Methodology for analyzing the trade‐offs associated with multi‐objective optimization in transportation asset management under uncertainty. (23rd November 2020)

• Record Type: Journal Article
• Authors: Bai, Qiang; Miralinaghi, Mohammad; Labi, Samuel; Sinha, Kumares C.
• Is Part Of: Computer-aided civil and infrastructure engineering. Volume 36:Number 4(2021)
• Page Start: 381
• Page End: 401
• Publication Date: 2020-11-23
• Subjects: Civil engineering -- Data processing -- Periodicals; Computer-aided engineering -- Periodicals; 624.0285
• DOI: 10.1111/mice.12637
• Languages: English
• ISSNs: 1093-9687

Abstract

Abstract: Trade‐off analysis, one of the key tenets of multi‐objective optimization for the project selection problem in Transportation Asset Management (TAM), can help decision‐makers quantify and comprehend the consequences of different resource allocations in terms of the multiple measures of system performance. In analyzing TAM trade‐offs, it is vital to account duly for the uncertainties associated with these system‐wide performance measures. In this paper, we present a methodology that integrates chance‐constraint programming, the Lindeberg Central Limit Theorem, and a hybrid NSGA II method, to address the performance uncertainties associated with the TAM multi‐objective optimization problem. Through analyzing the trade‐offs between expenditure and performance, and between different performance measures, we generate Pareto frontiers at different confidence levels using a hybrid NSGA II method. We demonstrate the proposed methodology using a case study involving real‐life assets and the expected cost and performance benefits of projects associated with these assets. Regarding the trade‐off between cost and performance, we determine the extent to which the strengths of these relationships vary across different confidence levels. We find that, generally, for a given network performance level, a higher expenditure is needed to achieve a high confidence level compared to the expenditure needed to achieve a low confidence level, and more importantly, measures these sensitivities. This is the "Price of Confidence" concept. Regarding the trade‐off between different pairs of performance measures under budgetary constraints, we show how these relationships vary with the confidence level specified for the analysis, and we measure the extent to which higher confidence level requirements translate into lower levels of overall system‐wide performance.