Prediction Error Adaptive Kalman Filter for Spectral Measurement Correction

This technical report presents a study on the application of Prediction Error Adaptive Kalman Filters for online spectral measurement correction and concentration estimation. The authors introduce an integrated discrete-time nonlinear model that considers dynamic aspects of the process and a physics-based sensor model. The study highlights the sensitivity of spectral measurements to external factors, such as temperature and pressure, which can affect measurement precision. Two alternative Prediction Error Adaptive Kalman Filters are proposed to estimate concentrations and sensor model parameters. A simulation of a ternary mixing process is conducted to compare the performance of the proposed filters against a standard Extended Kalman Filter. Results indicate that the adaptive filters can estimate concentrations and sensor parameters with minimal error, even under varying temperature conditions and measurement noise. The report outlines the model description, estimation problem, and the algorithms used, concluding with a discussion on future work in this area.