Cavity Resonance Sound Absorptive Material Parameter Optimization Based Image Recognition and Deep Learning

Traditional modelling and optimization methods for sound absorptive materials require manual testing and huge computation effort. Artificial intelligence is an ideal method for the prediction and identification of complex systems in these cases. This study presents an optimization approach based on image recognition combined with deep learning for cavity resonance sound absorptive material. The sound absorptive material model is conducted using the acoustic-structure coupling model in COMSOL Multiphysics for the sound pressure distribution and behaviour during different frequency excitation. The cavity material plane wave image recognition model was obtained using the CNN method of multiscale layered features. The image arrangement was then processed and sent to the convolutional layer for feature extraction. Subsequently, the feature record was entered into the loop. The association between the kinematic parameters of the cavity material and the sound pressure distribution is obtained through the prediction layer. Using the prediction of CRNN, the results indicate that the range of the cover thickness should be between 50~70 mm and the angle should be between 45~60°. It also showed that this optimization method could be successfully applied for similar materials and components, with certain universality.