Optimal Sensor Placement for Efficient Fault Diagnosis in Condition Monitoring Process; A Case Study on Steam Turbine Monitoring

摘要

Failure root cause analysis requires an optimum sensor network in the process of a complex system monitoring. Selection of the location, type and number of sensors are important metrics of sensor network optimization. Main aspects of this optimization can be categorized to failure detection, failures diagnosis from each other, the collected data from sensors and sensor reliability. In the process of sensor networks optimization, logical relationships are determined between components and sub-systems through different methods such as FMEA, FTA and RBD. In this paper, an augmented FMEA and FTA method is developed to extract for predicting failure causes in a condition monitoring process. The potential location of sensors is first determined through Sensor Placement Index (SPI). SPI depends on the Importance of failure modes and the cost of their monitoring processes. Due to the potential places of sensors, different scenarios are derived for sensor placement. Considering prior information about component state (operational or failed), system is simulated through Bays Monte Carlo method. By estimation of sensor detection probability, posterior probability of failure modes is calculated. Then the variance of proposed probabilities is added together and the result represents the uncertainty index. For determining the sensor reliability index, sensors are considered as system components. In this case, functional model of each scenario is developed and the scenario with less Top Event probability is selected as the optimal one. The main purpose of this paper is to show the difference between prioritization of scenarios based on two proposed criterion. It represents that both the uncertainty and reliability of sensors must be considered in the optimization process. But in some specific cases such high-reliable systems, the effect of sensor reliability index can be negligible. As a case study, optimization of sensor placement has been demonstrated on steam turbine and results are discussed.