Welding Defect and Robot Fault Prediction Methods Based on Historical Knowledge

Failures of different components of the welding robot system may cause the same welding defects, which will waste time and personnel energy in troubleshooting the causes of welding defects. In order to solve this problem, this paper proposes a method to predict robot equipment failure based on welding defects based on knowledge graph theory. This method first uses statistical ideas to preprocess manually recorded historical data to obtain the proportion of different robot component failures in each type of welding defect. Then the knowledge graph method is used to establish a relationship model between welding defects and robot equipment failures, and analyze the correspondence between welding defects and robot equipment failures. Finally, the welding defects produced during the robot welding process are input into the established relational model to predict the location of the robot equipment failure. Moreover, welding defects and robot equipment failures that occur again in actual production are allowed to be input into the established model, and new welding defects and equipment failures are also allowed to continue to be input into the established model. The established model can be continuously iteratively optimized through newly input welding defect and equipment failure data to continuously improve the completeness and accuracy of the relationship model. The practicality of the established relationship model is verified through the manually recorded welding defect and robot failure history data in actual production. The research results can be used to guide robot equipment failures caused by welding defects during the robot welding process.