wheel is mounted. The goal is to determine reliability, failure distribution and optimal maintenance strategies for the wheel. The results show that: (i) under specified assumptions and a given topography, the position of the locomotive wheel could influence its reliability and lifetime; (ii) the Bayesian Log-normal Regression Model is a useful tool.

This paper proposes a new approach to study reliability of locomotive wheels with Bayesian framework, utilizing locomotive wheel degradation data sets that can be small or incomplete. In our study, a linear degradation path is assumed and locomotive wheels’ installation positions are considered as covariates. A Markov Chain Monte Carlo (MCMC) computational method is also implemented. In the case study, data were collected from a Swedish railway company. This data includes, the diameter measurements of the locomotive wheels, total distances corresponding to their “time to maintenance”, and the wheels’ bill of material (BOM) data. During this study, likelihood functions were constructed for Expontional regression models, Weibull regression models, and lognormal regression models.

The results show that the locomotive wheels’ lifetimes are dependent on installation positions. For the studied locomotive wheels data, the Lognormal regression model is a better choice, because the model obtained the lowest Deviance Information Criterion (DIC) values. In addition, under current operation situation (e.g. topography) and current maintenance strategies (re-profiled, lubrication, etc.), the locomotive wheels installed in the second bogie have longer lifetimes than those installed in the first bogie; the wheels installed on the “back” axle have longer lifetimes than those on the “front” axle; and the right side wheels’ lifetime is shorter than that for the left side under a given running situation.