TH149122A - Devices to reduce data errors from sensors in multi-sensor measurement systems and such processes. - Google Patents

Abstract

DC60 (10/07/57) The process for continuous data from sensors in the monitoring system. It will automatically identify sensors that provide abnormal data. This is done by using a sensor model for each sensor, and estimating values from other sensors. The anomaly detection unit will use the estimated data from the sensor models to find the relationship between the standard deviation and mean of each sensor model data and the sensor data. The analysis will then analyze the change in the relationship between each model data and the sensor data. When the calculated value is lower than the acceptable value, the estimated model is considered abnormal. In the case where the relationship between the data values between the standard deviation and mean of the sensor data and the model data differs beyond the acceptable value, the estimated model is considered abnormal. In the case that the model is identified as abnormal, the processing unit calculates abnormal sensors and replaces the abnormal sensor data with the data from the sensor model with the best relationship. The relationship of each model data with the sensor data. When the calculated value is lower than the acceptable value, the model is considered to be abnormal. In the case where the relationship between the standard deviation and the mean of the sensor data and the model data differs beyond the acceptable value, it means that the model is abnormal. In the case that the model is identified as abnormal, the processing section calculates abnormal sensors and uses the data from the sensor model with the best relationship value to replace the abnormal sensor data.