Arc Sound Loudness Analysis for MIG Welding Droplet Transition State Identification

Literature Overview

This study by Huang Linran, Gao Yanfeng, Wang Qisheng, and Gong Yanfeng, published in Hot Working Technology (2020, Vol. 49, No. 11, pp. 132-135), proposes an arc sound analysis method based on the Moore loudness model for identifying different droplet transition states in MIG welding. The research was supported by the National Natural Science Foundation of China (Grant 51465043), the Jiangxi Provincial Natural Science Foundation (Grant 20171BAB206033), and the Jiangxi Provincial Key R&D Program (Grant 20171BBE50011). The work was conducted at the School of Aeronautical Manufacturing Engineering, Nanchang Hangkong University.

Core Technical Approach

The research addresses a practical and important problem in welding process monitoring: the real-time identification of droplet transition modes. In MIG welding, the droplet transition mode (globular, short-circuiting, spray, or pulsating) fundamentally determines weld quality, and operators need to know which mode is active to adjust parameters accordingly. Traditional methods for identifying droplet transition modes rely on high-speed imaging or current/voltage signal analysis, which can be expensive, complex, or unsuitable for harsh industrial environments. The arc sound method offers a simpler, lower-cost alternative.

The authors first analyzed the arc sound signals and their power spectral density (PSD) under different droplet transition states. They found that different droplet transition forms produce arc sound signals with distinctly different energy levels and frequency distributions. The globular transfer mode, characterized by large, infrequent droplets, produces lower frequency components with periodic intensity variations. Short-circuiting transfer, involving frequent wire-to-pool contact, generates broadband noise with higher amplitude fluctuations. Spray transfer, with its high-frequency, fine droplet stream, produces relatively steady high-frequency content. Pulsating transfer shows intermediate characteristics with periodic modulation at the pulse frequency.

Moore Loudness Model Application

The Moore loudness model, originally developed for human auditory perception, was adapted for welding arc sound analysis. The model integrates three acoustic parameters: sound pressure level (amplitude), frequency content, and temporal characteristics. This tri-parameter integration is advantageous because it provides a single scalar metric that captures the complex acoustic signature of different welding modes.

The key insight is that the Moore loudness curve serves as a fingerprint for each transfer mode. By establishing reference loudness curves for known transfer modes, the system can classify real-time welding operations by comparing the measured loudness curve to the reference database. This approach is robust because it does not rely on a single frequency component or amplitude threshold, but rather on the holistic acoustic signature.

Real-Time Detection Implementation

For practical implementation, the arc sound signal is captured by a microphone or acoustic sensor positioned near the welding torch. The signal undergoes preprocessing (filtering, windowing) and is then processed through the Moore loudness calculation algorithm. The resulting loudness value or curve is compared against threshold values or pattern-matching templates to classify the current transfer mode.

The advantages of this approach for industrial deployment include:

1. Low cost: A basic microphone and signal processing unit are sufficient, with total system cost potentially below 500 USD.
2. Non-contact: The acoustic sensor does not interfere with the welding process and can be positioned at a distance from the arc.
3. Environmental robustness: Unlike optical methods, acoustic sensing is less affected by arc light, spatter, and fume.
4. Real-time capability: Modern signal processing can compute loudness metrics at rates sufficient for closed-loop control (typically >100 Hz sampling).
5. Multi-purpose: The same sensor can detect other welding anomalies such as wire sticking, arc blow, and shielding gas failure.

Study Insights and Reflections

This work exemplifies the transfer of methods from one engineering discipline to another. The Moore loudness model, developed for audio engineering and hearing science, finds an unexpected application in welding process monitoring. This cross-disciplinary approach is characteristic of mature engineering practice, where well-established analytical tools are repurposed for new applications. The success of this transfer depends on the fundamental similarity between the physical phenomena: in both cases, complex signals with multiple frequency components and temporal variations need to be reduced to a perceptually or functionally meaningful metric.

From a welding engineering perspective, the acoustic monitoring approach complements existing process monitoring methods. In a comprehensive welding quality assurance system, acoustic monitoring can serve as a primary or secondary sensor in a multi-sensor fusion architecture. For example, current and voltage signals provide information about the electrical characteristics of the arc, while acoustic signals provide complementary information about the mechanical and fluid dynamic aspects of the arc and droplet transition. The combination of both can improve classification accuracy and provide redundancy for fault detection.

One area for future development is the extension of this method to automated welding systems where real-time parameter adjustment is required. If the acoustic monitoring system can reliably identify the transfer mode in real time, it can trigger automatic adjustments to welding current, voltage, or wire feed speed to maintain the desired transfer mode, thereby improving weld quality consistency without operator intervention.