The aim of this master´s thesis is to investigate if it is possible to extract more dynamic information out of physical signals from nuclear reactor noise measurements than what is possible today. This was achieved by investigating methods to examine and determine the process signal quality, and studying the corresponding transfer function.
Accurate measurements of for example the core stability and the control system interference are required for detailed process diagnostics. By analysing real reactor signals, (here neutron flux and reactor pressure), it is observed that they are correlated. Because the structure of the real system is not perfectly known, two hypothesis have been made regarding the real system.
Identification of the transfer function of the two simulated systems have been done using Matlab with process noise added to the system, with measurement noise added to the system, and with feedback added to the system. The identification models ARX (Auto Regression Moving Average), AR (a special case of ARX) and BJ (Box-Jenkin) have been used.
From the results, it follows difficult to adapt a good transfer function using the ARX model to data. This is because of bad coherence. When identifying the transfer function using a spectrum, an AR model, a good approximation was seen, since the approximation does agree well with the spectral estimate. Here the input is not used. When identifying using an uncorrelated noise vector as input, we get a bias in the approximation, since the output can not be fully explained by the input signal.
Source: KTH
Author: Andersson, Marika
>> Signal Processing Project Materials and PDf using Matlab
>> More Matlab Projects on Signals and Systems for Students
>> 200+ Matlab Projects based on Control System for Final Year Students