基于深度学习的轮胎花纹噪声预报方法研究  

Tire Pattern Noise Prediction Based on Deep Learning

作　　者：项大兵 张春生[2] 王红彦[2] 侯丹丹 XIANG Dabing;ZHANG Chunsheng;WANG Hongyan;HOU Dandan(E-Rubber Technology(Beijing)Co.,Ltd,Beijing 100084,China;Zhongce Rubber Co.,Ltid,Hangzhou 310018,China)

机构地区：[1]易瑞博科技(北京)有限公司,北京100084 [2]中策橡胶集团有限公司,浙江杭州310018

出　　处：《轮胎工业》2019年第8期503-506,共4页Tire Industry

摘　　要：在进行大量不同花纹轮胎通过噪声试验基础上,利用图像识别和深度学习方法建立了轮胎花纹几何结构与轮胎通过噪声试验结果之间的数学模型,并采用该模型对新设计的轮胎花纹进行通过噪声预测。结果表明:以1 dB声压级为预测精度时,模型预测准确度达到83%;利用花纹图像识别和深度学习理论所建立的轮胎噪声预报数学模型方便可靠,而且随着试验数据的不断积累和算法的改进,基于该方法的预报模型精度会进一步提高。Based on a large number of tire noise testing with various tread pattern,a mathematical model between the tire pattern geometry and tire noise level was developed using image recognition and deep learning technology,and the noise level of the new tire pattern was predicted by this model. The results showed that the model prediction accuracy reached 83% when the resolution was 1 dB. This model was reliable and convenient for application,and the prediction accuracy would be constantly improved with the accumulation of more testing data and improvement of the algorithm.

关 键 词：轮胎噪声预报 花纹识别 深度学习 神经网络 

分 类 号：TQ336.1[化学工程—橡胶工业]