Use of Data Analytics to Detect Loose Temperature Sensors in Harsh Environments

Abstract

Monitoring of thermal conditions in process plants can be challenging when sensors are installed for a long time, temperature ranges as well as associated thermal material stresses are large, and when the place of installation cannot be inspected visually to check if the sensor is installed properly. The use case investigated in this paper is the monitoring of pipe surface temperatures in a petrochemical process furnace. The idea we propose is to analyze the statistics of the sensor signals, specifically their standard deviation. As the air flow in the furnace is turbulent, we can assume that the sensor readings are affected by a certain stochastic variation. If the sensor is tightly attached to the pipe, this noise will be attenuated due to the thermal inertia. When the sensor is not firmly fixed, the attenuation effect will be smaller. In this paper, we test this hypothesis based on two thermowell sensors, finding that the statistics of the sensor signals do allow conclusions about the thermal coupling. This can significantly facilitate in-situ monitoring of sensors in inaccessible environments.