Prediction of Welding Electrical Current for Pipeline Welding Robot Based on Deep Learning Neural Network

Long-distance oil and gas pipelines play a key role in the transportation of oil and gas. Pipeline welding is a heavy task. When pipeline robots perform welding operations, misalignment between two pipelines can result in asymmetrical weld grooves on both sides, thereby leading to a poor welding quality. The commonly used laser vision method to identify the position of weld groove is easy to be interfered by strong light, so a new method is proposed based on welding current. This method can evaluate the actual position of weld groove by predicting the current change in the next stage. Firstly, the overall architecture of the welding current prediction neural network is constructed, which is the CNN-LSTM neural network. Subsequently, a current prediction algorithm based on the CNN-LSTM is developed. Finally, the validity of the proposed method is demonstrated through simulation experiments. The research offers robust technical support for the welding of long-distance oil and gas pipelines, thereby exhibiting significant practical value.