Multimodal Audio Detection Model for Casting and Online Learning Optimization 在线阅读 下载PDF
Author of the article:XU Congyu;WANG Biao;LIU Shanlin;HONG Jun;ZHU Qingbao
Author's Workplace：Hanshan Industrial Research Institute,Hefei University of Technology,Hefei 230009,China;School of Instrument Science and Opto-electronics Engineering,Hefei University of Technology,Hefei 230009,China;Anhui Haili Precision Casting Co.,Ltd.,Maanshan,238101,China
Key Words：casting detection model vibration mode online learning
Abstract：The resonance frequency generated by casting percussion is related to the the change of casting size.It is necessary to predict the spheroidization rate or verminization rate of the casting by resonance frequency.To solve this problem,this paper proposes a multi-mode audio detection model composed of multiple resonant frequencies,and the model coefficients were obtained from the audio detection data through online learning.The results show that the multi-modal audio detection model can effectively identify the change of the elastic modulus of the casting and the change law,and has nothing to do with the the change of casting size.The model coefficient can be further optimized through the improvement of casting quality and online learning.In some types of casting as an example,the measured 20 samples,by multimodal audio detection model,extract the model values to reflect the actual changes in quality of casting products and range,for casting products provide a basis for the online detection and quality classification.
