This paper describes a new method for diagnosis of reactor cores using noise signal processing
method. Detection and diagnosis of faults are based on the readings of signals coming from a set
of detector tubes, which are surrounded by fuel assemblies in the reactor core. Vibrations of the
detector tubes arise from the strong flow of coolant water in the reactor. If the vibration is large
enough, the detector tube may impact on the nearby fuel assemblies, which in turn will execute a
short, damped oscillation after each hit. This oscillation produces transient in neutron noise signal
that is too small to be obselved. The traditional frequency analysis is not an effective tool for
finding out transient in non-stationary signal and needs a reference or impact-free signal, that is
often impossible to obtain in practical applications. In this work, we suggest a method to detect
impact or transient in the signal, using one time the discrete wavelet transform @WT) and in the
other an adaptive wavelet filter. The adaptive wavelet filter is found to be very effective in
detection of detector vibration and impact problem early before its damage; therefore, the on-line
diagnoses of the reactor will detect this error and if necessary alert the operator. The reactor with
detector tube impact model is developed using Matlab-Simulink. The results show the
effectiveness of the proposed method.