On the Influence of Sensor Dynamics in Estimation of Reactor Noise Signals

The qualities of measured signals are always affected by the measurement systems and noise. If these effects can be modelled, the original signals can be restored. Two methods for estimation of signals are studied in this work, the causal Wiener deconvolution filter and the Kalman filter. The methods are first applied to two theoretical test cases and then to a pressure signal, recorded during a reactor process. The measurement system works basically as a lowpass filter, and therefore, the high frequency contents of the measured signal consist mostly of uninteresting noise. Because of this, it will not be possible to restore the true high frequency properties of the signal. However, most of the interesting information is in the low frequency range. One conclusion from this work is that some peaks that may indicate interesting dynamics can appear more distinctly in the power spectrum of the restored signal than in that of the measured signal. The methods recover the signals in the time domain, which is useful e.g. when physical models between different quantities are analysed. Finally, the Wiener filter is more straightforward to use in this application, than the Kalman filter is