Conceptual Reference Database for Building Envelope Research Prev
Next

Integrated bid preparation with emphasis on risk assessment using neural networks

Hegazy, T.
1993
Thesis, Ph.D. (Building), Centre for Building Studies, Concordia University, Supervisors: O. Moselhi and P. Fazio

Hegazy, T., (1993), Integrated bid preparation with emphasis on risk assessment using neural networks, Thesis, Ph.D. (Building), Centre for Building Studies, Concordia University, Supervisors: O. Moselhi and P. Fazio.
Abstract:
Construction estimating works as the basis for various strategic decisions regarding the preparation of bid proposals, procurement plans, various levels of schedules, and job cost control. Under the highly risky environment of the prevalent competitive bidding practice, preparation of realistic estimates pertaining to those management decisions has been a complex task that is often performed on an ad hoc and piecemeal manner. Conventional procedures and tools have proved inadequate to provide a structured decision aid that, under such environment, maximizes the contractor's chances of winning a job with maximum potential profit, and further generates practical baseline plans needed for job control to maintain this profit. Yet the situation has been translated into a high percentage of business failures, a high potential for claims, and at best a low profit margin in the industry. This research presents a methodology for an integrated cost estimation and bid preparation, with emphasis on the assessment of bidding risks and optimum markup estimation. The methodology utilizes available tools (algorithms, database management systems, and Al-based techniques) that can benefit from current industry practice and provide an adequate decision aid during bid preparation. The methodology facilitates integration among estimating, planning and scheduling, and bid unbalancing. It incorporates enhancements to the various functions that cover the quantitative aspects of an estimate including: direct and indirect cost estimation, planning and scheduling, and resource utilization. This enables detailed estimates of costs and durations to be generated for all the project tasks, with minimal redundancy and in less time. Such estimates also establish the baselines needed for efficient job control. For practicality, the methodology accounts for the qualitative (risk-related) factors that play a vital role in the preparation of competitive bid proposals (e.g., competition, market conditions, and contractor keenness for the job). The methodology utilizes Neural Networks, an Al-based technique that employs a learning mechanism and emulates the human ability to solve pattern recognition tasks similar to many problems encountered in construction. This technique is introduced as a new tool to the industry, incorporating several potential applications. A neural network model is designed and used to arrive at an optimum markup value that maximizes that contractor's potential profit and predicts the probability of winning the job at such level of profit, in response to the project risk pattern. The methodology then utilizes the data obtained through the detailed estimate to optimally unbalance the final bid, in an effort to improve the contractor's cash flow while maintaining his competitiveness. A PC-based prototype is developed to automate the bid preparation process and an example application is presented in order to demonstrate the effectiveness and practicality of the proposed methodology. The proposed integrated methodology contributes to current automation efforts in construction and its modular architecture allows for further enhancement and expansions. The developments made with respect to the markup estimation problem demonstrates the powerful capabilities of neural networks and the potential benefits of deriving analogy-based solutions to complicated construction problems that are characterized by high uncertainty. This approach could readily be utilized in other domains in construction management where solutions are based primarily on holistic analogy and traditional algorithmic solutions are inadequate.
Related Resources:

Related Concepts



CRDBER, at CBS, BCEE, ENCS, Concordia,