Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Ultrasound Phased Array Evaluation of Friction Stir Welds


New Developments in Ultrasound Phased Array for the Evaluation of Friction Stir Welds

Colin Bird, TWI 1
Olivier Dupuis, R/D TECH 2
Andre Lamarre, R/D TECH 2

1 TWI; Address; City, State, Postal Code, UK
2 R/DTech; 505, boul. du Parc-Technologique; Quebec, Quebec, G1P4S9, Canada

Paper presented at Friction Stir welding and Processing II, TMS annual meeting 2-6 March 2003


The use of friction stir welding techniques (FSW) to assemble thin aluminum plates in aerospace applications brought about the need for a high-resolution, non-destructive testing technology to find and characterize the small defects, such as kissing bonds, that may occur when using FSW. Phased array ultrasonics and eddy current arrays have been applied to FSW with success. This paper will discuss phased array ultrasonics and eddy current arrays and its application to friction stir welding of thin aluminum plates. Emphasis will be on detection of kissing bonds with both techniques.


This paper presents a novel NDT technology dedicated to the characterization of entrapped oxide defects (also known as kissing bonds) in aluminium friction stir welds. This technology was developed by TWI (UK), R/Tech (Fr.) and with other partners under the European Program Qualistir*.

The Industrial Objectives of this project are to overcome the current limitations of friction stir welding (FSW) technology by researching and developing:

  1. New NDT techniques and systems for the detection of 'kissing bonds' (lack of penetration defects) and other flaws (or defects) associated with the FSW process.
  2. New in-process monitoring systems to monitor FSW process parameters to improve joint quality by reducing manufacturing flaws/defects. The development of a NDT and in-process monitoring system, which is integrated with the FSW process will increase end user confidence in this new welding technology.
  3. Flexible FSW system for the application of the FSW process and its associated NDT and in-process monitoring system for the fabrication of 2 and 3 dimensional complex shaped welds thus increasing FSW applications.

This paper is dedicated to the presentation of the results concerning the ultrasonic NDT Technique.

Technique of inspection using ultrasonic phased array probe

The technique of inspection makes use of two phased array probes each being located on each side of the weld. Each probe is tilted at 26.4 degrees so to ensure a shear wave angle of 70 degrees in Aluminum part (cf following picture).

Representation of the inspection performed by


The qualities of ultrasonic phased arrays make it the most suitable solution to rapidly inspect FSW. Phased arrays use an array of elements, pulsed with different delays, to generate an ultrasonic beam. Once generated, the beam is identical to the beam generated by conventional transducers; phased arrays are merely a method of generating and receiving ultrasounds. The beams are generated and received with the use of focal laws, which define the elements to be pulsed and their delays. Software modelling programs and setup wizards are used to define the focal laws. For example, with the setup wizard, the operator defines the number of elements, the required refracted angle, the focal distance, the probe, wedge and material characteristics, and the computer calculates and displays the delays. The produced focal law is then used to generate and receive ultrasound. Different focal laws can be used to shape signal generation and reception.

Even though phased array probes are custom-built, they can be separated into a number of categories: circular, circular-annular, 2D matrix, and linear. Linear arrays are the most common because they are the cheapest and most versatile. Typical arrays have up to 128 elements, but there is no actual limitation in the number of elements, except price. There are physical limitations as to the size of the elements, the minimum being about 0.15 mm.

Phased arrays have big advantages over conventional UT. It is possible to change the beam angle for each pulse, thus sweeping the beam through a range of angles; this is called a sectorial or azimuthal scan. This type of scan isuseful to cover a range of angles for misoriented defects. Another type of scan, called a linear scan, is particularly useful for rapid scanning with linear arrays. Here the operator defines the focal law, then repeats it sequentially along the linear array to give very rapid coverage of the material to be inspected. Since R/D Tech phased array system FOCUS has a pulse repetition frequency of 20 kHz, electronic scanning is much faster than the equivalent mechanical scanning. If the system has a matrix (2D) transducer, it is possible to sweep the beam from side to side, known as lateral scanning. Lateral scanning is good for detecting misoriented defects, as in FSW inspections.

Elaboration of defective samples

A selection of the range of defects, which are significant in friction stir welding, has been selected in this project. This led to the manufacturing of samples (by TWI, GKSS, ALENIA) containing voids, pores, lack of penetration,and mainly kissing bonds. This paper presents the results of experiments with TWI samples.

14 plates of 1m length and nominally 7mm thick 7075-T351 Aluminium alloy were used for this section of the project; the thicknesses fell into 2 groups, 6.85mm and 6.95mm. The plates were each halved in length to produce ~0.5m welded samples.

To achieve the creation of the entrapped oxide defects and lack of penetration defects, a deliberately incorrect design of the FSW tool was developed.

As the FSW tools have fixed pin lengths, welding with a set pin length in the 6.95mm thick plates would leave an additional clearance of 0.1mm over that in the 6.85mm thick plates.

The usual tolerance for FSW to avoid a root lack of penetration defect is to have the pin within 0.1-0.2mm of the underside of the plate being welded. The initial pin was machined to 6.75mm and tried in pairs of plates that were6.85mm and 6.95mm in thickness, giving a 0.1mm and 0.2mm clearance respectively ( welds 1 and 2). After another trial using a different welding speed, which did not improve the weld surface ( weld 3), the rotation speed reverted to 350rpm and a trial was performed on 6.85mm plates in order to record the welding parameters, as the data from the welds may prove useful in analysing the defect assessment in NDT( weld 4, equivalent to weld 1).

The pin was then shortened to 6.55mm and welded through first 6.95mm and then 6.85mm plates, with 0.4mm and 0.3mm clearances respectively ( welds 5 and 6). Weld 6 showed a lower spindle torque than the previous weld records. Thereafter the welding was performed on the 6.95mm thick plates as the tool pin was shortened progressively by removing 0.2mm each time. Thus weld 7 used a 6.35mm pin, weld 8 a 6.15mm pin, and weld 9 a 5.95mm pin.

All of the welded plates in the series of varying clearances were subjected to a bend test. This is used by some fabricators as a check that there are no root defects present, while others rely only on tensile tests. The tests were conducted using a compliant strip over the weld sample to promote uniform bending, and after weld 1 had passed and weld 2 broken it was decided to perform interrupted bend tests, so that the failure had only started. In this way the initial stress raiser can be clearly seen, and welds 5-9 were tested in this manner. Macrosections were taken of the joints so that the visual record of the lack of penetration defects could be seen and quantified (cf weld T7 below containing a very thin kissing bond).


Figure : Left: Illustration of friction stir weld process - Right : Weld T7 (zoom) showing weld nugget structure (top) and coarse grain structure of the weld root (bottom). It can be observed that the pin penetration was of ~0.5mm from the bottom surface during the process(leading to the creation of thin entrapped oxide defects). This phenomenon can be monitored using ultrasonic phased array probes


After these trials 2 further specimens were welded in 6.95mm thick plates, with pin lengths of 5.75mm ( weld 10) and 5.55mm ( weld 11);

All plates were then machined to a smooth surface finish with a fly cutter, supporting the welded plates on parallels so that the machined surfaces would remain flat when placed in the water bath unclamped for ultrasound inspection.These welded plates and the bend samples have now been used to trial various ultrasonic techniques. Microsection photographs of the defects were taken.

The welding parameter records of welds 4-13 were recorded, for the progressive lengthening of the LOP defects. Micrographs were done on each side of each sample.

Processing ultrasonic signals for detecting entrapped oxide defects (kissing bonds)

The method is based on the fact that kissing bonds mainly occur because of a low penetration of the tool during the FSW process. This prevents the root region from being properly stirred (cf microphotographs). Because of that, the weld root contains many more little grains than the weld nugget. The weld root is therefore like the parent metal which is a noisier structure than the weld nugget. That difference in structure can be observed using ultrasound (cf following figure).


Figure: Top view (left) and side view (right) of the phased array inspection results. It appears clearly on the side view that the weld nugget signal is less noisy than the parent metal and weld root (light blue represents lower amplitude than dark blue).The technique hereby described consists of comparing mean signal M1 of weld nugget over mean signal of weld root M2. When the pin has penetrated deep enough, M2/M1 is close to 1. Decreasing values of M2/M1 indicate a less deep penetration and a higher probability of having entrapped oxide defects.


Figure: Illustration of the difference in ultrasonic noise:
histograms of amplitude level for weld nugget volume (red) and weld root volume (blue). the mean value of the nugget signal is lower than that of the root because the nugget is correctly stirred. Red histogram is narrower illustrating the correctly stired structure of the weld nugget.

When the root has been properly stirred, the ultrasonic noise level inside the root should be close to that of the weld nugget. By comparing the mean level inside the root to that of the weld nugget we provide the operator with a powerful tool for the estimation of the pin penetration and therefore to the probability of having kissing bonds.

The following figure is an illustration of the inspection of sample T7 containing entrapped oxide defects all along the weld. Only a few indications can be observed by pulse echo technique (red circles).


Figure: CScan image of FSW inspection:
the plate T7 contains kissing bonds (cf above micrograph)

Even if a few indications can be detected at the weld root (D1 to D3), it is difficult to say whether there is a kissing bond all along the weld or not. This can however be achieved by comparing the ratio M2/M1 of this plate to that of a reference plate containing no defect.

The following graph compares 2 curves: the red curve represents the ratio M2/M1 of T7 sample (containing kissing bond) and the blue curve represents the same ratio with sample T1 containing no defect.


Figure: Ratio M2/M1 along with the weld for sample of reference (blue curve) and defective sample (red curve) containing a very tight kissing bond observed on the microphotography)

The mean value of the ratio M2/M1 is more important when the sample contains kissing bond defects. Also, the deviation from the mean value of the ratio M2/M1 is much more important when the sample contains a kissing bond.

The difference in mean value between the weld nugget and the weld root was observed for most of the samples containing kissing bonds. The following figure represents the value M1 and M2 for the 13 samples. Samples 1 and 2 have no defect while all the other samples contain kissing bonds. It can be seen that the value of the root weld M2 tend to be equal to that of the parent metal. This indicates that the root has not been properly stirred (as the parent metal)which increases the probability of having kissing bonds.


Figure: Mean signal inside the nugget and inside the root for different samples


As a conclusion, many efforts were conducted under the program Qualistir for improving the inspection of kissing bonds in FSW using ultrasonic phased array probes. Samples containing different types of defects including entrapped oxide were manufactured. The inspection of all the samples led to the development of a new processing based on the comparison between the weld root ultrasonic signal and the weld nugget signal. Preliminary results tend to demonstrate that this processing provides a good indication of the presence of very tight kissing bonds. Further experiments will be conducted under Qualistir Program using other samples to improve the reliability of the processing. This processing is planned to be integrated into Quick view Software (provided by R/D-tech) for online monitoring.

For more information please email: