Machine deterioration identification for multiple nature of faults based on autoregressive-approximate entropy approach

Abstract

This paper presents a new methodology to monitor the condition of system due to the presence of multiple faults. The methodology is based upon autoregressive and approximate entropy approach through which intensity of deterioration on a healthy system can be effectively envisaged in its nascent stage. The healthy state of the system is altered by introducing a point defect in outer race of ball bearing and then condition of the system is further aggravated by shaft misalignment. After fitting an autoregressive model on the denoised signal, a statistical parameter known as approximate entropy is employed to assess the randomness of the signal. To establish the effectiveness of the proposed methodology the results are also computed without employing autoregressive-approximate entropy approach. The proposed methodology shows an increase of 14.3% in the approximate entropy as the condition of the system deteriorates, whereas only 0.3% increase in approximate entropy has been obtained without the proposed approach.