Engineering performance improvement based on the integration of genetic algorithms and artificial neural networks

Lei Zhang, Purdue University

Abstract

The sector of industrial facility construction has been experiencing unsuccessful project implementations for a long time. Both the industry and the academic world have realized the significant impacts of engineering activities on the success of project implementation. Improved engineering design performance leads to better project outcomes. However, industrial construction projects are complex processes involving large number of input and output. Therefore, first of all comes the need to understand how well the engineering activities are performed. Researchers and industry experts have been making efforts in measuring engineering performance. Better understanding of engineering performance lays the foundation for stepping forward to seek ways to improving engineering performance. Former studies on engineering performance improvement have focused on the promotion of certain techniques or products, or looked at specific engineering processes or areas. Few tried to make contribution to the whole facility development process. There is a lack of a systematic and analytical approach that improves engineering performance based on the understanding of the cause-effect relationships between engineering input and performance output from the perspective of the whole facility development process. This research proposes a neurogenetic system, which integrates genetic algorithms with artificial neural networks, for modeling engineering performance measurement and improvement in industrial construction projects. The system starts with a neural network model for establishing the cause-effect relationship between engineering input factors and engineering performance output measures. Because of its robust and efficient searching ability in complicated situations, genetic algorithms are employed to search for better engineering performance; the fitness function for the genetic search is the neural network model that predicts engineering performance. To make suggestions for possible engineering performance improvement, the research introduces the self-comparison evaluation that evaluates a project's engineering performance by comparing its actual engineering performance with its possible better engineering performance generated by the genetic search. Using real project data, the research developed and tested the proposed system. The testing produced significant results that demonstrated the plausibility of the GA-ANN integration in seeking the potential engineering performance and illustrated how the self-comparison concept could provide unique, project-specific, and objective engineering performance evaluation.

Degree

Ph.D.

Advisors

Chang, Purdue University.

Subject Area

Civil engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS