Integrating Prognostics and Health Management Information with Simulation-Based Optimization for Complex Maintenance System


Student thesis: Doctoral Thesis

View graph of relations

Related Research Unit(s)


Awarding Institution
Award date9 Sep 2019


For complex engineered systems, maintenance plays a critical role in improving system life and ensuring a satisfactory level of performance becomes more and more complicated due to rapid technological advancements. Hence, maintenance policies have seen a paradigm change from traditional maintenance policies (preventive maintenance and corrective maintenance) to classical condition-based maintenance (CBM) policy (which employs sensor technologies to monitor and recommend maintenance based on actual operating conditions), and in recent times to prognostics and health management enabled CBM policy (CBM supported by diagnostic and prognostic capabilities). However, to properly understand the benefits of integrating PHM-enabled CBM in complex maintenance systems for practical deployment, it is insufficient to develop prognostics algorithms alone without considering the analysis of the developed solution in actual settings. To demonstrate the potential advantage of prognostic capability in practical implementation, discrete event simulation (DES) is often used as opposed to using analytical models, due to intractability and inadequate realistic scenarios problems inherent in analytical models. This thesis, through different studies, investigates the benefits of adopting PHM for complex maintenance operations in different practical maintenance settings.

First, the idea of investigating the influence of resources and the role of prognostics information in CBM using DES without the simplifying conditions of perfect detection and diagnosis is considered in the following work.

Impact of resources and monitoring effectiveness on prognostic enabled condition-based maintenance policy. In the literature, the assessment of the role of prognostic information in Condition Based Maintenance (CBM) policy has been based on the assumptions of perfect condition monitoring and diagnostics. However, the key requirements for effective prognosis require both detection and diagnosis. This research focuses on practical CBM implementation from a new perspective by using a user-friendly Excel-based interface integrated with ARENA® based Discrete Event Simulation (DES) to assess and analyze the impact of joint resources and monitoring effectiveness on the key critical phases in CBM (detection, diagnostics, and prognostics) implementation while using the practical performance metrics of overall cost and operational availability. The objective of this thesis is to understand without and with optimization how the influence of resources and monitoring effectiveness affect asset availability and overall total cost under CBM policy, as well as to investigate under which monitoring condition the added value of prognostic information in CBM could be an advantage over classic CBM. The work demonstrated without optimization that prognostics-enabled CBM provided superior technical benefits but might not necessarily result in global overall total cost reduction on a system-wide level. However, with optimization, overall total cost-effectiveness was achieved with prognostics-enabled CBM policy. The proposed model can provide maintenance operation decision-makers implementing CBM with numerical evidence in assessing the benefits, and adoption of prognostics and health management (PHM) in their operation.

Using real-world degradation data to provide an actual representation of the real-world behavior of complex systems, DES is integrated with analytical degradation and RUL prediction models to implement the concept of multi-step RUL within DES framework.

A stochastic discrete event simulation incorporating multi-step prediction for predictive maintenance. Increased demand for multifunctional complex engineered systems to meet targeted high safety and reliability requirements has transformed the traditional role of maintenance from a reactive approach to a proactive approach. Prognostics and Health Management (PHM) exerts significant influence on maintenance decision making, specifically for imminent failure contingency response and mitigation plans. The ability to model complex systems due to flexible modelling assumptions, thus providing a better understanding makes DES model the preferred choice in the assessment of the impacts and benefits of prognostic information in maintenance systems. This work developed an integrated approach, wherein sufficiently accurate, computationally efficient, and environmentally portable analytical models based on recent statistical prognostic models (mixed-effects bearings prognostic model with multiplicative normal random error, and mixed-effects bearings prognostic model with multiplicative Brownian motion error) are used for degradation modelling and multi-step remaining useful life (RUL) prediction, while DES model is used for modelling the complex interactions and maintenance management system. The developed integrated framework is assessed with a case study of a multi-component manufacturing system using actual real-world vibration-based degradation data. Overall maintenance cost, throughput, and throughput profit are computed and compared with two other benchmark RUL prediction methods: one-step and two-step RUL prediction methods.

Using the similar degradation data, we proposed the use of RUL information in mission assignment and maintenance scheduling of degrading moving assets under resource constraints, such that the most value is obtained from the asset prior to maintenance being carried out.

Remaining useful life-based maintenance scheduling for degrading moving assets. In the literature, maintenance decision making of vehicular assets using predicted remaining useful life (RUL) information in prognostic and health management (PHM) applications rely on the use of simulated RUL information from random distributions to demonstrate the benefits of the approach. However, this approach might not be effective as these assets operate in highly dynamic environments and require the RUL information to be regularly updated. Hence, in this research work, we proposed the framework of RUL-based scheduling for degrading moving assets using actual real-world degradation data to efficiently and effectively assign required missions and recommend maintenance operations under the constraint of maintenance resources. Numerical results demonstrate the benefits of the proposed method and provide an implementation guide on how RUL information obtained from moving assets can be integrated into maintenance planning and scheduling optimization.

Through a series of research works related to maintenance of complex engineered systems, the thesis aims to better understand and assess the benefits of integrating PHM information in complex maintenance systems using discrete event simulation and optimization approach. The findings in this thesis provide both an implementation guide and numerical evidence for maintenance managers in service systems or manufacturing organizations that are considering adopting PHM in their organizations.