Back to results list
Please use this identifier to cite or link to this item:
|Title:||Optimal life-cycle management of transportation asset systems||Authors:||Zhang, Le||Advisors:||Gu, Weihua (EE)
Lee, Jinwoo (EE)
Fu, Weinong (EE)
|Keywords:||Transportation -- Management
Pavements -- Management
Pavements -- Maintenance and repair -- Management
Motor vehicle fleets -- Management
|Issue Date:||2019||Publisher:||The Hong Kong Polytechnic University||Abstract:||Transportation assets play a vital role in the national economy and in people's daily life. In this thesis, we develop new models of optimal life-cycle management for two common types of transportation asset systems: highway pavement systems and truck fleet systems. The pavement system models can be tailored for the optimal management of other infrastructure systems, such as bridges and tunnels; and the truck fleet models can be applied with modification to other vehicle asset systems, such as bus, airplane, and maritime vessel fleets. For the optimal management of highway pavement systems, we first present a novel optimization model with deterministic pavement deterioration. The model jointly optimizes the schedules of three common types of pavement management activities (treatments), i.e., preventive maintenance, rehabilitation, and reconstruction/replacement (MR&R), for a system of pavement segments under budget constraints. The objective is to minimize the total costs incurred by both highway users and pavement management agencies. We propose a Lagrange multiplier approach to relax the budget constraints, and employ derivative-free quasi-Newton algorithms to find the optimal solution. By relaxing the budget constraints, the solution approach decomposes the optimization problem for a pavement system into optimization subproblems for each pavement segment. Hence, it can be applied to pavement systems with any models of segment-level cost and treatment effectiveness, as long as the segment-level optimization subproblems are solvable. This approach also ensures a bounded optimality gap for a system-level solution, as long as the segment-level solutions have bounded optimality gaps. (In other words, it guarantees the near-optimality of the system-level solution if the segment-level subproblems can be solved to near-optimality.) Finally, the approach exhibits linear complexity with the number of pavement segments, which ensures computational efficiency, especially for optimizing large-scale systems.
We further extend this work to the stochastic case, which accounts for uncertainties in the pavement deterioration process. Inspection activities whose aim is to reveal the actual pavement state are included in the menu of management activities for scheduling. We formulate a semi-continuous model for selecting optimal inspection scheduling and management policies; and propose a statistical learning approach to update the model parameters sequentially after inspections. We employ approximate dynamic programming to solve the segment-level problem, which has a great advantage in terms of computational efficiency for solving large-size problems. Managerial insights are unveiled in numerical case studies, which can help highway agencies formulate more cost-effective inspection and management policies and budget allocation plans. The vehicle fleet management problem is different from the infrastructure management problem, mainly because: i) the fleet size can vary over time, while an infrastructure system is usually fixed in size; and ii) a vehicle's maintenance cost and replacement schedule are affected by its utilization, which can be controlled by the fleet manager (the utilization of highway pavement by private vehicles, in contrast, is difficult to control). In this thesis, the truck fleet management problem is formulated as a mixed-integer nonlinear program, which jointly optimizes the purchase, replacement, and utilization policies for the trucks under deterministic, time-varying demand. Our contribution is mainly methodological. We first approximate the original discrete formulation with a continuous-time one, where the numerous decision variables in the discrete-time model are replaced by a few decision functions in continuous time. This continuous approximation technique allows us to derive the analytical conditions for optimal utilization plans using the calculus of variations. These optimality conditions are then converted back to the discrete-time kind, and are used to develop a demand allocation rule, which greatly reduces the solution space of the original problem. The reduced program can then be solved efficiently using heuristic methods, such as tabu search. Based on extensive numerical case studies, we verify that this novel approach can produce better solutions with much lower computational cost than previous solution algorithms proposed in the literature. The thesis concludes with a discussion on potential extensions under the proposed methodological frameworks to account for more realistic features in real-world transportation asset management problems, and to model other types of infrastructure and fleet asset systems.
|Description:||viii, 143 pages : color illustrations
PolyU Library Call No.: [THS] LG51 .H577P EE 2019 Zhang
|URI:||http://hdl.handle.net/10397/80412||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|991022190958103411_link.htm||For PolyU Users||167 B||HTML||View/Open|
|991022190958103411_pira.pdf||For All Users (Non-printable)||4.09 MB||Adobe PDF||View/Open|
Citations as of Mar 12, 2019
Citations as of Mar 12, 2019
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.