Deadlock-free production using Dempster-Shafer and preset methods in predictive scheduling for multiagent controlled flexible manufacturing systems

Deadlocks can lead to indefinite delays and total production interruption, resulting in industrial losses. The control function is responsible for maintaining deadlock -free production schedules in manufacturing systems that permit product parallelism and resource sharing. Although some predictive scheduling algorithms can set optimal production schedules with parallel tasks for multiple products, they cannot detect or prevent deadlocks. This study proposes two approaches for minimizing the number of deadlocks in predictive production schedules defined by a flexible job -shop scheduling (FJS) algorithm. These approaches were evaluated in a virtual environment operated as a multi -agent controlled flexible manufacturing system (FMS). The first approach adopts the Dempster-Shafer evidence theory with a set acceptability value for the belief function. Agents convert information from different sources in a production system into decision outcomes that indicate the best control case from a case base to drive deadlock -free production. The second approach employs a decision table for product routing in the FMS and three corrective presetting methods to avoid deadlocks in production scheduling. A leaderless consensus is adopted in the proposed approaches to decision processes, ensuring that the shared and parallel activities executed by the multiple agents of the system remain promptly synchronized. In total, 32 FJS schedules containing 10-47 events were tested to evaluate the efficiency of the proposed approach. This paper presents the results of the experiments, an overall evaluation of the approaches, and data on the time complexity of the proposed algorithms. The second approach successfully avoids deadlocks in the FMS; however, both approaches can be adapted to other environments with order -driven production and predictive schedules.