LOAD DISTRIBUTION AMONG REAL-TIME CONTROL COMPUTERS - MULTIPROCESSOR CONTROL OF TANDEM ROLLING MILLS.

This paper concerns the application of a multiprocessor system to the supervisory control of a hot strip mill line. The computing load is decomposed into a set of small processes called tasks and the precedence relations among them is represented by a task graph. The parallel processing of a set of tasks on a multiprocessor system can be executed effectively on the basis of the static optimal or approximate scheduling algorithms developed by the authors. In the hot strip mill line control, dynamic scheduling scheme is required because the task set is generated repeatedly upon the arrival of a slab. This paper proposes both fixed and variable priority assignment methods for dynamic task scheduling. The proposed methods are evaluated and compared on a software simulator. The test results reveal that some of the proposed methods can improve markedly the responsiveness of the mill control system.