Application of ANN for the Identification of Discriminating Variables by Noise Analysis on Portable Motorised Equipment

Noise exposure is one of the main causes of occupational illnesses in agriculture. Typical machines exposing operators to this risk are motorised portable equipment, especially for hobby farmers. This work aims at analysing noise data emitted by this kind of machines and identifying key discriminatory variables, through an artificial neural network (ANN). This study allows a ‘diversified’ noise risk analysis for each type of machine, enabling the use of the most appropriate personal protective devices according to the different types of equipment. Using a class 1 Larson Davis 824 integrating sound level meter the noise emissions of 153 motorised portable equipment were analysed. Measurements involved more than 30 farms. The types of equipment used in different work operations were 55 brushcutters, 52 chainsaws, 17 hedge trimmers, 23 blowers and 6 olive harvesters. A classification model based on artificial neural networks (ANN) was implemented. This made it possible to investigate discriminating variables on noise emitted by different types of equipment. The analysis performed allows the identification of different operating machines through the emitted noise. Moreover, thanks to an accurate variable impact analysis it is possible to address their specific impact on the operator's health. Finally, by evaluating the noise emission of each type of equipment at different frequencies, it is possible to identify those responsible for the highest risks. Such investigations could provide the opportunity for the appropriate selection of personal hearing protection equipment, without the need for specific in-field measurements.