RPG Student List


Hongyi QU

Doctor Candidate in Chemical and Biomolecular Engineering

Supervisor: Prof. Furong GAO


Research Topic

Process control and optimization: from single agent to multi-agent cooperation for performances enhancement


Process control and optimization often work in dual in such commonly seen scenario: the lower level process control ensures that process output follows a given set-point or set profile generated according to certain economic targets by upper level real-time optimization (RTO). Nowadays, the parallel manufacturing, which means that there are many manufacturing machines (processes) operating simultaneously and in parallel to produce products, has been an inexorable trend in industrial applications as the fast development of large-scale manufacturing. This provides good chances for automation philosophy extension from traditional ‘self-learning’ to multi-agent ‘peers learning’ or inter-agent learning which makes multiple agents learn from each other for enhancement of individual performances. However, compared with the relatively mature developed multi-agent parallel RTO on optimization level, inter-agent learning in process control can be barely seen.

This research focuses on one of the common scenario in parallel manufacturing: similar types of (even identical) processes are running under the same/similar environment (and in a close neighborhood). Two problems are considered. The first one is how to develop the inter-agent learning mechanisms embedded in adaptive control methodology, which is considered to be the most suitable supporter for this mechanism. The second one is how to integrate multi-agent parallel RTO and the developed inter-agent learning adaptive control. For the first problem, the inter-agent learning mechanisms are sorted as two types: preferential learning and accumulating learning. Both of them will be embedded into the online estimators of adaptive control. Model free adaptive control (MFAC) will be used as application example of above mechanisms. For the second problem, the global searching algorithm (GSA) based parallel RTO is used in optimization level, and the unavoidable topics brought by the conflict between the basic assumptions of RTO and the natural of process control will be investigated. Above two problems will be studied under the cases of continuous process and batch process respectively.

Journal Publication

Qu, Hongyi; Zhang, Ridong; Gao, Furong, "‘State Déjà vu’ Inter-Agent Learning Adaptive Control Framework", (2016)