Целью работы является получение адекватного автоматического алгоритма дуговой сварки, использующего методы машинного обучения для мониторинга параметров сварки, прогнозирования дефектов шва и автоматической корректировки настроек с целью повышения качества и надежности сварных соединений. Настоящая работа посвящена созданию алгоритма для автоматизации дуговой сварки методом MIG/MAG с применением технологий машинного обучения, в частности модели Random Forest. Разработанная система направлена на непрерывный контроль таких параметров сварки, как температура шва, сила тока, скорость подачи проволоки, расход защитного газа и ширина шва. Основная задача – прогнозирование вероятности дефектов сварного соединения и автоматическая корректировка параметров для их предотвращения, что способствует повышению прочности и долговечности конструкций. Алгоритм реализован на языке Python с использованием библиотек Numpy, Sklearn и Matplotlib и адаптирован для работы в среде Google Colab, что упрощает проведение экспериментов и анализ данных. В ходе тестирования система продемонстрировала высокую точность предсказания дефектов (100%), подтвержденную метриками классификации и анализом ROC-кривой. Предложенный подход обеспечивает оптимизацию сварочных процессов, минимизацию производственных ошибок и открывает перспективы для адаптации к другим видам сварки, что делает его ценным инструментом для промышленного применения.
The aim of the work is to obtain an adequate automatic arc welding algorithm using machine learning methods to monitor welding parameters, predict weld defects and automatically adjust settings in order to improve the quality and reliability of welded joints. This work is devoted to the creation of an algorithm for automating arc welding by the MIG / MAG method using machine learning technologies, in particular the Random Forest model. The developed system is aimed at continuous monitoring of such welding parameters as weld temperature, current strength, wire feed speed, shielding gas flow rate and weld width. The main task is to predict the probability of welded joint defects and automatically adjust the parameters to prevent them, which helps to increase the strength and durability of structures. The algorithm is implemented in Python using the numpy, sklearn and matplotlib libraries and adapted to work in the Google Colab environment, which simplifies experiments and data analysis. During testing, the system demonstrated high accuracy in predicting defects (100%), confirmed by classification metrics and ROC curve analysis. The proposed approach ensures optimization of welding processes, minimization of production errors and opens up prospects for adaptation to other types of welding, which makes it a valuable tool for industrial applications.
1. Jie W., Zhifen Z., Zijian B., Shuai Z., Rui Q., Jing H., Guangrui W. On-line defect recognition of MIG lap welding for stainless steel sheet based on weld image and CMT voltage: Feature fusion and attention weights visualization // Journal of Manufacturing Processes. 2023. Vol. 108. P. 430–444. DOI: 10.1016/j.jmapro.2023.10.081. URL: https://www.sciencedirect.com/science/article/abs/pii/S1526612523010368 (дата обращения: 15.03.2025).
2. Zhang Z., Wen G., Chen S. Weld image deep learning-based on-line defects detection using convolutional neural networks for Al alloy in robotic arc welding // Journal of Manufacturing Processes. 2019. Vol. 45. P. 208–216. DOI: 10.1016/j.jmapro.2019.06.023. URL: https://www.sciencedirect.com/science/article/abs/pii/S1526612519301860 (дата обращения: 15.03.2025).
3. Yanxi Z., Deyong Y., Xiangdong G., Seiji K. Monitoring of Welding Status Based on a DBN Model During Laser Welding // Engineering. 2019. DOI: 10.1016/j.eng.2019.01.016. URL: https://www.researchgate.net/publication/334277053_Monitoring_of_Welding_Status_Based_on_a_DBN_Model_During_Laser_Welding (дата обращения: 17.03.2025).
4. Li D., Song Y., Ye F. Online monitoring of weld defects for short-circuit gas metal arc welding based on the self-organizing feature map neural networks // Proceedings of the International Joint Conference on Neural Networks. 2000. Vol. 5. P. 239–244. DOI: 10.1109/IJCNN.2000.861464. URL: https://www.researchgate.net/publication/3857970_Online_monitoring_of_weld_defects_for_short-circuit_gas_metal_arc_welding_based_on_the_self-organizing_feature_map_neural_networks (дата обращения: 12.03.2025).
5. Thakar H., Chaudhari M., Vora J., Patel V., Das S., Bandhu D., Gupta M., Reddy V. Performance optimization and investigation of metal-cored filler wires for high-strength steel during gas metal arc welding // High Temperature Materials and Processes. 2023. DOI: 10.1515/htmp-2022-0305. URL: https://www.researchgate.net/publication/377076428_Performance_optimization_and_investigation_of_metal-cored_filler_wires_for_high-strength_steel_during_gas_metal_arc_welding (дата обращения: 15.03.2025).
6. Wang Z., Li L., Chen H., Lin S., Wu J., Ding T., Tian J., Xu M. Recognition of GTAW weld penetration based on the lightweight model and transfer learning // Welding in the World. 2022. Vol. 67. P. 1–14. DOI: 10.1007/s40194-022-01396-0. URL: https://www.researchgate.net/publication/364294769_Recognition_of_GTAW_weld_penetration_based_on_the_lightweight_model_and_transfer_learning (дата обращения: 15.03.2025).
7. Khrais S., Hmoud H., Abdel A., Darabseh T. Impact of Gas Metal Arc Welding Parameters on Bead Geometry and Material Distortion of AISI 316L // Journal of Manufacturing and Materials Processing. 2023. Vol. 7. Article 123. DOI: 10.3390/jmmp7040123. URL: https://www.researchgate.net/publication/371969521_Impact_of_Gas_Metal_Arc_Welding_Parameters_on_Bead_Geometry_and_Material_Distortion_of_AISI_316L (дата обращения: 13.03.2025).
8. Kah P., Edigbe G., Ndiwe B., Kubicek R. Assessment of arc stability features for selected gas metal arc welding conditions // SN Applied Sciences. 2022. Vol. 4. Article 268. DOI: 10.1007/s42452-022-05150-5. URL: https://link.springer.com/article/10.1007/s42452-022-05136-7 (дата обращения: 15.03.2025).
9. Haichao L., Yixuan M., Mingrui D., Xin W., Tong C. Defects detection of GMAW process based on convolutional neural network algorithm // Scientific Reports. 2023. Vol. 13. DOI: 10.1038/s41598-023-48698-x. URL: https://www.researchgate.net/publication/376146088_Defects_detection_of_GMAW_process_based_on_convolutional_neural_network_algorithm (дата обращения: 17.03.2025).
10. Alnecino N., Francois N. B., Paul K. Optimization of GMAW Process Parameters in Ultra-High-Strength Steel Based on Prediction // Metals. 2023. Vol. 13. Article 1447. DOI: 10.3390/met13081447. URL: https://www.researchgate.net/publication/373090987_Optimization_of_GMAW_Process_Parameters_in_Ultra-High-Strength_Steel_Based_on_Prediction (дата обращения: 18.03.2025).
11. Sakthivel R., Venkadeshwaran P., Sridevi R., Meeran R., Kamaraj C. Effect of Welding Current, Arc Voltage and Gas Flow Rate on Depth of Penetration during MIG Welding of AA2014 Plate // International Journal of Scientific & Engineering Research. 2015. Vol. 6. URL: https://www.researchgate.net/publication/279920913_Effect_of_Welding_Current_Arc_Voltage_and_Gas_Flow_Rate_on_Depth_of_Penetration_during_MIG_Welding_of_AA2014_Plate (дата обращения: 12.03.2025).
12. Asif K., Zhang L., Derrible S., Indacochea E., Ozevin D., Ziebart B. Machine learning model to predict welding quality using air-coupled acoustic emission and weld inputs // Journal of Intelligent Manufacturing. 2022. Vol. 33. P. 1–15. DOI: 10.1007/s10845-020-01667-x. URL: https://www.researchgate.net/publication/346315747_Machine_learning_model_to_predict_welding_quality_using_air-coupled_acoustic_emission_and_weld_inputs (дата обращения: 15.03.2025).
13. Chen G., Sheng B., Luo R., Pengzhen J. A parallel strategy for predicting the quality of welded joints in automotive bodies based on machine learning // Journal of Manufacturing Systems. 2022. Vol. 62. P. 636–649. DOI: 10.1016/j.jmsy.2022.01.011. URL: https://www.researchgate.net/publication/358478950_A_parallel_strategy_for_predicting_the_quality_of_welded_joints_in_automotive_bodies_based_on_machine_learning (дата обращения: 15.03.2025).
14. Kumaresan S., Aultrin K., Kumar S.S., Anand M. Transfer Learning With CNN for Classification of Weld Defect // IEEE Access. 2021. DOI: 10.1109/ACCESS.2021.3093487. URL: https://www.researchgate.net/publication/352848476_Transfer_Learning_With_CNN_for_Classification_of_Weld_Defect (дата обращения: 12.03.2025).
15. Jinquan L., Jie L., Jinli X., Ding L. Welding quality analysis and prediction based on deep learning // The International Journal of Advanced Manufacturing Technology. 2021. P. 405–416. DOI: 10.1109/WCMEIM54377.2021.00045 URL: https://www.researchgate.net/publication/363118146_Welding_quality_analysis_and_prediction_based_on_deep_learning (дата обращения: 15.03.2025).
16. Kim I.-S., Lee M.-G., Jeon Y. Review on Machine Learning Based Welding Quality Improvement // International Journal of Precision Engineering and Manufacturing-Smart Technology. 2023. Vol. 1. P. 219–226. DOI: 10.57062/ijpem-st.2023.0017. URL: https://www.researchgate.net/publication/372169737_Review_on_Machine_Learning_Based_Welding_Quality_Improvement (дата обращения: 14.03.2025).