Estimación de ruido atenuado con paredes utilizando redes neuronales artificiales
Nowadays, noise significantly impacts people’s quality of life, influencing daily comfort and posing a risk to auditory health. Controlling sound pressure levels in workplace environments with attenuation walls involves engineering designs that require high costs, long construction times, and potent...
Gardado en:
| Autor Principal: | |
|---|---|
| Formato: | bachelorThesis |
| Idioma: | spa |
| Publicado: |
2024
|
| Subjects: | |
| Acceso en liña: | http://dspace.unach.edu.ec/handle/51000/14408 |
| Tags: |
Engadir etiqueta
Sen Etiquetas, Sexa o primeiro en etiquetar este rexistro!
|
Engadir etiqueta | Summary: | Nowadays, noise significantly impacts people’s quality of life, influencing daily comfort and posing a risk to auditory health. Controlling sound pressure levels in workplace environments with attenuation walls involves engineering designs that require high costs, long construction times, and potential failure risks after fabrication. This study aimed to estimate the effectiveness of airborne noise attenuation in an impedance box with attenuation walls using Artificial Neural Networks (ANN). The researcher created an impedance box consisting of an emitter and a receiver. It was essential to traditionally evaluate the noise attenuation in the impedance box following ISO 9612:2009 standards for field sampling and IEC 61252 standards for equipment. Subsequently, the effectiveness of an ANN with a Backpropagation Algorithm (ANN-BP) was simulated in MATLAB®, using a matrix of input and output variables based on accurate attenuation data. The impedance box contained melamine wood and gypsum-cork materials for attenuation. The input and output variable matrix considered three dimensions: a) construction materials, b) thermo-hygrometric conditions, and c) noise sampling data distributed across 21 elements. The developed ANN-BP consists of four layers: a) the first input layer with 20 static receptor neurons, b) a second layer with nine hidden neurons, c) a third layer with one dynamic neuron, and d) a final output layer with one static neuron. The average error percentage in estimating noise attenuation with walls using the ANN-BP was 0.1477%. The independent samples T-test demonstrated that the ANN-BP generates similar results for estimating noise attenuation compared to the traditional method. |
|---|