Operational Control as a Means of the Evaluation of Quality of Welded Connections for Flash-Butt Welding of Modern High-Strength Steels

1Kuchuk-Yatsenko, SI  https://orcid.org/0000-0002-1166-0253
1Rudenko, PM  https://orcid.org/0000-0002-7770-2145
1Gavrish, VS  https://orcid.org/0000-0002-2154-9719
1Didkovsky, ОV  https://orcid.org/0000-0001-5268-5599
1Antipin, Ye.V  https://orcid.org/0000-0003-3297-5382
1Ziakhor, IV  https://orcid.org/0000-0001-7780-0688
1Paton Electric Welding Institute, NAS of Ukraine
Sci. innov. 2020, 16(2):66-71
https://doi.org/10.15407/scine16.02.066
Section: Research and Engineering Innovative Projects of the National Academy of Sciences of Ukraine
Language: English
Abstract: 
Introduction. Flash-butt welding (FBW) of rails is controlled in real time based on the tolerances of the main process parameters according to the data of specifications. The operational control algorithm enables real-time detection of low quality weld and inadmissible trends in the process.
Problem Statement. In addition to the existing method, in order to control the compliance of welding of new high-strength steel rails with the specifications, it is necessary to take into account the width of the heat-affected zone (HAZ). The known numerical methods for calculating the HAZ in real time cannot be implemented because of insufficient computational capabilities of modern control systems.
Purpose. To develop an algorithm for real-time monitoring of FBW with predicting the width of the HAZ, in compliance with technical specifications.
Materials and Methods. A numerical method for calculating thermal fields during flash-butt welding, a regression analysis for HAZ prediction. The HAZ width is calculated based on the process parameters at the burning-off stage and on the upsetting.
Results. A real-time algorithm has been developed for controlling FBW of modern high-strength steels with prediction of the HAZ width. The algorithm is based on mathematical modeling of joints formation during flash-butt welding.
Conclusions. The regression equation in the form of a second-order polynomial or MLP neural network with a structure of 3 neurons in the input layer — 2 neurons in the hidden layer — 1 neuron in the output layer can be used for calculating the HAZ width in real time with the required accuracy for practical use. Prediction of the HAZ width during operational control expands the possibilities of its use for resistance butt-welding of high-strength rails. The developed algorithm has increased the accuracy and reliability of operational control of FBW in real time.
Keywords: flash-butt welding, fuzzy logic, heat-affected zone, high-strength rails, regression model, Sugeno classifier