Title

Optimization techniques for fault diagnosis and organizational design

Date of Completion

January 2007

Keywords

Engineering, Industrial

Degree

Ph.D.

Abstract

In this thesis, we model engineering problems and develop efficient optimization algorithms in the areas of fault diagnosis and organizational design. Two topics are addressed on fault diagnosis. The first topic deals with a multi-mode test sequencing problem, while the second solves the dynamic multiple fault diagnosis problem. Three topics on organizational design are covered. The first is on effects-based design of robust organizations, the second is on patrol planning in stochastic environments, and the third is on an agent-based simulation model for organizational analysis. ^ Multi-Mode Test Sequencing (MMTS) problem is a binary identification problem wherein one needs to develop a minimal expected cost test procedure to determine which one of a finite number of possible failure states, distributed among multiple modes, if any, is present. We propose computationally efficient heuristic algorithms based on information gain and rollout strategies for its solution. Dynamic Multiple Fault Diagnosis (DMFD) problem involves real-time inference of the most likely set of faults and their time-evolution based on blocks of unreliable test outcomes over time. We decompose the DMFD problem into a series of decoupled sub-problems, which are solved by a fast and high-quality deterministic simulated annealing method. A local search and update scheme is then applied to further improve the solution. ^ In Effects-based design of robust organizations, we model the dynamic system associated with the mission environment as a finite-state Markov Decision Process (MDP). Using this model, we determine a near-optimal action strategy that specifies which action to take in each state of the MDP model by the Monte Carlo control method. The range of missions, action strategy, and platform utilization measures are used to synthesize a robust organizational structure. The patrolling problem has the following characteristics: Patrol units conduct preventive patrol and respond to call-for-service. The patrolling locations have different priorities, and varying incident rates. We design a patrolling scheme enabling locations near-optimally visited based on their importance and incident rates. Finally, in agent-based simulation model for organization analysis work, we propose an approach for analyzing and simulating the performance of an organization in project-based mission environments based on the congruence model of organizations. ^