Neural Network–Spreadsheet Integration for Earth‐Moving Operations

Abstract

Abstract: This paper examines the potential to combine trained artificial neural networks with application development software to expand their applicability to construction management problems. The problem of determining optimal equipment combinations for the dozer/scraper earth‐moving team, under various physical conditions of the haul road, was chosen for its limited scope and lack of effective solution methods currently available. An introduction to the problem provides the current state of technology for handling this problem. The problem is analyzed to determine the information and resource requirements as well as the input and output forms that would be desirable for such a program. The paper proceeds through the design process and development of the program schematic. Programming of the system is examined, including the development process used to train the neural network, conversion of the network to a spreadsheet form, and spreadsheet manipulation of the network outputs. The paper concludes with recommendations for expansion of the program for more complex haul road conditions and equipment types.