Fuzzy System Identification Methods in TIG Welding Process Modeling

Literature Overview

This 1998 study by Li Wen, Sun Hui, and Chen Zigang from Dalian Railway Institute, published in the Railway Journal, addresses the application of fuzzy system identification to TIG welding process modeling. The work was supported by the Dalian Municipal Youth Academic Leader Project and the Liaoning Provincial Natural Science Foundation. At the time, intelligent modeling of welding processes was in its early stages in China, and this paper represents a pioneering effort to apply fuzzy logic to capture the nonlinear, multi-variable behavior of the TIG arc welding process.

Core Technical Content

The TIG welding process involves a complex interplay of electrical parameters (current, voltage, pulse frequency), geometric parameters (travel speed, torch angle, stick-out distance), and consumable parameters (tungsten electrode diameter, shielding gas flow rate) that collectively determine the weld bead geometry, penetration depth, and microstructure. The authors recognized that traditional first-principles modeling approaches struggle to capture this nonlinearity, particularly when transitioning between different welding conditions or material combinations.

Fuzzy system identification, as applied in this study, involves constructing a set of fuzzy rules that map input variables (such as welding current I, voltage U, travel speed v) to output variables (weld width W, reinforcement height H, penetration depth P). The key advantage is that fuzzy rules can be derived from expert knowledge and experimental data without requiring a precise mathematical transfer function.

Key Modeling Parameters

Parameter Category	Input Variables	Output Variables	Typical Range
Electrical	Current I (A), Voltage U (V)	Penetration depth P (mm)	I: 80–250 A; U: 12–22 V
Geometric	Travel speed v (mm/min), torch angle α (deg)	Bead width W (mm)	v: 50–300 mm/min
Consumable	Electrode diameter d (mm), gas flow Q (L/min)	Spatter rate, porosity index	d: 1.6–4.0 mm; Q: 8–20 L/min

The study employed Mamdani-type fuzzy inference with triangular membership functions. The defuzzification method used was the centroid method, which provides a smooth output surface suitable for process control applications.

Connection to Cladding and Overlay Practice

In the context of weld overlay and cladding operations, the TIG process is frequently employed for single-pass overlay layers, particularly for nickel-based alloy coatings on carbon steel substrates or for the first-pass welding in multi-layer cladding sequences. The fuzzy modeling approach presented here has direct relevance to optimizing overlay parameters:

For Inconel 625 overlay on SA-516 Gr.70, the arc stability and dilution control are critical. Fuzzy rules can encode the relationship between pulse parameters (base current, peak current, pulse duration) and the resulting dilution percentage, which typically must be controlled below 30% to maintain the corrosion resistance of the overlay layer.
In strip cladding applications where GTAW is used for the cap pass, the fuzzy model can predict the dilution gradient across the clad layer, enabling optimization of the strip composition to achieve uniform corrosion resistance.

Engineering Practice Implications

The dilution control problem in TIG overlay welding can be formulated as a fuzzy optimization problem. The input variables include the number of passes, interpass temperature, and heat input per pass. The output is the predicted dilution at the clad-base metal interface. By constructing a fuzzy rule base from qualified weld coupons (per NB/T 47014 qualification procedures), engineers can predict dilution before committing to full-scale fabrication, reducing the risk of intergranular corrosion or stress corrosion cracking in the final product.

Key Insights and Reflections

The 1998 publication date places this work at the beginning of intelligent welding research in China. The methodological approach—using fuzzy logic to bridge the gap between empirical welding knowledge and quantitative process control—remains relevant today. However, the study was limited by the computational resources available at the time, which constrained the complexity of the fuzzy rule base.

For modern cladding operations, the fuzzy identification framework can be extended by:

Incorporating more input variables, such as substrate thermal conductivity and alloy composition effects on dilution.
Using adaptive fuzzy systems that update their rule base based on real-time sensor feedback (arc voltage monitoring, acoustic emission, infrared thermography).
Combining fuzzy logic with finite element thermal simulation to predict residual stress distributions in clad plate assemblies.

The fundamental insight remains valid: welding processes are inherently nonlinear and uncertain, and fuzzy logic provides a practical framework for encoding expert knowledge into actionable process models. This is particularly valuable in overlay welding, where the interaction between dissimilar metals introduces additional complexity that is difficult to capture with purely analytical models.

Reference Value and Outlook

This paper serves as an important historical reference for understanding the evolution of intelligent welding research in China. The fuzzy system identification methodology, while conceptually simple, provides a robust foundation for process modeling that complements more modern approaches. In the context of bimetal pressure vessel fabrication, where weld overlay qualification per NB/T 47014 requires extensive coupon testing, fuzzy modeling could significantly reduce the number of trial welds needed to establish qualified welding procedures. The approach is particularly promising for novel material combinations where limited empirical data exists, such as zirconium-clad pressure vessels or advanced nickel-alloy overlays for hydrogen service.