Designing a fuzzy Q-learning multi-agent quality control system for a continuous chemical production line – A case study

• Designing a new MAS for the quality control of the continuous chemical production line.
• Applying the Q-learning method to equip the proposed FQL-MAQCS with a self-learning mechanism.
• Utilizing fuzzy rule-based and knowledge-based systems jointly to accelerate the quality control process.
• Testing the proposed FQL-MAQCS through the application in a real case study.

In continuous chemical production lines, final products are produced through irreversible and continuous reactions in successive departments. So, the quality of the final products depends on the quality of the input materials and departments’ situation. Any out of the control situation along a production line leads to operation time, resource and financial lost and spoiled materials. Undoubtedly, decreasing the controlling communication time between different departments and handling unexpected conditions are critical issues in such production lines. In this paper, a Fuzzy Q-learning Multi-Agent Quality Control System (FQL-MAQCS) is proposed to control a continuous chemical production line. The proposed FQL-MAQCS can control unexpected conditions by applying a multi-agent based system consisting of the quality control executor, process data analyst, central decision maker, departmental decision maker, and knowledge and rule manager agents. In addition, the proposed system is equipped with a self-learning mechanism, whose knowledge is gradually formed based on the results of the learning process and stored in the knowledge base. To this end, the fuzzy Q-learning method and rule production mechanism are used. The proposed FQL-MAQCS is tested in a real case study and the associated results demonstrate the usefulness and capability of the developed quality control system.