Automation of Roof Construction Management by Means Artificial Neural Network

Abstract

To enable efficient managerial decisions in planning and generation of construction process, we need technologies to analyze and evaluate their conditions and to predict their further evolution. One of advanced methods to generate them is based on artificial neural networks (ANN). This paper represents the outcome from a research in construction process organization and management to build roof structures at the planning and operation stage, assisted by the ANN methodology. Relying on systems analysis and expert survey, the authors have designed a 4-layer feedforward ANN with a node pattern of 15-6-3-1. The authors used the concept of fuzzy sets to measure input and output of the network. The paper clearly illustrates the function of ANN and describes the learning algorithm by the backward propagation of errors relative to the resulting ANN. The authors have identified problems related to generation of the labeled training data in the researched field and developed a network learning approach that combines extraction of expert knowledge about the system with learning on training examples from real-life construction project.