Research On Robot Welding Quality Evaluation Based On Deep Learning And Multi-sensor Information Fusion

Welding technology is a very important processing method,which is widely used in manufacturing and processing.The welding quality not only affects the service performance and life of the welding products,but more importantly,it affects personal and property safety.In order to deeply study the influence law of welding parameters on welding quality,analyze the welding quality under different parameters,guide parameter optimization,this paper proposes a neural network algorithm based on deep learning and multi-sensor information fusion for welding quality evaluation and parameter optimization.A multi-sensor welding acquisition system based on arc welding robot platform was built to realize synchronous acquisition of voltage,current and sound.The experimental scheme was determined and the welding data collection of different welding qualities was completed.In order to eliminate the noise data in the welding data,wavelet packet filtering algorithm was used to filter the welding data,retain the effective information in the data,and the filtering effect showed that wavelet packet filtering greatly eliminated the noise interference in the welding data.In order to solve the problem of large fluctuation of data values and unclear features in welding data,data normalization was carried out to unify the scale of data and lay a foundation for data fusion and neural network learning process.A welding quality evaluation model is proposed.The model includes welding quality evaluation index,scoring system and evaluation results.Finally,five different levels of welding quality are obtained,and corresponding parameter optimization directions are given.A three-layer bidirectional long short-term memory network model based on attention mechanism(3L＿Attn＿BiLSTM)is proposed.By training 3L＿Attn＿BiLSTM training set,parameters are obtained.Through experiments,it is proved that the proposed network model can achieve high learning rate and low loss value when dealing with large amount of data,and its accuracy can reach 95.15% with loss value of 0.15.In order to solve the practicality of parameter adjustment after defects occur in welding process,a parameter adjustment prediction algorithm is proposed.According to quality evaluation,corresponding parameter adjustment prediction direction is given.Through experiments,it is proved that this algorithm can effectively improve welding quality and reduce occurrence of welding defects.The research in this paper applies new methods in artificial intelligence to robot welding quality evaluation and parameter optimization control.Through theoretical analysis and experimental verification,good results have been achieved.The research results provide ideas and references for research on welding quality control,which has positive significance for improving stability and yield of welding quality.