Time-Frequency Independent Component Analysis for Multi-Damage Detection on a Rotating Machine

Maintenance of the plant machinery plays a crucial factor for maintaining continuity of industrial processes. This paper reported a development of an acoustic-emission-based (A E) technique of identifying multi-damage to the machine remotely using two sensors. In implementation, we emphasized in the separation of sound signals emitted by multiple machines using Time Frequency Independent Component Analysis (TFICA) recorded with a microphone array using the technique of mixing assuming a single source. Overall this study aimed to identify the unbalance, misalignment and bearing faults. Each machine had simultaneously two different damages, i.e. the bearing fault and unbalance, unbalance and misalignment, and bearing fault with misalignment. Separation process was performed using several types of techniques, namely, time domain ICA, frequency domain ICA. Observations FDICA superior in separation rather than TDICA with high MSE values are: 1.7 × 10^-5. From the experimental results showed that the distance between the microphone so the shorter the distance the smaller the spatial aliasing occurs.