Absolute Value Principal Components Analysis (AVPCA) and Parameter Estimation (PE) to bearing fault detection using rotor speed signal monitoring — A comparative study

Abstract

In this paper, a comparative experimental study between the Parameter Estimation (PE) technique and the Absolute Value Principal Component Analysis (AVPCA) algorithm to bearing fault detection using rotor speed signal monitoring is represented. The PE technique relies on the residuals between the input/output (Voltage/Speed) signals of the real system and of the estimated model. AVPCA, in other hand base on the Sum Square Error (SSE) distance between the training-databases and the tested-databases from just only the output signal (Speed) and its minimum. The experimental results reveal that the AVPCA algorithm is more effective in detecting bearing faults than the PE technique using rotor speed signal monitoring.