A condition-based failure-prediction and processing-scheme for preventive maintenance

Abstract

Preventive maintenance (PM) is an effective approach for reliability enhancement. Time-based and condition-based maintenance are two major approaches for PM. In contrast, condition-based maintenance can be a better and more cost-effective type of maintenance than time-based maintenance. However, irrespective of the approach adopted for PM, whether a failure can be detected early or even predicted is the key point. This paper presents a failure prediction method for PM by state estimation using the Kalman filter. To improve preventive maintenance, this study uses a hybrid Petri-net modeling method coupled with fault-tree analysis and Kalman filtering to perform failure prediction and processing. A Petri net arrangement, viz, early failure detection and isolation arrangement (EFDIA), is used; it facilitates alarm, early failure detection, fault isolation, event count, system-state description, and automatic shutdown or regulation. These functions are very useful for health-monitoring and preventive-maintenance of a system. This study implements EFDIA to an application-specific integrated circuit on a Xilinx Demonstration Board. A condition-monitoring system of a thermal power plant is used as an example to demonstrate the proposed scheme. Linking the Kalman filter to the EFDIA Petri net, a condition-based failure prediction and processing scheme has been completed for preventive maintenance. This paper presents a failure prediction and processing scheme for PM via the thermal power-plant example, by using a hybrid Petri net modeling method endowed with fault-tree analysis and Kalman filtering. The FPN (Petri net dealing with system failure) has to be constructed beforehand. The next step is to obtain control charts for all fault places in the FPN in order to prescribe thresholds and increment times for every step in Kalman prediction. Afterwards, the system model of each place in the FPN must be derived to perform Kalman filtering. With these prerequisites, this method can be applied to any system. The proposed Petri net approach not only can achieve early failure detection and isolation for fault diagnosis but also facilitates event count, system state description, and automatic shutdown or regulation. These capabilities are very useful for health monitoring and PM of a system. Since the triggering signal of S/sub i/ place of the EFDIA in Section IV (S/sub i/ is a place for the Kalman-predicted indicator value of the sensing signal for the Petri net dealing with system failure) indicates that subsystem #i performance is going to reach the prescribed failure threshold, the signal can be provided via the Kalman filtering method in Section III. Linking the Kalman filter to the EFDIA Petri net, a condition-based failure prediction and processing scheme has been completed for preventive maintenance.