Studying Acoustic Behavior of BFRP Laminated Composite in Dual-Chamber Muffler Application Using Deep Learning...

Area of Science:

• Acoustics
• Materials Science
• Computational Engineering

Background:

• Investigating muffler materials' acoustic behavior requires extensive, time-consuming calculations, especially for advanced composites.
• Traditional methods pose challenges for optimizing muffler design with novel materials like basalt fiber reinforced polymer (BFRP).

Purpose of the Study:

• To develop a deep learning-based monitoring system for predicting the acoustic behavior of BFRP composite mufflers.
• To reduce the time and effort required for muffler material selection and design optimization.

Main Methods:

• Developed two deep neural network (DNN) architectures: recurrent neural network with long short-term memory (RNN-LSTM) and convolutional neural network (CNN) in Python.
• Created a dual-chamber laminated composite muffler (DCLCM) model in MATLAB to generate acoustic datasets (e.g., transmission loss, power transmission coefficient).
• Optimized model training parameters using Bayesian genetic algorithms (BGA) and validated results against experimental data.

Main Results:

• Both RNN-LSTM and CNN models achieved over 90% accuracy on test and validation datasets.
• The proposed deep learning approach significantly reduced the time and effort for muffler material selection and design.
• Validated the accuracy and reliability of the predictive technique by comparing with existing experimental results.

Conclusions:

• The developed indirect dual-chamber muffler concept using deep learning is an efficient method for acoustic performance prediction.
• This approach accelerates the selection of muffler materials and optimal designs, benefiting industrial applications.
• Future muffler designs may integrate deep learning algorithms for enhanced performance and efficiency.