A modified Auto Associative Kernel Regression method for robust signal reconstruction in nuclear power plant components

The application of the Auto Associative Kernel Regression (AAKR) method to the reconstruction of correlated plant signals is not satisfactory from the point of view of the robustness, i.e. the capability of reconstructing abnormal signals to the values expected in normal conditions. To overtake this limitation, we propose to modify the traditional AAKR method by defining a novel measure of the similarity between the current measurement and the historical patterns. An application of the proposed modified AAKR method to the condition monitoring of a pressurizer of a Pressurized Water Reactor (PWR) Nuclear Power Plant (NPP) shows benefits with respect to the traditional AAKR method, in terms of earlier detection of abnormal conditions and correct identification of the signals responsible for triggering the detection. (Baraldi et al. 2010, Baraldi et al. 2013b). The remainder of the paper is organized as follows. In Section 2, the fault detection problem is introduced. In Section 3, the AAKR method is briefly recalled. Section 4 shows the limitation of the traditional AAKR approach to condition monitoring and states the objectives of the present work. In Section 5, the proposed modification of the traditional AAKR is described and discussed. In Section 6, the application of the proposed method to a case study concerning the monitoring of 6 signals in the pressurizer of a Nuclear Power Plant is presented. Finally, in Section 7 some conclusions are drawn.