A Novel Online Error Correction Scheme for Sensors in Aerospace Applications

Sensors that form an integral part of any measurement system are prone to various types of errors, which are contributed by fabrication processes, calibration schemes, operating environment etc. Different types of sensors, which are used in aerospace applications, cannot afford to have such errors as the measurements are very crucial for a successful mission. Many of the error correction schemes are based on modeling of sensor characteristics and are highly complex and non-flexible. Moreover, these schemes are not ideal for online error corrections, which are essential for realtime systems. Artificial Neural networks (ANN) with proper learning strategies prove to be good candidates in such cases. The proposed Meta-cognitive Extreme Learning Machine (McELM) based ANN framework, employed for error correction, automatically captures the non-linear behaviour of sensors to external environmental disturbances like temperature, structural vibration etc. A clustering based thresholding scheme has been used to improve the adaptiveness and stability of the ANN network. The performance of the algorithm has been evaluated on experimental data obtained from pressure sensors used for absolute pressure measurement onboard Indian launch vehicles. The proposed cognition based scheme builds a computationally less complex network with lesser training samples and a simpler hidden neuron layer for sensor error correction.