Control inertia and fuzzy brakes

A new control property namely ldquocontrol inertiardquo is introduced in this article. In this research, control techniques not needing a mathematical model of the system are subject to study. Neuro-predictive (NP) method is a non-model based technique works for a wide variety of nonlinear systems, and lets us compare different systemspsila behaviour. In this paper, two different nonlinear systems, a model helicopter and a tank reactor, are controlled similarly by neuro-predictive method in simulation environment. Although tank reactor is controlled successfully by NP method, a repeated significant overshoot is observed when model helicopter is controlled (leading very long settling time and a considerable amount of energy consumption). This discrepancy in control behaviour is explained by a property of systems, called ldquocontrol inertiardquo. In this paper, control inertia is defined as the ratio of control input to the second temporal derivative of systempsilas output. It is indicated that the undesirable control behaviour of model helicopter (repeated overshoot and its consequences) is influenced by its high control inertia. In order to improve the control behaviour, a fuzzy inference system is designed and added to the control circuit to decelerate system when it is approaching setpoint. This fuzzy inference system is called ldquofuzzy brakerdquo, which improves the performance significantly in case of high inertia. Having a general understanding of systempsilas dynamics (not necessarily a mathematical model), it is possible to judge whether the system is high inertia or low inertia, and whether a fuzzy brake is needed or not. In general, the concept of control inertia can be used in intelligent control system design together with input-output data and fuzzy rules derived by experience/observation.