Real-Time Expert Systems for Process Supervision

Abstract

The way of controlling and supervising industrial processes has been notably changing during the last decade. The technology and the equipment of old control rooms with their long control panels have been replaced by computers and video screens. The operator’s task is also evolving. In the past, he was directly manipulating the different means of process control; this was essentially a manual job and required a thorough knowledge of the process behaviour. Today, these tasks are automated and integrated in a computerized system, and the operator has become a supervisor. In general, operators are satisfied with their new assignment in that they find their work more interesting, more motivating and more rewarding. But they also agree that it is more demanding. Their responsibility is increasing — they supervise a larger geographical area — but above all, their hands-on experience concerning the process has diminished: it has moved from partial but detailed knowledge to global but diffuse knowledge [1]. In normal operation, this has no notable consequence, but in the case of a crippling situation, intuition and know-how become of prime importance once again. Misinterpretation of abnormal situations may lead to out-of-specification production or, at worse, to a threat to the safety of people. Paradoxically, a solution to this problem may again come from ... computerization. Indeed, it is precisely such cases that open the path to artificial intelligence techniques and particularly to expert systems (or more widely knowledge-based systems) technology for developing diagnosis and decision support tools that can help plant staff to understand unusual situations and provide them with proper guidance. The interest in this technology lies in the fact that it can take into account the human aspects of the problem by its ability to explain situations and adapt itself to a specific context.