Design of a Performance-Driven Control System Using a Support Vector Machine

Abstract

In control system, it is necessary to maintain a high control performance. However, it is difficult to maintain the performance by retirement of skilled workers and loss of their knowhow. Therefore, from the perspective of green innovation, the problem must be solved by automatic control. In recent years, several researches of control performance assessment have been proposed. A performance-driven controller that redesigns the controller when detecting that the control performance is deteriorated has been proposed. This method is based on a result of the control performance assessment, however, it does not consider a deterioration cause of the control performance. Here, the control performance deterioration caused by disturbance and/or system variant is considered as an example. When the reason of a poor control performance is caused by disturbance, the incorrect parameter adjustment may lead more poor control performance, eventually, the control system may fall into an unstable state. In this paper, the reason of the control performance deterioration is detected by Support Vector Machine in advance, and the controller is redesigned only when the control performance deterioration is caused by system variant. Thanks to this strategy, the incorrect parameter adjustments are avoided, and a high control performance can be maintained.