A Brief History of In-Process Monitoring for Welding and Joining

This article presents a historical perspective of the research that fueled the initial development of in-process monitoring for real-time feedback of welding and joining process quality.

In-process monitoring evolved independently out of three fields of study in the late 1960s.  The first group was focused on sensing, modeling, and control and consisted of researchers from the fields of electrical and mechanical engineering.  While this first group bypassed the concept of in-process monitoring altogether, they laid the foundation for much of the research and development in weld process monitoring (and control) that followed in the 1980s and onward. The second group was composed of researchers in the nondestructive evaluation field and, in particular, from the application of acoustic emission techniques for nondestructive testing. Researchers from this second group realized that the post-process, acoustic emissions generated from the sudden relief of concentrated stresses in structures and weldments can also occur in-situ (i.e., during the actual welding or joining process). And finally, the third group was composed of welding engineers performing process development in their respective fields of process expertise.

The initial research into sensing, modeling, and control in the late 1960s and early 1970s was motivated by the need for achieving robust mechanized and robotic automation [1].  Most of the early research primarily involved arc welding and focused on arc sensing [2].  Arc sensing (also commonly referred to as through-the-arc sensing) monitors the change in current (constant voltage power sources) or voltage (constant current and pulsed current power sources) in the arc.  Systems employing arc sensors rely on the principle that the resistance between the welding torch's electrode and the workpiece is a function of the distance separating the two.  The obvious advantage of arc sensing is the use of the arc itself as a sensor.  This reduces costs and avoids the technical challenges associated with instrumenting external sensors in or around the harsh environment of the welding arc.

In addition to through-the-arc sensing, external sensors were also investigated early on to facilitate control of the welding process.  Surface-based, temperature sensing was one of the first feedback techniques for closed-loop control of the heat input to accommodate variations in the fit-up, fixturing, and material thickness during arc welding [3].  While temperature measurements have certain drawbacks (including intimate contact with the workpiece or sensor calibration to account for differences in thermal emissivity of the workpiece), they provide a direct interrogation into what is happening at the workpiece surface and an indirect means of inferring depth of penetration and molten puddle size. 

Finally, weld pool oscillation sensing as a means of indirectly assessing the depth of penetration in an autogenous GTA weld was first proposed by Hardt, et al. [4].  Originally recognized by Cheever and Howden [5] and later investigated further by Kotecki, et al.[6], the authors were able to empirically verify that the pool's natural resonant frequency increased as the pool's radius decreased. The coupled sensing and modeling technique is founded upon the premise that the weld puddle's oscillations are a function of (1) the external force, (2) the properties of the molten region, (3) the surface tension, and (4) the shape of the container. 

In 1989, following over 20 years of heavily active research in sensing, modeling, and control of welding and joining processes, Cook summarized the progress in real-time quality control as follows [7]:

Sensing and control of discontinuity formation, i.e., real-time quality control, is in its infancy, and a great deal of research remains before seeing it in wide use on the production floor.

One of the primary contributions to the development of in-process monitoring that bore out from the sensing, modeling, and control research effort is the ability to use process attributes (for example, voltage during gas tungsten arc welding) to infer in real-time difficult-to-measure product attributes (for example, depth of penetration) [8].  Directly sensing the product attributes of interest provides the most robust feedback to a closed-loop control system. However, in certain cases, directly sensing many of the product attributes associated with welding is impractical and/or infeasible in a production environment. Therefore, indirect sensing techniques are employed where a sensor is coupled with a model to “indirectly” sense, or infer, the DWP of interest.

A second large body of research evolved out of the nondestructive evaluation field in terms of extending the application of acoustic emission testing from a post-process technique to an in-situ technique. The work primarily involved inspecting welds for nuclear power plants where quality control and quality assurance best practices are paramount. Consequently, this body of research was focused on ways to supplement and enhance the rigorous post-process nondestructive tests that were currently in place. Unlike the sensing and control research that was presented earlier, this group of researchers was primarily interested in defining, evaluating, corroborating, and maturing the concept of in-process monitoring from an intellectual curiosity to an established component in a manufacturing quality assurance best practices.

One of the first documented references to the concept of in-process monitoring was published by W.D. Jolly in 1968 [9]. Jolly went on to publish several more reports extending his original findings which are presented here. In 1969, Jolly demonstrated that the level of acoustic emissions increases in the presence of cracks, porosity, and inclusions during submerged arc welding [10]. Based on these initial findings, he was able to verify these results on a number of different materials and from a variety of different processes, including gas tungsten arc welding, resistance spot welding, and brazing. His work in gas tungsten arc welding revealed the extremely sensitive nature of acoustic emission by detecting hairline cracks that otherwise went undetected by radiography [11]. At the time, all nondestructive methods used for weld inspection were applied post-process. In light of this, Jolly emphasized the real-time benefit of acoustic emission as a nondestructive, in-process test for welds:

The most important feature of this technique as opposed to other nondestructive testing methods is that the data is real-time.

While Jolly acknowledged that acoustic emission is unable to differentiate between types of defects, he suggested one of the more important take-aways from his findings is how real-time acoustic emission can be used to supplement existing post-process nondestructive weld tests:

The [acoustic emission] technique is not qualified as a final inspection test, but when used to supplement established weld inspection methods acoustic emission monitoring can result in the reduction of re-work costs and improvement of weld quality.

In 1973, Romrell demonstrated the application of acoustic emission monitoring to closure welds for nuclear reactor fuel pins [12]. Romrell reaffirmed Jolly's findings regarding the sensitivity of acoustic emission and the importance of a real-time, nondestructive test method especially when other traditional post-process test methods cannot be applied.

Finally, a very limited body of research in welding process development yielded noteworthy findings regarding the concept of real-time feedback and the evolution of in-process monitoring. In 1972, Ellis proposed an in-process method of weld quality for continuous drive friction welding [13]. Ellis investigated the effect of process parameter variations on tensile strength during continuous drive friction welding of mild steel. In doing so, he noted a relationship between ultimate tensile strength and burn off rate:

There exists a relatively consistent relationship between ultimate tensile strength and burn off rate which is apparently independent of pressure, speed of rotation and diameter of workpiece of the ranges examined.

In light of this evidence, Ellis suggests monitoring the burn off as an in-process means for assessing bond quality:

Although the reliability of the process has been shown to be of a high order, there has nevertheless been a pressing demand for an effective means of measuring the quality of friction welds preferably by an in process technique.

In 1974, motivated by Jolly's findings, Baeslack demonstrated the ability to differentiate nominal welds from flawed welds using acoustic emission monitoring during arc stud welding [14]

The principle behind [in-process] monitoring is that the acoustic emissions produced during the fabrication of a quality product differ from those produced in the fabrication of a defective product.

Baeslack concluded by stating one of the key advantages of an in-process monitoring capability is immediate structural integrity feedback:

The effective implementation of in-process ... systems offers the advantage of immediate structural integrity determination not found in other [post-process] nondestructive testing procedures...

In summary, the concept of in-process monitoring was born by three separate groups of researchers in the late 1960s and early 1970s. While the sensing, modeling, and control group contributed indirectly to the conception of in-process monitoring, their real contribution was to lay the foundation for the rigorous research that ensued in the 1980s and onward. The work published by Jolly reflects the earliest known mention of in-process monitoring for welding and joining processes.

--

[1] J.D. Lane. "Robotic Welding State of the Art." Robotic Welding, IFS (Publications) Ltd., UK, pp. 1-10, 1987.

[2] E.P. Vilkas, "Automation of Gas Tungsten Arc Welding Process." The Welding Journal, vol. 45, pp. 410–416, May 1966.

[3] W. M. McCampbell, G. E. Cook, L. E. Nordholt, and G. J. Merrick. “Development of Weld Intelligence System.” Welding Journal, vol. 45, pp. 139S, March 1966.

[4] D.E. Hardt, K.M. Masubichi, H.M. Paynter, E. Unkel, J. Converti, and M. Zachsenhouse. "Improvement of Fusion Welding Through Modeling, Measurement, and Real-Time Control." International Conference on Welding Technology for Energy Applications, May 1982 (Gatlinburg, TN), p 281–299.

[5] D.L. Cheever and D.G. Howden. "Technical Note: Nature of Weld Surface Ripples." The Welding Journal, vol. 48, pp. 179s-180s, April 1969.

[6] D.J. Kotecki, D.L. Cheever, and D.G. Howden. "Mechanism of Ripple Formation during Weld Solidification." The Welding Journal, vol. 51, pp. 386s–391s, August 1972.

[7] G.E. Cook, K. Andersen, R.J. Barnett. "Feedback and Adaptive Control in Welding." Recent Trends in Welding Science and Technology, Proceedings of the Second International Conference on Trends in Welding Research, S.A. David and J.M. Vitek, Ed., May 14–18, 1989 (Gatlinburg, TN), ASM International, p 967–971.

[8] D.A. Hartman, D.R. DeLapp, G.E. Cook, R.J. Barnett. "Intelligent Fusion Control Throughout Varying Thermal Regions." 34th Annual Meeting -- IEEE Industry Applications Society Phoenix, Arizona, October 3-7, 1999.

[9] W.D. Jolly. "An In Situ Weld Defect Detector Acoustic Emission." BNWL-817, Battelle Memorial Institute, Pacific Northwest Laboratory, January 1968.

[10] W.D. Jolly. "The Use of Acoustic Emission as a Weld Quality Monitor." BNWL-SA-2727, Battelle Memorial Institute, Pacific Northwest Laboratory, September 1969.

[11] W.D. Jolly. "The Application of Acoustic Emission to In-Process Weld Inspection." BNWL-SA-2212, Presented at 1969 Spring Conference of the ASNT, Los Angeles, California, March 1969.

[12] D.M. Romrell. "Acoustic Emission Weld Monitoring of Nuclear Components." The Welding Journal, vol. 52, pp. 81s-87s, February 1973.

[13] C.R.G. Ellis. "Continuous Drive Friction Welding of Mild Steel." The Welding Journal, vol. 51, pp. 183s-197s, April 1972.

[14] W.A. Baeslack. "Acoustic Emission Monitoring of the Arc Stud Welding Process as a Method of Determining Weldment Quality." M.S. Thesis, The Ohio State University, 1974.

 

© 2023 Manufacturing Behavioral Science LLC. All Rights Reserved.

Search