Research On Welding Quality Inspection Method Based On Neural Network Optimization Case-Based Reasoning

In recent years,with the strategic policy of "made in China 2025",manufacturing industry has developed rapidly.Welding is an important link in manufacturing industry,and it has many applications in automobile manufacturing,aerospace,mechanical electronics,ship navigation and so on.As one of the many industries in manufacturing industry,automobile production is also changing to the direction of intelligence.For example,the traditional factory manual welding mode has been upgraded to the welding mode of intelligent robot.Therefore,it is impossible to meet the requirements of modern industry simply relying on the artificial inspection of welding quality.So how to intelligently detect welding has become an urgent problem.In view of the above,this thesis introduces semantic ontology and improved BP neural network into case-based reasoning welding quality detection method.The main research work is as follows1.Study the method of extracting the bottom image and visual features of welding defects,collect them through 3D vision cameras,vision sensors,etc.,then perform geometric size definition and feature extraction,and finally store these features in the Allegro Graph graphics database through ontology instance mapping,which is the welding quality Prepare for testing.2.Aiming at the lack of case representation in case-based reasoning in welding quality testing,the ontology modeling criteria,evaluation criteria and modeling methods are studied,welding geometric features are extracted and the associations between ontology are established,and the ontology modeling tool Top Braid Composer is used.The welding defect ontology is constructed,the information extracted from the welding defects of the production line is described semantically,and the welding quality inspection ontology model is constructed.3.Aiming at the problem of slow case retrieval speed and low retrieval accuracy in case reasoning in welding quality inspection,research on the welding quality retrieval method based on genetic algorithm to optimize BP neural network,because BP neural network is easy to fall into local optimality,slow convergence speed and other shortcomings In this paper,the genetic algorithm is used to optimize the initial weights and thresholds of the BP neural network.The results of simulation experiments show that the genetic algorithm optimizes the BP neural network as a case-based reasoning retrieval can improve the retrieval rate and accuracy.4.Building a welding quality inspection platform to test the effectiveness of the welding quality inspection of the production line.This article uses laboratory software and hardware resources,and then combines related technologies to build a verification platform and verify its functional modules.The verification results prove the effectiveness of the welding quality inspection method based on neural network optimized case reasoning.