Curve Fitting Method for Dual-Pulse MIG Welding Expert Database

Literature Overview

This paper by Lin Fang, Chen Xiaofeng, Wei Zhonghua, Cui Longbin, Gao Liwen, and Xue Jiaxiang, published in Electric Welding Machine in 2011, proposes a novel data fitting and evaluation method for developing an expert database for dual-pulse MIG welding. The authors conducted process experiments to obtain 50 sets of data matching given welding current and wire feed speed, then used MATLAB to perform curve fitting at different orders. This work was supported by multiple funding sources including the National Natural Science Foundation of China.

Core Technical Methodology

The dual-pulse MIG welding process employs two distinct pulse amplitudes within each cycle: a high-current pulse for penetration and a low-current pulse for wire feeding and arc stability. The expert database maps the relationship between welding current and wire feed speed to achieve stable arc conditions without wire sticking or back-burning. The authors developed a systematic approach to populate this database through experimental data and mathematical curve fitting.

The experimental approach involved collecting 50 data points where the welding current and wire feed speed were matched to produce stable arc conditions. These data points were then fitted using polynomial curves of varying orders in MATLAB. The optimal fitting order was selected based on the combined evaluation of R-square, adjusted R-square, and root mean square error (RMSE). The resulting optimal fitting curve provides a complete expert database that maps any given welding current to the corresponding stable wire feed speed.

Dual-Pulse Process Characteristics

Dual-pulse MIG welding is particularly advantageous for welding thin sheets, aluminum alloys, and high-conductivity materials where the conventional single-pulse process may produce excessive heat input or unstable arc conditions. The low pulse maintains arc stability and controls wire feeding, while the high pulse provides sufficient penetration without excessive thermal distortion. The expert database developed in this study enables digital welding machines to achieve single-variable control, where the operator sets only the welding current and the machine automatically selects the appropriate wire feed speed.

Engineering Practice Implications

The expert database concept is directly applicable to the development of user-friendly welding equipment where operators with limited experience can produce high-quality welds by simply selecting the desired welding current. The single-variable control approach reduces the complexity of parameter selection and minimizes the risk of operator error. The experimental verification confirmed that welding parameters selected from the expert database produced stable welding processes without wire sticking or back-burning, and the resulting weld quality was good.

This approach can be extended to other welding processes and materials by developing process-specific expert databases through similar experimental and fitting procedures. For pipe welding applications, where joint geometry and material thickness vary, the expert database can be adapted to include additional variables such as joint fit-up, material thickness, and welding position.

Study Insights and Reflections

The methodology presented in this paper represents a practical approach to knowledge-based welding system development. By combining experimental data collection with mathematical curve fitting, the authors created a compact and efficient expert database that captures the essential relationships between welding parameters. The use of multiple evaluation metrics (R-square, adjusted R-square, and RMSE) for selecting the optimal fitting order demonstrates methodological rigor. The approach is transferable to other welding processes and can be adapted for different materials and joint configurations, making it a valuable tool for welding process optimization and equipment development.