Join our welcoming International Design Community.
Complete this application form to join now.
Author: Melo, Andrés Felipe
Supervisor: Clarkson, P. John
Institution: Department of Engineering, University of Cambridge
This thesis describes a computational model for aiding design process planning. The view is taken that planning is concerned with selecting and ordering design tasks by considering the timely success of the project, and its corresponding risk of network.
The development of the computational model necessitates a mathematical model of decision and risks, and a knowledge representation of the design process; implementation and adaption of search and optimisation algorithms, and definition of a plan representation that can be used for scheduling. The design process is modelled as a Markov decision model, which uses classical decision theory. For this, the representation is based on describing explicitly the states of the design; and design actions are defined in terms of the changes they can produce on the state.
Efficient algoritms and an economic planning representation were developed for finding near-optimal plans. The resulting algorithms enable the analysis of large models that cannot be handled with standard Markov analysis due to computational complexity. The plan representation adds functionality for scheduling, by indicating the possibility and consequences of changing them, in terms of risk.
The model and algorithms were evaluated with three industrial case studies to determine the fitness of the representation, efficiency of the algorithms, and use of the results. the results are satisfactory in three senses and provide grounds for future research in planning decision and risk management.