An adaptive approach to evaluation of implementation prospects of new construction technologies
Abstract
A neural network-based approach to estimating the adoption potential of new construction equipment technologies is presented. The research goal is to develop a method that helps select a new equipment technology for implementation for a considered construction operation. The main reason for using neural networks for this method is the requirement of performing complex cause-effect mappings in technology evaluation. To facilitate the analysis, the problem is broken down into three levels and at each level a neural network is developed to provide the corresponding estimation mechanism. The neural network at the top level establishes the relationships between the performance characteristics and the acceptability of a technology for an operation based on examples of existing alternative technologies and preferences of users. Eigenvalues of technology comparison matrices form the input vector while the result from a poll forms the output of each performance-acceptability pattern used to train the network. An illustrative case study is provided, in which simulated polls on alternative technologies for a concrete placement operation were used to generate training and testing data. The other two neural networks together perform evaluation of the productivity of a new technology in a given operation scenario and provide an important input to the technology acceptability estimate. The network at the bottom level estimates the production capacity of the technology based on its performance observed in example job conditions. The network at the middle level estimates the operating efficiency of the technology based on production data for comparable technologies in various operation conditions. For illustration, an experiment with a desktop robotic excavator model was developed to simulate excavation tasks in sample performance conditions for producing cycle time data for training and testing the production capacity estimate network. Separately, an operation simulation program was developed to simulate an excavation and hauling operation for generating sample operation attribute data for training and testing the operating efficiency estimate network. Since test results show that a sufficiently accurate estimate can be produced with a limited data collection effort for all three networks, this approach has the potential to provide an efficient method for estimating new construction technology acceptability.
Degree
Ph.D.
Advisors
Skibniewski, Purdue University.
Subject Area
Civil engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.