Proceedings of the 11th IMA International Conference on Modelling in Industrial Maintenance and Reliability

Authors A. Raza and V. Ulansky

https://doi.org/10.19124/ima.2021.01.1

Abstract

This study proposes a mathematical model for assessing the trustworthiness indicators of the operability checking for a deteriorating system. The set of mutually exclusive events at the time of operability checking are analyzed. Correct and incorrect decisions correspond to events such as true–positive, false–positive, true–negative, and false–negative. General expressions for computing the probabilities of possible decisions when checking the system operability at a discrete time are proposed. The paper introduces the effectiveness indicators of corrective maintenance such as average operating costs, total error probability, and a posteriori probability of failure–free operation. We illustrate the developed approach by calculating the probabilities of correct and incorrect decisions for a specific stochastic deterioration process.

Z. Zhang and C.G. Lee

https://doi.org/10.19124/ima.2021.01.2

In recent decades, there has been significant growth in the development of rechargeable battery–powered devices, such as electric vehicles, leading to a huge demand for batteries with high reliability and quality. End of life (EoL) is a critical indicator of battery health and can be estimated by either adaptive stochastic processes or advanced machine learning techniques. However, these approaches follow the degradation path that can be modelled as simple mathematical form such as linear or exponential function and lack interpretability due to its black–box nature. To address these shortfalls, an GRU–driven degradation process is proposed to learn complex battery degradation patterns, in which degradation progression is controlled by a recursive Gaussian distribution with its mean learnt from an GRU–driven degradation pattern. Due to the non–Markovian state transitions, a joint–learning sampling–based expectation maximization algorithm is developed to estimate model parameters based on historical observations. To
validate the superiority of the proposed methods, a case study of battery data was implemented. The results show a better performance with respect to EoL accuracy than that achieved with traditional methods.

R. Yan, S.J. Dunnett and L.M. Jackson

https://doi.org/10.19124/ima.2021.01.3

Abstract

Condition monitoring has demonstrated its effectiveness in improving the economic return of wind turbines. However, a wind turbine consists of hundreds, even thousands, of mechanical, electrical and power electronic components. The failure of any one of them may lead to the shutdown of the turbine. For this reason, a variety of component monitoring systems have been developed dedicated to monitoring these different components. Consequently, a wind turbine usually needs to be monitored simultaneously by several different types of component monitoring systems that benefit wind turbine operation and maintenance to
different extents. This not only increases the complexity of the hardware configuration but also increases the costs of the entire condition monitoring system. How to achieve a condition monitoring system that can monitor the most vulnerable components whilst bringing the most economic benefit to the wind turbine operator is an important question. The aim of this paper is to answer such a question with the aid of the Petri net modelling method. The model developed in the paper will investigate the influences of condition monitoring systems and fault detection using wind farm Supervisory Control and Data Acquisition (SCADA)
system on the economic return of wind turbines, thereby providing a feasible tool for constructing an optimal wind turbine condition monitoring system.

R. S. Lopes, P. Do, C.A.V. Cavalcante and B. Lung

https://doi.org/10.19124/ima.2021.01.4

Abstract

Maintenance decisions in multi–component systems are of great interest to maintenance managers. The equipment during its operation, produces and stores a large amount of data, especially discrete event data such as alarm, failed situation, change of operation modes, stop of the systems, and so forth, and produced via processings supported by the programmable logic controller (PLCs), supervision system, SCADA.
Considering this data to assist in maintenance management and decisions is an area with a growing interest in maintenance management. In this paper, we study the stochastic dependency in a multi–component system through data from PLCs database. We use appriory algorithm and affinity function to identify failure
dependence in multi–component systems. The results of failure dependence can be used as input for planning group maintenance, purchase spare parts, or planning opportunistic maintenance.

S. Rezvani and N. Almeid

https://doi.org/10.19124/ima.2021.01.5

Abstract

In construction projects worldwide, Main Contractors (MC) have to choose Sub–Contractor (SC) with varying degrees of previous experience. This adds an extra layer of uncertainty to the project that may impact costs and time directly or indirectly. Adequately assessing and selecting amongst alternative SCs is a well–known risk management strategy to mitigate threats and realize opportunities. But this is often done empirically. This paper discusses a mean to identify and weight SC and MC characteristics in each project’s context in a way that enables a systematic and optimized decision–making process when selecting SCs in construction projects. Multi–criteria decision analysis (MCDA) and M–MACBETH software is applied to an empirical case study from a construction company in Malaysia. A cabling project of a double electrified railway in Malaysia provides the context for the decision–making problem, criteria, audit data and information. A sensitivity and robustness analysis is included in the paper.

W. Li and R. Dwight

https://doi.org/10.19124/ima.2021.01.6

Abstract

Because of the complexities of the railway track system compositions and
their degradation mechanisms, the existing models built for relevant LCC estimation are either too simplified or focused on a part of the system. In this paper, a simulation – based model is proposed which is to incorporate all of the major track components such as rail, ballast and sleepers. A cost breakdown methodology is adopted to estimate the cost on a component-by-component and activity-by-activity basis. The interactions of track components and between maintenance and track components can be considered during the simulation process. Some activities whose implementation is purely experience-based can be loosely added as extra cost elements by disregarding its connection with other activities. Though attempting to embrace all, the simulation-based model is still considered easier to implement and potentially faster to run compared to the more complex and probably more powerful models including the Petri-Net models: the tools for its development are readily available. Apart from that, only knowledge on Monte Carlo simulation is required. In addition to providing a tool for the planning of maintenance actions, the model may also be used to evaluate different
track form options. By replacing the simulation-based models with parametric models or limiting the focus to one or several track components, the model can be simplified conveniently.

A.C.J .Santos, C. A. V. Cavalcante and S.Wu

https://doi.org/10.19124/ima.2021.01.7

Abstract

The delay time concept has been used in a wide variety of applications in maintenance policy optimisation. In some situations, the sojourn time of an item at a defective state may hamper the performance of the item due to its reduced efficiency, and may therefore incur more repair cost. Consequently, the longer the sojourn time in a defective state, the worse the influence may cause. Despite these facts, in the literature, there is little research investigating the influence of the sojourn time in the defective state on its relevant cost. In addition, a limited number of papers study delay time modelling for the case that repaired items can be reused. This paper therefore extends a recently published paper by Santos et al. (2021) and presents some insights into the delay time modelling for second–hand items. In general, the proposed model may be applied to the scenarios where used items may be replaced by a new or refurbished one. The main contribution of this paper is on its exploration of methods of item reuse, from which new directions for further studies are established

D. Brüggemann, R. Chan, H. Gottschalk and S. Bracke

https://doi.org/10.19124/ima.2021.01.8

Abstract

The computer vision task “semantic segmentation” forms a crucial building block in the interaction vehicle. This allows distinguishing irrelevant FNs from potentially relevant FNs and thus provides more safety- abilities but also on distance information of VRUs within a safety-relevant region of interest ahead of the ego semantic segmentation models for autonomous driving not only based on their classification and localization the VRU to be found. In this work we therefore introduce a sophisticated evaluation framework that assesses consistently specify how well the VRU instance must be covered by the perception model in order to consider relevant as detection errors of VRUs on the path ahead. Moreover, standard evaluation approaches do not road users (VRUs, e.g. pedestrians) far away from the probable travel path of the ego vehicle are not as safety- be considered in a more differentiated way. For example in autonomous driving, detection errors of vulnerable overlooked by the perception model. From a practitioner’s point of view, however, faulty detections need to one commonly considers the number of false negatives (FNs), i.e. counting instances that have been of several redundant systems. As metric to evaluate the performance and reliability of this perceptual function,

Y.R. Melo, C.A.V. Cavalcante, R.S. Lopes and P.A. Scarf

https://doi.org/10.19124/ima.2021.01.9

Abstract

In this paper, we examine the impact of opportunities on systems that are maintained periodically. In a fleet of systems, such as a wind farm, an irregularity in one can serve as an opportunity for others, allowing maintenance actions to be carried out at a lower cost. In this context, we propose a maintenance policy with two phases. In the first phase there are inspections, and in the second phase, only corrective maintenance is carried out. Our main idea is to introduce flexibility to phase two, whereby opportunistic replacements may be carried out. Besides the fact that any action can be taken at predefined fixed times, we also consider defaulting which is the impediment to perform maintenance actions related to the limit of resources available or any external event. The proposed model is based on the delay time concept, where a single component system has three states: good, defective and failed. The system operates in both the good and the defective state. The defective state is only observable by inspection. The purpose of this study is to provide some insights for the decision–maker related to best practice when a system is accessible only at pre–planned, fixed times, where a default can occur, and installation problems can affect the reliability of the system and significantly change its expected life. The results of the numerical study show that the proposed policy has advantages when opportunities offer a less expensive means of preventive replacement and are frequent. For the benefit of the maintenance manager, especially when facing large logistic costs, maybe it is more interesting to be aware of what to do, when opportunities happen, than follow a restricted sequence of inspections.

S. Najafi, J.Y.J Lam, C. G Lee

https://doi.org/10.19124/ima.2021.01.10

Abstract

Maintenance planning plays an essential role in improving the reliability of a system and its effectiveness. Condition–based maintenance (CBM) suggests preventive actions to avoid failures, considering the state of a system. In this paper, an opportunistic CBM policy is developed for a series system composed of two units whose conditions are monitored regularly. The reliability functions are
obtained using the proportional hazards model (PHM) to estimate the system’s remaining useful life. When the deterioration of a unit exceeds a predefined threshold, a soft failure occurred, which can be detected through inspection. Hard failures are rectified by a corrective action immediately, including minimal repair, general repair, and replacement. The objective is to find an optimal policy that minimizes the total expected cost of the system. The problem is formulated in the semi–Markov decision process (SMDP) framework, and a reinforcement learning algorithm is proposed to find the optimal control policy and the long–run expected average cost per unit time. This study extends previous models, suggesting a new opportunistic CBM policy using the PHM in which actions with different levels can be performed on a two–unit series system.

S. Rezvani and M. C. Gomes

https://doi.org/10.19124/ima.2021.01.11

Abstract

Transportation network, especially highways, is considered a national or international asset, and by proper maintenance system, public and private organizations can prioritize the budgeting of repair and reconstruction. The problem is to have a reliable and practical model creating a solid understanding of the pavement degradation condition by inexpensive measurable parameters for municipalities. This study focuses on the road pavement condition, particularly the statistical evaluation of the processes of degradation involved in various road sections. Quantitative statistical analysis of a sample taken from the Iowa Department of Transportation (DOT) in the United States provides a better understanding of the needs in pavement maintenance processes. In addition, it can identify the critical factors of pavement maintenance. Through a case study, it is shown that organizations can develop a solid based statistical decision–making model using basic and low–priced parameters. The model has two approaches, with and without pavement type (used by creating several dummy variables to include each pavement type as independent variables). This study will positively enhance the pavement degradation prediction through a statistical analysis model and a case study of the Department of Transportation (DOT) of Iowa, USA. These details include explanatory models, bivariate correlation, principal component analysis, hierarchical and non–hierarchical clustering, creating dummy variables, and developing multivariate regression models.

E.T. Senalp

https://doi.org/10.19124/ima.2021.01.12

Abstract. Availability, Reliability and Maintainability (ARM) are three attributes that collectively affect the economic life–cycle costs of a platform (e.g. Ship, Aircraft). Whilst there are different approaches, this paper presents modelling of Intrinsic Availability and Reliability (Ai&R) of Key Platform Functions (KPF) and System Functions (SF) adopted as a modelling approach during the platform design and manufacture (D&M). Ai is a measure of the proportion of time, which required functions are successfully provided, when used in stated conditions. It is a useful indication of the potential availability that the platform offers separate from operational and logistics constraints. When high reliability is achieved and supportability optimised, corrective maintenance is minimised, hence in–service availability realised tends towards the Ai and a limited number of failures are expected that result in support / logistic delays. Ai&R modelling using increasingly mature data, during the D&M phases provides visibility of potential impacts on the Ai&R implicit in the design. Estimating the Ai&R at equipment, SF and KPF levels gives a consistent approach within a complex platform design and allows improved comparative analysis. This paper presents benefits and outcomes of the application of functional ARM for platform modelling and simulation.