on 23rd January 2019, Guang Zou (ESR5) became the second TRUSS ESR in successfully defending his doctoral thesis in front of an examination panel appointed by the Postgraduate Committee at University College Dublin (UCD). The thesis, titled “Probabilistic Methods for Life Cycle Management of Steel Structures under Fatigue” and directed by Dr. Kian Banisoleiman from Lloyd’s Register EMEA, Associate Prof. Arturo Gonzalez from UCD and Prof. Michael Habvro Faber from Aalborg University, was submitted on the 18th December 2018. The defence was held in meeting room 3.10 of Lloyd’s Register Global Technology Centre (GTC) in Southampton Boldrewood Innovation Campus. The jury was composed by Prof. Torgeir Moan (External Examiner from the Department of Marine Technology in Norwegian University of Science and Technology (NTNU)), Dr. Daniel McCrum (Internal Examiner from UCD), and Dr. John O’Sullivan (Chair from UCD). The VIVA was also attended by Kian Banisoleiman and Arturo Gonzalez.
Unlike the first TRUSS PhD VIVA, which was a public defence in UPC, the current VIVA took place under UCD regulations, where a VIVA is not opened to the public. Only the candidate, internal and external examiners and the chair are present. Supervisors can be present, but only as witnesses if the candidate authorizes it. The thesis is assessed in an oral defence VIVA examination going through the chapters of the thesis one by one, with the outcome being: (a) Award the Doctoral degree – no corrections required, (b) Award the Doctoral degree – corrections required, (c) Award the Doctoral degree – revision without re-examination, (d) Revise thesis and submit for re-examination, (e) Do not award the Doctoral degree – recommendation that the candidate transfer to an appropriate graduate programme, and (f) Do not award the Doctoral degree. More details of the UCD examination procedure can be found here. Guang’s defense began with a 10-minute presentation focused on main contributions of thesis, after which, the Examiners posed questions to Guang on a chapter by chapter fashion, concluding at 1:00 pm when the Examiners considered that all aspects of the thesis have been examined to their satisfaction. Following a short period of deliberation, the Examiners awarded the doctoral degree subjected to minor corrections.
-
-
A moment during VIVA discussions
-
-
From left to right: Arturo Gonzalez, Kian Banisoleiman, Guang Zou, Torgeir Moan and Daniel McCrum
The format of the thesis was in the form of a compendium of papers. The thesis was structured into 7 chapters and 9 appendices, including 4 papers submitted for publication in prestigious journals with high JCR impact (currently under review). Additionally, Guang has published 13 conference papers. His main contributions are summarized in the abstract of the thesis, which is reproduced below.
Thesis abstract
Structural deterioration as a result of crack initiation and growth is acknowledged to be the cause of many failures, the consequences of which are often unaffordable, especially when the structure is operated in extreme environments, such as marine and offshore structures, nuclear plants, aerospace crafts, etc. It is clearly needed to adopt efficient design, inspection and maintenance strategies that reduce the failure risk as well as the costs associated with these engineering strategies and activities. Optimization of design, inspection and maintenance is often challenging, as both fatigue deterioration and performances of (inspection and maintenance) interventions are subjected to a high degree of uncertainty.
Reliability and risk analysis together with probabilistic modelling represent strong tools for explicit representation of the effects of uncertainties, fatigue deterioration and maintenance interventions and for risk-informed and traceable engineering decisions under uncertainty. However, limitations and research gaps with current reliability/risk-based structural design and maintenance methods include: (a) almost all maintenance optimization is based on cost models which can be difficult to obtain, (b) probabilistic inspection optimization methods are usually computational demanding, and (c) typically, structural design, inspection and maintenance are only marginal addressed or optimized.
This thesis develops probabilistic methods and models to reveal the relationships between uncertainties, current engineering strategies and future outcomes, that allows developing optimal strategies accordingly, based on time-variant reliability analysis, life cycle risk assessment, value of information (VoI) theory and Bayesian decision analysis. The focus is placed upon optimization of maintenance strategies, especially condition-based maintenance (CBM). Besides, optimization of inspection and design strategies are jointly addressed from the perspective of reducing maintenance costs. Six main contributions are noted under each of the following bulleted points.
- A probabilistic method to quantify the maximum benefits of CBM to structural reliability and risk reduction. It is generally recognized that compared to time-based maintenance (TBM), maintenance costs can be reduced by CBM (introducing inspections and maintenance criteria, i.e. critical crack size for maintenance), and normally the CBM with the best compromise between maintenance costs and reliability is regarded as optimal. The method finds that the max reliability under CBM can be higher than under TBM (or the min risk can be lower), depending on the specific intervention times, which indicates that higher reliability can be achieved by less maintenance and thus some maintenance can be unbeneficial. Hence, reducing maintenance costs do not necessarily lead to a compromise in reliability, i.e. the objectives of reducing maintenance costs and increasing reliability sometimes agree with each other.
- A new maintenance classification method: beneficial, ineffective and unbeneficial maintenance. Such classification facilitates qualitative analysis towards fully utilization of VoI: the VoI of inspections under CBM lies in distinguishing beneficial maintenance from ineffective (or unbeneficial) maintenance. A compromise between maintenance costs and reliability is only necessary for beneficial maintenance, while ineffective (or unbeneficial) maintenance should always be identified and avoided.
- A probabilistic optimization method for maintenance criteria and for identifying ineffective maintenance. The method derives optimal maintenance criterion based on reliability. This reliability-optimal criterion is the smallest maintenance (crack size) criterion which defines the upper limit of maintenance, as adopting a smaller criterion than the reliability-optimal would lead to ineffective (when perfect maintenance is assumed) or unbeneficial maintenance (when imperfect maintenance is adopted). Such a limit is of great significance for designing maintenance strategies, especially when exact cost-models are not available. The investigation concludes that a higher reliability can be achieved adopting CBM and reliability-optimal criteria, than adopting TBM. This is true for any intervention times.
- An optimal CBM strategy with time-varied maintenance criteria. When a sequence of maintenance interventions is scheduled for the whole service life, the developed maintenance optimization method shows that it is not optimal to use the same criterion for all interventions. Both the reliability-optimal and cost-optimal (crack size) criteria become larger as the remaining service life becomes shorter.
- A novel and efficient computation method for the VoI of a sequence of inspections and for optimization of inspection strategies. VoI is a strong tool for understanding the impacts of uncertainties on decisions, but its computational requirements become prohibitive in complex maintenance decision contexts. An efficient method is proposed based on a holistic treatment of unknown information (i.e. a sequence of unknown inspection results) and an adaptive maintenance strategy, by which combined effects of a sequence of maintenance interventions are captured while greatly reducing computational time. For well-designed inspection strategies, VoI > 0 and the adaptive maintenance strategy is the optimal; when adaptive maintenance strategy is not optimal, VoI = 0 (signifying no need for inspections) or at least part of inspections is of no value (indicating that the inspection strategy is not optimal and needs to be optimized). It is shown that the VoI is strongly dependent on the available maintenance action alternatives and their possible sequence, and adaptive maintenance strategy is more likely to be the optimal strategy when more interventions are scheduled. In addition, more inspections and/or more accurate inspection methods do not necessarily bring more value. Lastly, the optimal inspection strategies by VoI-based and cost-based optimization can be different, although the VoI is assessed based on expected costs.
- A risk-informed method for holistic optimization of structural design, inspection and maintenance against fatigue. The main idea is that when design and CBM strategies are marginally addressed, the derived strategies may not be optimal from the perspective of the whole life cycle, e.g. the optimality of CBM derived for an existing structure is dependent on the design reliability (or safety margin). As such, a risk-informed method that envelopes major engineering strategies and uncertainties affecting fatigue performance is developed. The latter optimizes design and CBM strategies holistically, based on life cycle risk quantification. The method is compared to reliability-based methods and to sequential optimization methods. It is shown that a compromise between design costs, maintenance (including inspection) costs and reliability can be achieved by the method. The derived optimal design and CBM strategies are associated with less life cycle costs (LCC). The method allows deriving the fatigue reliability level by risk-informed optimization (i.e. it is not prescribed) and the derived optimal strategies are shown to be adaptive to failure consequences. The decision making for design and CBM strategies is consistent and traceable, as it is based on the same risk quantification model used for fatigue failure. -> Read more about the research by Guang in TRUSS