Acoustic Emission Short-Time Energy Based TIG Weld Penetration State Identification

Literature Overview

This paper by Wang Qisheng and colleagues from Nanchang Hangkong University, published in Hot Working Technology (2020, Vol. 49, No. 21, pp. 107-110), presents a method for online identification of TIG welding penetration states using acoustic emission (AE) signals processed through short-time energy (STE) analysis. The work was supported by the National Natural Science Foundation of China (51465043), the Jiangxi Provincial Natural Science Foundation (20171BAB206033), and the Jiangxi Provincial Key R&D Program (20171BBE50011).

Core Methodology

The research addresses the critical challenge of online weld quality monitoring during TIG welding. The authors developed a signal processing pipeline that begins with AE signal acquisition from the base material, followed by time-frequency domain analysis. A key preprocessing step involves limiting-filtering to suppress pulsed interference, which is particularly important in industrial environments where electromagnetic noise from rectifiers, transformers, and auxiliary equipment can mask the weld-related acoustic emissions.

The core of the method is the short-time energy analysis, which enhances the signal-to-noise ratio and highlights energy mutation information inherent in the AE signals. The STE is computed over sliding time windows, and the average STE value is extracted as the discriminating feature for penetration state classification. Three penetration states were investigated: under-penetration, critical penetration, and over-penetration.

Technical Parameters and Results

The results demonstrate a clear monotonic relationship: the average short-time energy increases sequentially from under-penetration to critical penetration to over-penetration. This trend is physically intuitive — under-penetration produces weaker acoustic activity because the arc has not fully interacted with the root of the joint, while over-penetration generates stronger emissions due to excessive heat input and greater fluid turbulence in the molten pool.

Engineering Practice Integration

From my experience in pipe welding quality control, online penetration monitoring is one of the most persistent challenges. In the production of API 5L X70 pipeline girth welds using TIG root passes, operators rely heavily on visual observation of the backside weld appearance and occasional radiographic testing. The AE-based method described here offers a non-invasive, real-time alternative that could be integrated into automated welding cells.

A practical consideration is sensor placement. In pipe welding applications, the circumferential geometry constrains the positioning of AE transducers. The authors used base-material-coupled sensors, which is feasible for flat plate joints. For girth welds on pipes with typical diameters of 219 mm to 1219 mm, transducer placement on the outer circumference at strategic clock positions would be necessary. The attenuation of AE signals through curved geometries must be characterized empirically for each pipe diameter and wall thickness combination.

Another important consideration is the influence of welding parameters on the AE signal baseline. In pipe welding, parameters such as travel speed, current, voltage, and electrode stick-out are continuously adjusted to accommodate joint geometry variations, fit-up tolerances, and thermal distortion. The SVM classifier trained on flat plate data may require recalibration for each new welding condition. A robust implementation would incorporate parameter normalization or adaptive thresholding.

Key Reflections

The 98% recognition accuracy is impressive for a laboratory study, but several questions arise for practical deployment. First, the paper does not address the effect of welding current and voltage fluctuations on the AE signal characteristics — in practice, even minor parameter drift can alter the energy content of the AE signal independently of penetration state. Second, the method assumes that the three penetration states produce distinctly separable STE distributions, which may not hold when the welding parameters are near the transition boundaries between states.

Nevertheless, the concept of using STE as a penetration indicator is sound and complements existing techniques such as back-side infrared thermography and in-situ CT scanning. For critical applications such as nuclear-grade stainless steel pipe welding (per NB/T standards) or offshore platform piping (per DNV-ST-F101), where full penetration is mandatory, this method could serve as a supplementary real-time indicator alongside conventional post-weld NDT.

Reference Value and Outlook

This work provides a viable pathway for real-time weld penetration monitoring using acoustically coupled sensors, which are relatively inexpensive and non-intrusive compared to optical or thermal methods. Future research should extend the method to pipe geometry, incorporate multi-parameter adaptive classification, and validate the approach under production conditions with varying base material thicknesses and compositions. The integration of AE-STe features with other online monitoring signals — such as arc voltage, current, and travel speed — could significantly enhance classification robustness.