Fuzzy rule-based human fault avoidance method under Grafcet

This paper describes a real time algorithm to be implemented by Grafcet on a PLC designated to avoid plant damages or system failure during plant operation due to interactions between human operator and plant, that is to avoid incorrect human decision and/or human actions. Every human action on the plant as an input command is supervised by means of the proposed algorithm. If man-plant interaction is correct, then, operation is performed according human demand. If interaction between operator and plant is not safe, then such operation is blocked, indicating the state of the components of the plant that do not satisfy safety conditions. It means that any human interaction with system will be only possible if such command is proper to the actual plant stage, that is, if such action do not cause damages according rule based criteria designed off-line by expert human operators. Fuzzy modeling and control is a technique to handling qualitative information in a formal, rigorous and efficient way if such information are properly used. The author presents an alternative method to control man-machine interfacing processes where fuzzy logic are to be applied. The developed algorithm for decision on the man-machine interface blocking avoiding faults and detecting man-machine incorrect actions, is particularly emphasized. Such algorithm exploits the knowledge of a rule-base created under fuzzy ideas to be defuzzified by means of the linearisation procedure by method of small perturbation theory of a polynomial function achieved by a multi-dimensional least-squared method and alternatively by Lagrange interpolation.<>