Digital twin dynamic modeling method based on real-time driving of multi-source video stream

Abstract

The present application relates to the technical field of digital twin modeling, in particular to a digital twin dynamic modeling method based on real-time driving of multi-source video streams, comprising the following steps: multi-source data synchronous acquisition and standardized processing; dual-source feature extraction, fusion modeling and maintenance instruction generation; twin model incremental updating and state calibration; abnormal alarm binding and visual display.The present application relates to the field of industrial equipment monitoring, and through FPGA timestamp hard synchronization acquisition of visual and vibration and temperature sensor data, distortion correction and noise reduction, the problems of single data dimension and poor acquisition quality are solved.Improved YOLOv8 algorithm containing ECA mechanism and dynamic anchor frame and multi-scale optical flow method are used to extract component features, entropy weight-hierarchical analysis method is used to quantize reliability and dynamically adjust weight, material fatigue accumulation theory and Transformer network are used to predict residual life, and MAE / RMSE optimization model is used to improve modeling comprehensiveness and prediction reliability, and solve the problem of insufficient accuracy of the prior art.

Images