Prediction of noise using artificial neural networks modeling and statistical methods in the woodworking industry

ahmadi, saeid and salehian, fatemeh (2021) Prediction of noise using artificial neural networks modeling and statistical methods in the woodworking industry. Masters thesis, Gazvin University of Medical Sciences.Ghazvin.Iran.

[img] PDF
Download (164kB)


Introduction: Noise pollution is one of the most important pollutants in the work environment and is almost one of the harmful factors for workers' health. Sound prediction is one of the important aspects of sound control in industries. Forecasting is important in the carpentry industry, which is an important part of the woodworking industry and workers are exposed to excessive noise. There are many methods for predicting noise. Materials and Methods: This research is a study Descriptive, analytical - cross-sectional that was carried out in 6 main phases, which include: 1- Identifying and collecting data 2- Determining the evaluation criteria of statistical models and artificial neural networks 3- Constructing multiple regression 4- Implementing artificial neural networks 5- Optimizing the weights of artificial neural networks It is model sensitivity analysis with genetic algorithm. In the first stage, data was collected from 375 carpentry workshops in Tehran province, Khavaran, Chahardangeh, Nematabad and Delavaran industrial towns. From the 10 main characteristics of acoustic, structural and carpentry processes that affect sound, in the next step, evaluation criteria were presented for comparison and accuracy of both statistical models and artificial neural networks. Then statistical analysis of multiple regressions was done. Then, artificial neural network modeling was implemented with the help of MATLAB software. In the next step, the weights of artificial neural networks were optimized using the genetic algorithm, then the sensitivity analysis of the model was performed using calculations. Discussion: With the help of evaluation criteria, two models of artificial neural networks and statistical methods were compared. The results showed that artificial neural networks provide more accurate prediction than multiple regression. The best neural network can accurately predict the equivalent sound level, our results showed that the developed experimental methods can be a useful tool for the analysis of noise pollution and enable occupational health professionals to use these methods. Conclusion: The artificial neural network model showed higher accuracy compared to linear and non-linear regression statistical models. In this study, the artificial neural network was trained 13,000 times by the gradient descent algorithm, which showed higher accuracy compared to similar studies where the repetition rate of the training algorithm was much lower, so this study showed that by increasing the repetition, the prediction accuracy can be increased. . Finally, a graphical user interface program was presented using factors affecting sound to predict noise in the woodworking industry. Key word: Sound prediction, artificial neural networks, wood industry, sound exposure

Item Type: Thesis (Masters)
Subjects: R Medicine > RN Health > RN1003 Occupational Health
Divisions: University Thesis > Faculty of Health > Professional Health
Depositing User: Health School Students
Date Deposited: 25 Dec 2022 05:55
Last Modified: 25 Dec 2022 05:55
URI: http://eprints.qums.ac.ir/id/eprint/12247

Actions (login required)

View Item View Item