Machine learning based process parameter optimization method for laser breach robot

Abstract

The application discloses a laser breaking robot process parameter optimization method based on machine learning, which comprises obtaining sensor data and material characteristics, performing space-time fusion based on sensor characteristics on various sensor data to obtain multi-modal sensor fusion characteristics, combining corresponding robot process parameters and material characteristics to construct a breaking effect prediction model, performing pre-breaking on a to-be-broken project and fine-tuning the breaking effect prediction model, constructing a three-layer process knowledge graph according to domain knowledge, continuously acquiring sensor data by performing a breaking operation, performing self-adaptive multi-objective Bayesian optimization on the robot process parameters according to the three-layer process knowledge graph and the fine-tuned breaking effect prediction model, and outputting optimal robot process parameters. The method can greatly reduce process debugging time and reduce equipment loss risk, and has important engineering value for improving the autonomous operation capability of the laser breaking robot in a complex and unstructured environment.

Images