Method for complex increase of welding production control efficiency based on swarm intelligence algorithms and evolutionary modeling

Abstract

The article presents comprehensive methods for enhancing the efficiency of welder training management through the implementation of swarm intelligence algorithms and evolutionary modeling. It introduces an information-measurement system that automates the evaluation of welders' experience and work quality, facilitating objective task allocation and the improvement of training processes. Central to this system are swarm intelligence algorithms – specifically, bee algorithms, ant colony algorithms, and firefly algorithms – which optimize the selection of training courses and learning trajectories for welders. These algorithms streamline educational pathways and identify the most suitable courses for each welder, thus reducing training time and enhancing the quality of training. Evolutionary modeling algorithms assist in the efficient allocation of welding tasks among specialists based on their performance and work quality. The article details the processes involved in identifying and measuring weld defects, assessing weld quality, predicting defects, and improving training management efficiency. It also discusses the application of neural networks for weld defect analysis, which enhances assessment accuracy and automates quality control processes. Practical testing was conducted at OAO "BELGAZSTROY" and OOO "INVESTAP-MAIND," demonstrating a 20–30 % reduction in training time and a corresponding decrease in the number of weld defects. These results validate the effectiveness of the proposed approach and highlight its potential for improving welding work quality and welder training in industrial settings.