Fault Diagnosis of Water Pump Based on Acoustic Emission Signal Using Fast Fourier Transform Technique and Fuzzy Logic Inference

Acoustic emission technique was used to examine four different water pumps in order to monitor the condition. The audio signal from a microphone was captured for the following conditions: normal pump, unbalance, misalignment, and bearing fault. The data were recorded 20 times for each pump. The sampling frequency used was 48 kHz and a measured time duration was 5 s. To execute damaged pattern classification, Fuzzy Inference System was applied and processed data extraction in time and frequency domain. The features in time domain were extracted from audio signal into several parameters, for example Root Mean Square (RMS), Kurtosis, Crest-Factor, Shape-Factor, and Skewness. Meanwhile, frequency domain data was extracted into instantaneous frequency parameter using the Fast Fourier Transform (FFT) approach. The experimental results showed that the classification accuracy yielded 90%. Therefore, the usage of FIS in acoustic emission analysis could potentially detect different fault categories.