Virtual Sensing for Fault Detection within the LUMEN Fuel Turbopump Test Campaign

Abstract

During the operation of a rocket engine and its subsystems, sensors are exposed to high thermal and mechanical loads and are therefore susceptible to potential failure. However, control and health monitoring systems rely on accurate data acquisition of key engine parameters such as the combustion chamber pressure or the rotational speed of the turbopump. In case of sensor degradation, model-based and data-based approaches for sensor validation can serve as a complementary layer to ensure reliable data acquisition. These virtual sensors allow a real-time estimation of a target value based on other process variables. As the target can be rebuilt with different inputs, a network consisting of mathematically redundant virtual sensors is developed in this work. This approach is robust to degradation of physical sensors and allows the detection and isolation of faulty sensors. Therefore, the methodology used in this work offers a cost-effective and powerful approach to ensure continuous health status and performance assessment in case of sensor faults.