Separation of Acoustic Signals on a Compressor Using FastICA

Separation of mixed signals has been extensively studied for telecommunications, health, damage analysis, and others. This research focuses on the separation of acoustic signals to support the analysis of engine conditions. The compressor engine is the object of this research. Mixed signals will be separated using FastICA. The acoustic signal from the compressor engine parts is mixed with synthetic mixing and real mixing. Separation results will be evaluated through LSD and SDR values. Two machines with normal bearing condition and bearing fault condition successfully separated well in synthetic mixing with the largest SIR value was 5.60dB and the smallest LSD was 0.510.