Introduction
Examples of good, undersize and stick spot welds were solicited from our industrial clients. The decision of what constituted an "undersized" weld was left to the manufacturer. For suppliers of automotive sub-assemblies, this meant producing a weld with a nugget of insufficient size. The manufacturers of products not made according to such requirements based this decision on their own guidelines of quality control.
The response was strong: five manufacturers provided us with 18 welds (6 good, 6 undersize, 6 stick welds) from 33 different combinations of material, thickness and coatings. In total, 594 welds were submitted for testing.
From the literature, we knew that 4-point resistivity testing and ultrasonic inspection using through-weld compressional waves were the two most widely used methods. Our efforts focused on these techniques.
Nondestructive Tests
This technique had been extensively researched by AT&T. Various publications documented the successful application of this test method for testing spot welds on automotive steel, stainless steel and aluminum. Such documentation notwithstanding, some manufacturers contacted by the IDL had previously tried to implement this technique, with little or no success. Additionally, in-house experience at the Iowa Demo Lab suggested only sporadic success. This project attempted to reconcile this disparity between reported results and anecdotal information.
In this approach, a probe with 4 spring-loaded contacts is pressed against the test sample. The orientation of the probe across the spot weld is shown in the diagrams below (taken from the AT&T Microhmeter User's Guide).
As the probe is held against the spot weld and activated, direct current passes through the outer contact tips; the inner points measure the resulting voltage. With this information, the resistivity across the weld may be calculated. Absolute resistivity measurements are possible, but in this study we calculated the ratio of the resistivity of the weld to the resistivity of the base metal. In our tests, we took two readings on each of the 6 welds of each classifications: good, undersize, and stick. An overall average was then calculated as a representative value for that group of welds.
Our approach in this study focused on careful attention to technique. The probe tip was steadied on the weld by the operator using one hand; it was triggered with the other. The trigger was held until the activation light acknowledged the reading. No casual probe alignment was permitted, and the sample/probe contact was carefully handled . These may seem to be trivial points at first glance. They are not. Some support staff at the IDL (part time students), as mentioned in the introduction, had been previously unable to obtain consistent readings on a given set of samples. When their implementation technique was closely evaluated, and changes consistent with the guidelines just stated were made, marked improvement in repeatable readings was achieved.
To critique our data for validity, it was expected that a "well behaved" sample would exhibit certain trends in the resistivity ratio. A good weld would provide a significantly larger path for the injected current flow and should show the lowest resistivity. An undersize weld was expected to provide less additional path for the current, and the resulting resistivity ratio should be greater. And a stick weld, with very little bonding between the layers of material being welded, should show a resistivity approaching that of the base metal, or a resistivity ratio approaching unity.
With this criteria in mind, we reviewed the resistivity ratios for our samples. Again, we expected the results to show an increasing resistivity ratio as we went from good to undersize to stick welds.
Good results were obtained on the steel samples from one
manufacturer. These samples were essentially the same kind of steel, but
in various thicknesses. Some difficulty was encountered in separating good
from undersize welds, but overall the trends observed were as expected.
Our tests also showed better results on cold rolled steel
than on HSLA steels (left image below), while some heavier gage steel samples
(right image below) gave mixed results.
Hot dip galvanized material gave better results than some
welds on cold rolled steel produced with capacitive discharge (below, left)
while all of the welds on aluminum yielded good results (below, right).
Additionally, some stainless steel samples proved untestable in that the resistivity readings appeared to "saturate" and precluded our getting any meaningful readings from them. The reason for this is not immediately known, and in fact is contrary to results in some published literature.
All told, of the 28 combinations of weld samples that
tested adequately, 20 groups produced expected trends. In other terms,
70% of the samples studied in this report yielded results that correlated
with expected behavior. Indeed, when sample groups from individual manufacturers
were looked at after destructive peel tests to reveal weld nugget diameter,
some very good correlations were found between the nondestructive test
results and weld nugget size.
The plot above shows a family of calibration curves developed on aluminum samples of varying thickness. As nugget diameter increased, resistivity ratios were seen to decrease.
To conclude our appraisal of the 4-point resistivity test,
it would appear that some samples can present problems for using this method
of determining weld quality. However, the technique also appears to suffer
from an undeserved reputation, based on our experience. The test cannot
be implemented in a casual or haphazard fashion. It is possible that the
very simplicity of the equipment design, complete with green and red go/no-go
lights deceives the operator into an unattentive mode of application.
One significant caution, however, must be raised. This method seems to have a very limited commercial representation with respect to equipment vendors. While 4-point resistivity devices that measure bulk resistivity in silicon wafer applications areavailable from several vendors, it does not seem that production of such equipment for weld testing has a robust marketplace. This could conceivably lead to poor availability and/or service of the equipment.
- Develop a generic approach for the interpretation of waveform features that would distinguish one weld from another
- Validate this approach on our broad spectrum of samples
- Optimize equipment settings by linking pulser/receiver gain settings to the highest number of correct weld interpretations using the above criteria
- Determine if transducers other than the captive water column typically used could provide good results while being easier to implement
As with resistivity testing, many instances of the successful application of ultrasonic inspection for spot welds are available in published literature. What appears to be lacking, however, are tangible guidelines for such things as selecting the optimum gain setting on the ultrasonic pulser/receiver and quantitatively setting limits to features in the waveforms obtained on different welds. Our approach in this portion of testing had several goals:
The use of ultrasound to interpret weld quality is based on the detection of reflections from the weld interface along with attenuation of through-weld signals due to coarse grains in the weld nugget. In very simple terms, the images below illustrate the signal features obtained on various welds and seen on an ultrasonic flaw detector.
It should be seen that a good weld will exhibit reflections form the backwall of the welded structure, with peaks visible that correspond to roughly twice the thickness of an unwelded region. As the weld becomes undersized, smaller reflections will be detected from the weld interface, with peaks occuring on the time/depth scale that correspond to a single thickness of material. These interfacial reflections will be stronger as the weld nugget gets smaller, leaving more unwelded material to reflect the signal back to the transducer from a single thickness depth. Additionally, a weld nugget's thickness will likely increase as its diameter increases. This increase in coarse-grained material will act to decrease the amplitude of the through-weld reflection due to attenuation effects.
The position of the through-weld signals can closely be predicted from the backwall reflections on unwelded material: every even-numbered multiple reflection of this single distance will correspond to a through-weld signal, with odd-numbered signals considered interfacial reflections. (This assumes that the operator will be able to distinguish the reflection that will always occur where the transducer contacts the weld surface.)
These basic principles can seem to over-simplify the challenge to adequately inspect randomly chosen welds. What is missing from the equation is a guide for establishing an appropriate gain setting on the ultrasonic instrumentation. In our previous work on spot weld inspection, it was determined that a rnage of gain settings could be employed, coaxing a variety of changes in signal characteristics. As the images below show, a reduced gain would tend to minimize interfacial reflections, and correspond to an increase in false accepts, or bad welds that pass inspection. Alternately, higher gain settings could accentuate these reflections, and lead to an increase in false rejects. This was precisely our experience on one manufacturer's welds.
It should be apparent that an intermediate gain setting should provide an optimized signal, with the highest number of correct calls. In our research, a link was noticed between this point and the characteristics of a backwall reflection on unwelded material.
We monitored the signal characteristics of single-thickness backwall reflections at various gain settings, while correlating features of weld signals with the results of destructive tests. It appears from our work that the optimum gain setting for spot weld inspection occured when the 2nd backwall signal, on unwelded material, was saturated at 100% FSH, and then the gain was increased by an additional 12 dB. Schematically, it works this way:
The next step in our evaluation of an optimized weld inspection technique dealt with developing a simplified means of assessing the features from different signals. It was desired that the algorithm be easy to implement, but of course discriminate between different weld conditions adequately.
We obtained weld signals from all of our welds, while following the gain setting guideline described above. We then looked at the peak amplitudes of the various echoes corresponding to 1 thickness, 2 thicknesses, etc. (1st echo, 2nd echo, etc.) and entered these values into our database. For each weld condition, in each category, we determined the average values for each of these echoes.
Looking at the interfacial reflections (1st, 3rd, 5th, 7th echoes) we determined the maximum value from that weld. We then evaluated the peak amplitudes of the three weld classifications (good, undersize, stick) by using logical qualifiers such as "greater than" or "less than." For example, we determined if the maximum interfacial reflection for the stick weld was greater than that for the undersize weld, and in turn if the maximum interfacial reflection from the undersize weld was greater than that obtained on the good weld. If both conditions were met, we deemed this weld group a "correct call" as far as interfacial reflections were concerned.
We then looked at the through-weld reflections (2nd, 4th, 6th, 8th echoes). The average value for these peaks was obtained in each weld classification, and a linear regression fitted to this data. The value of the slope of this regression was evaluated in our database. The slope of the line fit to even-numbered echoes of the undersize weld shouldless than the slope for the same echoes on agood weld, due to attenuation effects. If this was the case, we considered this weld group a correct call for through-weld reflections.
During our tests, we also employed solid delay lines on standard diameter transducers. We tried to match the size of the transducer element to the anticipated nugget size, but this was established in a fairly easy manner. We experimented with delay lines "off the shelf" as well as ones that had their ends chamfered or rounded over. While we will not go into excessive detail here, the solid delay line that had a 1/32" radius round-over to break its edge worked best in our study. After being proven in some initial tests, this delay line was used for the bulk of our work.
We determined that 60% of our welds could be correctly tested and characterized by applying the instrument set-up guidelines we described, and submitting the data to the analysis procedures outlined here. Permitting some minor adjustments to be made to our signal feature evaluation algorithms, this level rose to 87% correct weld classifications. Recall that this approach was aimed at a testing any generic spot weld. Refinements to the evaluation criteria on a case-by-case basis should improve this percentage.
We feel that this work represents a significant step toward assisting manufacturers in implementing spot weld inspection. It provides guidelines for a weld inspector in equipment set-up and in signal characterization. And the most significant features of the waveform analysis presented here should be readily imlemented on a flaw detector by the use of threshold detection gates and the use of overlays and/or grease pencils to record peak decay.
The following image is submitted to emphasize the main points of our work with ultrasonic inspection of spot welds:
Conclusion
We hope that the material we have presented here is helpful in guiding quality engineers and weld inspectors who are concerned with the evaluation of spot welds. Other techniques than the two we highlighted here are in various stages of developement, but these two methods have certainly been documented the most. Of the two we discussed, the ultrasonic approach perhaps holds the highest promise of versatility. Users will no doubt develop refinements to our general approach for a particular set of welds. But our overall excellent results, obtained on such a broad spectrum of welds, is quite encouraging.
