Adaptive control for safe and quality rebar fabrication

Rebar fabrication is a labor intensive operation that uses scrap or "trash" steel For raw materials and therefore can benefit greatly from improvement s in safety, productivity, and quality. Shared control through a human-mach ine interface may be the best alternative for achieving highest quality sta ndards and improving worker performance in safety and productivity. This pa per develops a control scheme for automated rebar bending within the framew ork of computer integrated construction and presents research focused on th e task level control to compensate for springback in the bent rebar. Three major problems are addressed: (1) Conception of a hierarchical computer int egrated construction control structure that links rebar Fabrication to the other construction project functions: (2) comparative evaluation of alterna tive algorithms For prediction of springback; and (3) portability of a spri ngback control model that uses real-time electronic sensing. Bending tests were conducted with both a laboratory prototype and an actual shop table be nder to experiment with alternative models for in-process springback predic tion including a neural network model. Limitations in control system portab ility were realized in the transfer from the laboratory prototype bender to the shop bender. Springback model evaluations revealed that empirical stat istical models, neural networks, and in-process relaxation performed equall y well.