关键词: 数字孪生, 扰动识别, 卷积神经网络, 云—雾—边缘协同

Abstract: Modeling and simulation has become one of the core technologies in manufacturing industry.However,the existing simulation technology is insufficient for the dynamic response of the manufacturing process.Digital twin provides a solution for dynamic mapping of the simulation system to manufacturing process.At present,there are still existing the following problems: ①it is difficult to recognize dynamic disturbances in the process of equipment operation,and the training process of disturbance identification model is limited by computing and data resources,which is time-consuming;②data utilized for identification of dynamic disturbance and update of constraints are offline or semi-offline,which can not meet the requirements of real-time response.To solve the above problems,a method for responding to dynamic disturbances in digital twin manufacturing simulation process based on cloud-fog-edge collaboration was proposed.In the cloud,a universal model was trained by public data sets,and a personalized model further was trained by edge data to improve the accuracy of disturbance recognition.In the fog,the personalized model was deployed to ensure the speed of disturbance recognition,while the recognized disturbance was updated to the digital twin simulation model.In the edge end,real-time signals were collected and uploaded to the fog end.The experimental results showed that the disturbance constraint updating mechanism could accurately update the disturbance and make a quick response to the disturbance during operation.

Key words: digital twin, disturbance identification, convolution neural network, cloud-fog-edge collaboration