MIG Welding Bead Quality Optimisation Using Taguchi DOE on Mild Steel

MIG Welding Bead Quality Optimisation Using Taguchi DOE on Mild Steel is an experimental study that treats a welding booth the way a process engineer does: as a system whose inputs — current, travel speed and shielding-gas flow — can be tuned deliberately instead of by habit, using a formal design of experiments.

In small fabrication units in clusters like Peenya (Bengaluru) or Coimbatore, MIG parameters are usually set by the operator's experience, giving inconsistent bead geometry — too little penetration on some joints, heavy reinforcement and undercut on others. A full factorial study would need 27 or more plates, which no student team or small shop can afford. The Taguchi L9 orthogonal array solves exactly this: nine carefully chosen runs let three factors at three levels be evaluated independently, and the signal-to-noise (S/N) analysis identifies the setting that is not just good on average but robust against variation.

The team welds nine trial plates of IS 2062 E250 mild steel on the college MIG set, sections each bead transversely, etches it in 2% nital so the fusion line is visible, and measures bead width, reinforcement height and penetration under a stereomicroscope. Depth-to-width penetration ratio is analysed with the larger-the-better S/N ratio, reinforcement with the smaller-the-better, and ANOVA separates the factors that dominate from noise. A confirmation run at the predicted optimum closes the loop — if the measured bead lands within the confidence band, the study stands; if not, we say so and discuss why.

MIG welding quality in small Indian fabrication shops depends on parameters set by habit: current, travel speed and shielding-gas flow are chosen from memory or a generic chart, and the resulting beads vary widely in penetration, reinforcement and appearance, with no documented basis for the settings. Poor penetration weakens joints; heavy reinforcement and undercut create stress raisers and rework. A complete factorial experiment to find the right settings needs more trials than a small shop or a student team can run.

There is a need for a statistically designed, measurable parameter study — using a Taguchi L9 orthogonal array on mild-steel bead-on-plate welds — that identifies which of the accessible machine settings dominate bead geometry, quantifies their contribution through S/N ratios and ANOVA, and verifies the recommended setting with a confirmation experiment, all within a 14-week academic project using only a college welding lab and a metrology bench.

Abstract

This project applies the Taguchi method to MIG bead-on-plate welding of mild steel. Three parameters — current, travel speed and gas-flow rate — are studied at three levels each using an L9 orthogonal array. Bead penetration, width and reinforcement are measured on etched macro-sections, S/N ratios and ANOVA identify the dominant factors, and a confirmation weld validates the predicted optimum within a stated error band.

Introduction

Bead geometry is the first fingerprint of weld quality: penetration governs joint strength, excessive reinforcement creates stress concentration and wastes filler, and width irregularity signals unstable transfer. These responses are set by the heat input (current and speed) and the arc environment (gas flow), interacting in non-obvious ways that operator intuition cannot separate.

Existing System and Its Limits

• Shop-floor practice sets MIG parameters from a wire-make's generic chart or the welder's habit; small fabricators rarely document what was actually used.
• One-factor-at-a-time experiments miss interactions and need dozens of trials to cover the space.
• College welding labs teach joint making but not parameter justification: students can lay a bead but cannot say why that current or that speed.
• Limitation: no affordable, statistically defensible study links the welding variables a student can access to the bead geometry they can measure.

Proposed System

• A Taguchi L9(3^4) design with three factors — current (120/160/200 A), travel speed (20/30/40 cm/min) and gas flow (10/15/20 L/min of Ar–CO2) — and the fourth column reserved as an error estimate, giving nine welded plates instead of twenty-seven.
• A metrology protocol: each plate transversely sectioned, ground through 800 grit, etched in 2% nital, and measured for penetration depth, bead width and reinforcement height with three repeat readings per dimension.
• A two-response analysis: larger-the-better S/N for depth-to-width penetration ratio, smaller-the-better S/N for reinforcement, plus ANOVA on the L9 layout to rank factor influence and quantify the error column.
• A confirmation experiment at the predicted optimum with a pre-agreed acceptance rule (measured mean within the 95% prediction interval), and an honest failure analysis if it is not met.

Feasibility

• Technical: the welding lab already owns the machine and the metrology bench; the Taguchi maths is spreadsheet-level and documented in the NIST/SEMATECH statistics handbook and standard quality-engineering texts.
• Economic: nine plates, cutting discs, emery, etchant and gas come to roughly ₹2,500–4,000; the machine time is about two lab sessions.
• Operational: welding and sectioning are scheduled in weeks 5–8, leaving the rest of the 14-week window for analysis, confirmation and documentation, which suits the two-phase VTU timeline.

In scope

• Bead-on-plate (no joint) MIG welding on 10 mm IS 2062 E250 mild steel, 1.2 mm ER70S-6 wire, Ar–CO2 shielding.
• Three factors at three levels in an L9 array; two responses (penetration D/W ratio, reinforcement height); appearance observations (spatter, undercut) as qualitative notes.
• Macro-section metrology and S/N + ANOVA analysis; one confirmation weld.

Out of scope

• Mechanical testing of joints (tensile, bend, impact) — the study stops at bead geometry and macro-structure.
• Wire chemistry, base-metal composition variation or preheat as factors.
• Multi-pass welding, positions other than flat (1F/1G), and pulsed-MIG transfer modes.
• Residual-stress or distortion measurement.

The project follows a design → weld → section → measure → analyse → confirm experimental sequence, mapped onto the two VTU project phases.

Testing approach: experiment-level control (randomised run order to cancel time trends, constant stick-out, machine warm-up, gas-leak check), measurement-level control (three repeat readings per dimension, etch repeated if the fusion line is not sharp), and statistical control (the L9 spare column as an internal error estimate; ANOVA F-ratios checked before any factor is declared significant; the confirmation run as the final external check).

The study is organised as a closed experimental loop: design, execution, metrology, analysis, confirmation — with every step feeding documented records. The flowchart below shows the pipeline; the analysis stage branches into two responses because bead quality is not one number.

flowchart TD
  F{"Factors and levels"} -->|"current 120/160/200 A"| D["Taguchi L9 array + randomised run order"]
  F -->|"speed 20/30/40 cm/min"| D
  F -->|"gas 10/15/20 L/min"| D
  D --> W["Nine bead-on-plate welds, fixed stick-out and wire feed"]
  W --> C["Transverse sectioning, grinding, 2 percent nital etch"]
  C --> M["Metrology: penetration, width, reinforcement"]
  M --> SN{"S/N analysis"}
  SN -->|"D/W ratio - larger the better"| R1["Response table 1"]
  SN -->|"reinforcement - smaller the better"| R2["Response table 2"]
  R1 --> A["ANOVA with error column"]
  R2 --> A
  A --> P{"Predicted optimum levels"}
  P --> CW["Confirmation weld + metrology"]
  CW -->|"within 95 percent band"| REP["Validated results chapter"]
  CW -->|"outside band"| FB["Failure analysis and revised model"]

Prose description. The design stage fixes the statistical structure before any arc is struck, so the nine runs carry the information of a much larger campaign. Execution and metrology convert each run into three measured numbers plus observations. The analysis stage runs twice — once per response — because the best setting for penetration is not automatically the best for reinforcement, and the final confirmation stage closes the loop exactly as Taguchi practice requires: the prediction is checked against reality, and the study reports the error either way.

• Mechanical response extension: tensile-shear and bend tests on butt joints welded at the optimum, linking bead geometry to joint strength.
• Fourth factor: add wire feed speed or stick-out as the fourth L9 column instead of the error estimate, moving to interactions-aware designs.
• Hardness mapping: HAZ microhardness traverses as a third response with the same DOE framework.
• Robustness study: deliberately vary stick-out and gas pressure around the optimum to demonstrate the Taguchi noise-factor analysis.
• Shop transfer: package the optimum settings and the confirmation protocol as a one-page procedure card for a small fabrication unit, and measure their before/after rework rate.