Robust signal reconstruction for condition monitoring of industrial components via a modified Auto Associative Kernel Regression method

• We addressed the problem of detecting abnormal conditions in industrial components.
• We propose a modification of the AutoAssociative Kernel Regression method.
• The new method allows to provide an earlier detection.
• The method is tested on real data collected from an energy production plant.

In this work, we propose a modification of the traditional Auto Associative Kernel Regression (AAKR) method which enhances the signal reconstruction robustness, i.e., the capability of reconstructing abnormal signals to the values expected in normal conditions. The modification is based on the definition of a new procedure for the computation of the similarity between the present measurements and the historical patterns used to perform the signal reconstructions. The underlying conjecture for this is that malfunctions causing variations of a small number of signals are more frequent than those causing variations of a large number of signals. The proposed method has been applied to real normal condition data collected in an industrial plant for energy production. Its performance has been verified considering synthetic and real malfunctioning. The obtained results show an improvement in the early detection of abnormal conditions and the correct identification of the signals responsible of triggering the detection.