Abstract
In underground mining, there are risks involved in operating heavy machinery such as Load–Haul–Dump and rock breakers. In order to minimize those risks, that can be fatal, it was proposed to study the feasibility of employing both vision and sound to improve either human-based teleoperation or robotized machinery. Earlier attempts have considered only visual input data. In this work, recognition results carried out on acoustic emissions generated by a rock breaker during normal operating activities, are presented. These acoustic data were recorded in-situ and then analyzed using four different techniques: FFT, FFT-Wavelet, Cepstrum, and Time–Frequency. The results indicate that Power Cepstrum analysis provides the best performance both in terms of finding noticeable differences between the signals representing the operating states of the rock breaker as well as in terms of data compression capabilities. A back-propagation neural network was then trained to classify the signals into the different operating states of the machine. [Work partially supported by Project Dicyt 009913SS, University of Santiago, Chile.]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have