Content area
Abstract
Modular construction is a method for constructing units of a project remote from the final project site. It brings the advantage of the manufacturing processes to the construction industry. It presents an opportunity to improve a variety of performance parameters relating to the project, such as, reduction of construction time, reduced labor costs, increased quality and efficiency, and simultaneous construction.
The research performed has addressed the following issues of modularization, acquiring and structuring the knowledge of the experts in the field through knowledge engineering procedures, developing a decision model for modularization feasibility (pre-screening, detailed analysis and economic analysis), and developing a neural network system for modularization feasibility decision support.
The original prototype system used a pure expert system approach. Based on various decision factors, it recommends an appropriate degree of construction modularization. If modularization is recommended, it also provides the user with an approximate change in costs and schedule due to modularization. The system is then trained using neural network approach, utilizing multi-layered unsupervised learning. This trained version of the system has some advantages over the initial expert system, such as the ability to generalize, performing better under incomplete information, and improving the basis of the initial expert system.