부산대학교 도서관

Alarm systems, as a layer of protection in processes and industrial complexes, have attracted much attention in both industry and academia. A simple yet extremely common approach to monitor processes is to raise an alarm when the value of a process variable exceeds its predetermined limits. This approach, however, is not effective when an abnormality or fault changes the variance of a process variable, and not necessarily its mean value. In this paper, a method based on generalized delay-timer is proposed to detect such changes. Subsequently, an algorithm is developed for optimization of the generalized delay-timer parameters based on the properties of the process data. Simulation results confirm the effectiveness of the proposed methods.