| The lack of specific data from field-based operations, such as those of construction, has often hindered management's goal of achieving increased productivity and job efficiency through implementation of existing analytical modeling methodologies such as simulation and non-linear production optimization.;The studies described in this dissertation used state-of-the-art instrumentation technologies in conjunction with proven artificial intelligence techniques to better implement a non-linear production optimization theory for excavation operations known as the "load-growth curve." The research focused on the use of on-board microprocessor-based sensors for real-time data acquisition from a specific construction machine, the push-loaded earthmoving scraper. Electronic instrumentation and computer data acquisition technologies were further enhanced through data analysis by real-time, rule-based logic.;The study placed particular emphasis on system performance and the ability to manage the desired optimization functions in real-time. Therefore, decision-making logic is implemented in the form of a series of distributed and integrated knowledge-based expert systems. The knowledge and the heuristics used for this problem domain are represented in the form of IF dots THEN production rules, in three separate expert system modules, all sharing data, knowledge, and information via a simple "blackboard" or shared database.;Module 1 is a knowledge-based expert system that determines the suitability and economic applicability of a variety of scraper types for a specific earthmoving task. Module 2 acts as a performance simulator for the machine type recommended by Module 1. Concepts of fleet balance and overall performance issues are also handled by this module. The expert system in Module 3 handles real-time communication with on-board sensors and production optimization, based on machine performance under real-world conditions.;The concept and methodology used in this research were validated using a physical model experiment at Stanford. The model equipped with load sensors for data acquisition was linked to the Exsys expert system shell running in a microcomputer environment for decision-making and real-time feedback.;The integrated approach used in this research identified a number of barriers, both theoretical and practical. These need to be carefully addressed before significant advancements in partial or full automation of field-based construction machines can be achieved. |
