Dynamic compensation for an infrared thermometer sensor using least-squares support vector regression (LSSVR) based functional link artificial neural networks (FLANN)

A novel functional link artificial neural network (FLANN) architecture is presented and applied to dynamic compensation for an infrared thermometer sensor. The identification results between a generic FLANN and a least-squares support vector regression (LSSVR) are verified to be similar. A new method to update the FLANN weights is derived from LSSVR. Compared with the generic FLANN, the improved one differs markedly in solving a set of linear equations instead of an iterative problem. As a result, more accurate weight evaluations are obtained, and a faster learning course can be expected. The infrared thermometer sensor dynamic compensator is established based on the principle of inverse model rectification, and the improved FLANN is used to describe the compensator. The actual calibration data of the infrared thermometer uIRt/c are used to validate the feasibility of the present method. The experimental results show that the improved FLANN is faster in training speed, higher in precision and more robust