Proposal of a quality index applied to fault detection method in electrical valves

Abstract

In modern Intelligent Maintenance Systems, the machine or equipment robustness also depends on its capability to automatically generate reliability and safety reports. This paper describes an approach to autonomously identify if a faulty signal report has been correctly classified. The proposed approach builds on our previous experience in developing embedded intelligent maintenance systems and helps in avoiding the occurrence of “false positive” interpretations, which means, when the maintenance system indicates a possible fault that does not occur. This index would be useful for real-time monitoring and evaluation on fault detection systems, taking into account several degradation model characteristics. In order to validate the proposed methodology a test bench was developed in a lab reproducing some common faults and degradation processes that may occur in the field. The proposed approach makes use of a data acquisition equipment to store information from sensors to monitor specific physical variables from mechanical components such as gears. A test sequence is applied to the valve control actuators with the following steps: a few seconds of faulty free operational cycle sensor data (which means the opening and closing operations are executing without failure) are collected and then a faulty system behavior is emulated changing some mechanical actuator parts to faulty ones. In the faulty emulation case, a malfunction event must be identified and reported by the fault detection system. The preliminary results indicate that the index is extremely useful especially when the degradation stage of a system is below, for example, catastrophic failure or a predefined level.