Identification of Fault Components in Diesel Engine Sounds on Train Using Neural Network

Abstract

The diesel engine damage is a typically found-problem in a train. One of the examples from such damage is the damage on components or known as fault components. Fault components occur because the engine does not attain its maximum performance due to the various causes starting from injector damage to injection pump damage. These damages have different sound characteristics. Based on the differences in sound characteristics, expert engineers might identify and categorize the type of damage within the diesel engine by using condenser microphone installed within 1 meter from the source of the engine sound with HPF and FFT (Fast Fourier Transform). The three sources of sound have different frequency. The normal diesel engine has dominant frequency under 1 kHz while the diesel engine sound with injector damage has the dominant frequency under 2 kHz and the diesel engine sound with injection pump damage has dominant frequency above 2 kHz. The data are gathered and processed using Neural Network with RMSE (Root Mean Square Error) 0.001.