Development of a grid service for multi-objective design optimisation

Abstract

The emerging grid technology is receiving great attention from researchers and applications that need computational and data capabilities to enhance performance and efficiency. Multi-Objective Design Optimisation (MODO) is computationally and data challenging. The challenges become even more with the emergence of evolutionary computing (EC) techniques which produce multiple solutions in a single simulation run. Other challenges are the complexity in mathematical models and multidisciplinary involvement of experts, thus making MODO collaborative and interactive in nature. These challenges call for a problem solving environment (P SE) that can provide computational and optimisation resources to MODO experts as services. Current PSEs provide only the technical specifications of the services which is used by programmers and do not have service specifications for designers that use the system to support design optimisation as services. There is need for PSEs to have service specification document that describes how the services are provided to the end users. Additionally, providing MODO resources as services enabled designers to share resources that they do not have through service subscription. The aim of this research is to develop specifications and architecture of a grid service for MODO. The specifications provide the service use cases that are used to build MODO services. A service specification document is proposed and this enables service providers to follow a process for providing services to end users. In this research, literature was reviewed and industry survey conducted. This was followed by the design, development, case study and validation. The research studied related PSEs in literature and industry to come up with a service specification document that captures the process for grid service definition. This specification was used to develop a framework for MODO applications. An architecture based on this framework was proposed and implemented as DECGrid (Decision Engineering Centre Grid) prototype. Three real-life case studies were used to validate the prototype. The results obtained compared favourably with the results in literature. Different scenarios for using the services among distributed design experts demonstrated the computational synergy and efficiency in collaboration. The mathematical model building service and optimisation service enabled designers to collaboratively build models using the collaboration service. This helps designers without optimisation knowledge to perform optimisation. The key contributions in this research are the service specifications that support MODO, the framework developed which provides the process for definining the services and the architecture used to implement the framework. The key limitations of the research are the use of only engineering design optimisation case studies and the prototype is not tested in industry.