Hardware Implementation of Artificial Neural Networks for Vibroacoustic Signals Classification

Abstract

This paper studies the architecture of a neural classifier designed to identify technical condition of machines, based on vibroacoustic signals. The designed neural network is optimized for implementation on Field Programmable Gate Arrays (FPGA) programmable devices. FPGA allows massive parallelism and thus real-time classification as each neuron can execute arithmetic operations simultaneously. The classifier of vibroacoustic signals was designed and tested for the self -organized neural network. The teaching vectors are based on estimates derived from processed vibroacoustic signals generated by rotary machines. The created classifier was applied for recognizing technical state of demonstrative toothed gear DMA1 in variable operating conditions.

Neural net implementation of steam properties on FPGA

“…Recognition of digits using a network with 300 inputs and 10 outputs with a single neuron is reported by Latino et al [27]. A method of configurable MLP with a single neuron block with floating point add and multiply units along with activation function as LUT for a smart position sensor of solar panels has been studied by Dąbrowski et al [28]. A neural classifier with fixed point representation and 12-bits for detecting damaged toothed gears using vibroacoustic signals is highlighted by Polat and Yildirim [29].…”

A signal pre-processing algorithm designed for the needs of hardware implementation of neural classifiers used in condition monitoring

“…In this paper, the signal pre-processing algorithm allowing to calculate the technical state vectors for a planetary gearbox working in non-stationary operation is proposed, and it is dedicated for hardware implementation on FPGAs [19][20][21]. The purpose of the algorithm is to estimate features of the vibration signal that are related to the investigated failures, then these features can serve as components of an input vector for a neural network.…”