System of Multiple ANNs for Online Planning of Numerous Building Improvements

Abstract

The aging infrastructure in North America and worldwide mandates large investments in repair and improvement (R&I) activities. For organizations that own many assets, managing a large number of R&I activities is not a simple task and requires accurate estimating and scheduling so that proper budgeting and resource allocation decisions can be made. To support these decisions, this paper introduces a Web-based system that estimates the cost and duration of a user-requested R&I activity and provides alternative schedules based on resource availability. For estimating, the Web-based system hosts 32 artificial neural networks (ANNs), trained on actual historical data, for 32 common R&I activities in building projects. Each ANN incorporates a sensitivity analysis to consider the uncertainty in the input parameters on the estimate, and is linked to a central scheduling algorithm for resource allocation based on a first-come first-serve basis. The automated system helps practitioners in planning numerous R&I requests with least time, cost, and paper work. Details on system development are provided in this paper along with perceived benefits and the opinion of users on its performance.