Digital Twin-based Operating Condition Anomaly Detection And Visual Analytics Methods Of Cranes

The crane is a kind of indispensable lifting equipment in the process of steel production.In order to achieve high quality and stable production,it is necessary to ensure reliable and stable operation of the crane.It is also necessary to make corresponding adjustment actions.Digital twin provides a viable solution for operational status detection and closed-loop control of equipment.This paper establishes anomaly detection algorithm models based on digital twin to address the problems of noisy and fluctuating collected data and poor adaptability of traditional anomaly algorithms.This paper introduces the visual analytics method to launch the research on the efficiency and accuracy of crane operating condition anomaly detection,mainly including the following:（1）An integrated digital twin model for crane operating condition anomaly detection is constructed.The anomaly detection visual analysis data flow model is established to describe the multi-stage analysis process of anomaly detection.The geometric model and the intrinsic relationship model are established to give a comprehensive digital twin integration model of cranes.（2）An anomaly detection method i Forest Vis DM is proposed.Statistical indicators for multi-dimensional data of cranes are designed and the dimensions suitable as input data are filtered using Decision Tree.The Isolation Forest is used to identify anomalies and combined with key parameters to complete the anomalies confirmation.Finally,the higher accuracy of the i Forest Vis DM method is confirmed by comparing the metrics of recall,precision and F1 of the detection results.（3）The reasoning and feedback rules for the cause of the anomalies of the cranes are studied.The rule base is built based on the Drools rule engine,and the closed-loop control of the traveling digital twin system is implemented based on the DQN reinforcement learning algorithm.（4）The digital twin prototype system for visual analytics of crane operating condition anomaly detection is designed.The system integrates the identification of anomaly data,abnormal cause reasoning and feedback operation solving methods.Finally,the better application value of the method in this paper is shown by case verification.