Construction Noise Prediction Using Stochastic Deep Learning

Ooi, Wei Chien (2022) Construction Noise Prediction Using Stochastic Deep Learning. Master dissertation/thesis, UTAR.

Preview

PDF Download (8Mb) | Preview

Abstract

Construction noise is an occupational noise that is potentially harmful, and it usually originates from earth-moving machines in construction sites. The impact of construction noise on the health and safety of construction workers is one of the main concerns in the industry. The adverse impacts arising from construction noise may jeopardize public welfare, particularly for those who live nearby the construction site. Therefore, this research aims to develop a reliable noise prediction model on the basis of stochastic modelling and deep learning technique. Stochastic modelling was applied in this study to manipulate the several major parameters such as the randomness of three different duty cycles, coverage angle of the noise receiver, and position of dynamic machinery in the construction site, to generate a set of randomized data as the input for the deep learning model. The deep learning model was trained with stochastic data to predict the noise levels emitted from the construction site. The programming algorithm of stochastic modelling was executed in MATLAB, whereas the deep learning model was established by using Python 3.6 programming language in Spyder. Ten case studies were conducted in this study to validate the predictive performance of the stochastic deep learning noise prediction model. The stochastic deep learning model showed high accuracy of prediction results with an average absolute difference of less than 1.2 dBA having the relative percentage error of less than 4 % among the case studies as compared to the measurement. The reliability of the results from the prediction model was high. In conclusion, the stochastic deep learning model was established and provided a promising outcome with satisfactory predictive performance. Lastly, the model is worthwhile to be further developed to fully exploit the potential of the stochastic deep learning noise prediction model in construction industries as a planning, managerial, and monitoring tool.

Actions (login required)