Deterministic-stochastic Subspace Identification for Filtering of X-ray Analyzer Measurements

O. Ojala, H. Hyƶtyniemi, and J. Miettunen (Finland)


Filtering, subspace identification, dynamic models, Kalman filter, measurement noise


In this paper, multivariate data analysis is applied to filter measurements of mineral contents in an X-ray analyzer. Traditionally, filtering is carried out by exponential smoothing; however, this way the signals are always delayed and cannot react rapidly to the changes in the measurements. To enhance this situation, deterministic stochastic subspace identification algorithm N4SID is applied to construct an optimized Kalman filter using measurements from three channels (Zn, Cu, Fe) of the analyzer. Models are constructed using incremental variables. The Kalman filter models and first order exponential smoothed measurements are compared and ideas for future research are introduced.

Important Links:

Go Back