Application of Noise Reduction Technique in the Inverse Analysis Using Kalman Filter

Abstract

The inverse techniques usually employ the sensor measurement to estimate the unknown quantities. Regardless of sensor accuracy, the measurements contain some degrees of uncertainty and error, inadvertently. Inasmuch as, the inverse problems are ill-conditioned in general term, the measurement errors cause instabilities, perturbations, and excursions in the solution procedure. To handle the noise difficulties, a novel approach is proposed in the current study. In this method, the measurement errors are filtered to alleviate the noise priori to utilization of inverse method. The Kalman filter is implemented to remove the noise from the original sensor readings. Thereafter, the Levenberg–Marquardt method is implemented to predict the unknown. To evaluate the accuracy and robustness of the developed approach, a high nonlinear test case containing moving boundary heat conduction problem is investigated. Comparing the obtained results illustrates the improvement of inverse solution procedure by employing the noise filtering technique.