Estimation, Analysis and Smoothing of Self-Similar Network Induced Delays in Feedback Control of Nuclear Reactors

Abstract

Abstract-This paper analyzes a nuclear reactor power signal that suffers from network induced random delays in the shared data network while being fed-back to the Reactor Regulating System (RRS). A detailed study is carried out to investigate the self similarity of random delay dynamics due to the network traffic in shared medium. The fractionality or selfsimilarity in the network induced delay that corrupts the measured power signal coming from Self Powered Neutron Detectors (SPND) is estimated and analyzed. As any fractional order randomness is intrinsically different from conventional Gaussian kind of randomness, these delay dynamics need to be handled efficiently, before reaching the controller within the RRS. An attempt has been made to minimize the effect of the randomness in the reactor power transient data with few classes of smoothing filters. The performance measure of the smoothers with fractional order noise consideration is also investigated into.

Fractional order fuzzy control of nuclear reactor power with thermal-hydraulic effects in the presence of random network induced delay and sensor noise having long range dependence

“…This typical case is known as non-Gaussianα -stable distribution of the random variable which makes the mathematically tractable solutions for suppression of such stochastic phenomena with fractional order characteristics quite difficult [43]. Figure 7 shows that the run-time variance of such spiky random variables (packet delays) does not converge to a finite value and have a non-Gaussianα -stable distribution [42].…”

“…The random network induced delay consideration has been motivated from the study reported in [42]. Majumder et al [42] used various smoothing filters for removing the stochastic nature of the power signal which is continuously fed back from the reactor house to the RRS through a long communication medium.…”

“…For case study, a heavily loaded LAN data have been collected as shown in Figure 6 for estimation of the degree of selfsimilarity indicated by its associated Hurst parameter. Some typical characteristics of such network induced random delays in a loaded network have been discussed by Majumder et al [42]. Detailed study regarding statistical characterization of delays in heavily loaded network has been done by Tejado et al [30], [31].…”

Signal de-noising methods for fault diagnosis and troubleshooting at CANDU ® stations

“…De-noising detector data using different types of wavelets was carried out for a particular case of Self Powered Neutron Detector (SPND) where the signal was presumed to be corrupted (Majumder et al, 2011a). Minimizing the effect of randomness in the reactor power transient data by using different classes of smoothing was investigated in feedback control of a large Pressurized Heavy Water Reactor (PHWR) (Majumder et al, 2011b(Majumder et al, , 2012 along with other work related to investigation and minimization of sensor noise with respect to reactor control (Das et al, 2011(Das et al, , 2013.…”

“…For cases when the smooth width is very large or signal distortion is highly undesirable, the Savitzky-Golay (Majumder et al, 2011b) algorithm can be applied. This method is based on the principle of least-square fitting of polynomials to segments of the data.…”