机电工程技术2023，Vol.52Issue(12)：23-26,85,5.DOI:10.3969/j.issn.1009-9492.2023.12.006
基于神经网络的超声波焊接功率预测方法
Power Prediction Method of Ultrasonic Welding Machine Based on Neural Network
摘要
Abstract
Because the ultrasonic welding process has nonlinear time-varying characteristics,the traditional control method is no longer suitable for the accurate control of active power,but the neural network algorithm is more applicable to this scenario because it does not rely on specific mathematical models and can solve the mapping relationship between multiple variables and multiple objects.Therefore,two neural network models to predict ultrasonic welding power are designed and trained with welding data in real engineering.The results show that by using the two neural network models,the ultrasonic welding power can be predict accurately,and the loss of training verification sets of both models can be reduced to less than 10-2.Comparing the two training results,it can also be found that the training verification set of the predictive model based on convolutional neural network has less loss,better prediction effect,stronger model generalization ability and more robustness,and can be used as the preferred algorithm for power prediction in ultrasonic metal welding power control.关键词
超声波焊接/神经网络/预测Key words
ultrasonic welding/neural network/prediction分类
通用工业技术引用本文复制引用
周俊雄,周明欧,范鹏,卢其辉..基于神经网络的超声波焊接功率预测方法[J].机电工程技术,2023,52(12):23-26,85,5.基金项目
国家重点研发计划(2022YFB4702500) （2022YFB4702500）
