American Welding Society ...

Raising the Bar for Inspections of Steel Moment–Frame Connections
Ultrasonic technicians will have to perform to a higher level if they are to meet the requirements of FEMA 353 for inspection of welded connections used in locations prone to seismic activity


Fig. 1—Typical welded moment–resisting connection prior to 1994.
Steel moment–frame buildings are designed to resist earthquake ground shaking. The assumption prior to the January 1994 North Ridge, Calif., earthquake was that steel moment–frame buildings would be capable of extensive yielding in the presence of an earthquake and any damage would be expected to consist of moderate yielding and buckling of the steel elements, not brittle fracture. Steel moment–frame connections, as shown in Fig. 1, were also predicted to be capable of withstanding a great deal of plastic rotation without significant reduction of strength.

The North Ridge earthquake was found to have damaged steel in buildings with moment–frame connections. While the building structures remained essentially elastic, the observed damage caused by the earthquake was the result of brittle fractures initiated within the connections. Most of the fractures occurred in the complete–joint–penetration weld between the beam bottom flange and column flange —Fig. 2. FEMA 353, Recommended Specifications and Quality Assurance Guidelines for Steel Moment–frame Construction for Seismic Application, includes photos of the damage found after the earthquake (two examples of which are included here in Fig. 3).

Fig. 2 — Common zone of fracture initiation in beam–column connection.
The bottom flange to column flange weld joint is often poor in quality due to various fusion–related discontinuities. Some of these discontinuities, especially those at critical locations, are crack initiators when the connection is subjected to severe stress and strain. The basic configuration of the welded connection, which utilizes a backing bar, makes it extremely difficult to detect discontinuities within the weld and at the root. Prior to 1994, backing bars were typically left in place following weld completion. The bars prevented visual examination of the weld root and proper interpretation of ultrasonic indications due to backing bar–related reflectors.

Over the years, designers have relied on ultrasonic testing (UT) as the principle NDE method for examining complete– penetration weld joints. Ultrasonic test methods and acceptance criteria have been established in AWS D1.1, Structural Welding Code —Steel, for structural weld joints; however, in July 2002, the Federal Emergency Management Agency (FEMA) issued additional UT guidelines for steel moment–frame connections for seismic applications, the FEMA 353 document mentioned previously. Appendix E, "Supplementary Ultrasonic Technician Testing" of FEMA 353 was developed as a method to further the qualification of UT technicians in the detection and sizing of typical flaws in steel moment–frame welded joints. The basis of FEMA 353 Appendix E is summarized in FEMA 355B, State–of– the–Art Report on Welding and Inspection, Chapter 7, "Inspection and Acceptance Criteria."

Fig. 3 — Brittle fracture initiated within weld connection between beam bottom flange and column flange in steel moment-frame building damaged during the 1994 earthquake in North Ridge, Calif.
Specifically, FEMA 353, Part I, Chapter 5, requires that moment–frame welded connections be scanned to the procedures prescribed in AWS D1.1, Annex K for UT flaw detection and sizing. Also, ultrasonic technicians are required to demonstrate proficiency on weld joint mock–ups that include butt, corner, and T configurations. The mock–ups must include at least 20 known intentional reflectors that represent typical flaws and orientations.

Getting Started
Compliance to the FEMA recommendations was untested until the spring of 2002 when a federally funded project in Seattle imposed the FEMA recommended specifications. NDE Professionals, Inc. (NPI), was assigned the task of developing a UT practical examination and test samples for the Northwest Council of Engineering Labs (NWCEL). The examination program was to reflect the FEMA program recommendations. After conferring with the project quality assurance representative and the NWCEL, it was decided that the initial practical examination be modified to first evaluate the technician competency to AWS D1.1, Section 6, Part F, for flaw detection. If the technicians' competency proved to be adequate, additional steps would be made to further develop the program to include the full Annex K evaluation method.

Prior to the initial examination, three ASNT UT Level III individuals evaluated the first eight test samples selected for the examination. The initial test samples had a nominal thickness of 0.600 in., represented only butt and T configurations, and had no backing bars. The Level III group individually verified all the known reflectors in the test samples and compared their results. The Level III evaluation recorded each reflector/indication detected with regard to dB rating, depth, length, and XY location. Although each Level III found all known indications, a consensus as to the exact value for each indication attribute (dB, depth, length, and XY location) could not be reached. It was decided a preliminary "range" or "tolerance" need be established for testing purposes (ref. Table 1, Initial Range). NPI intended to evaluate all the data after the initial examination and validate or adjust the tolerances as appropriate.

A draft procedure was authored that set the parameters of the exam. The scoring was to be as described in FEMA 353. In March 2002, a group of 12 ultrasonic Level II technicians from the Seattle area (Group 1) was given the first set of practical examinations.

Grading the Examination
Scoring of the examination was based on three factors:

  • Number of detected indications
  • Number of missed indications
  • Number of false indications.

In order to receive credit for a detected indication, the detected indication attributes had to fall within the established attribute tolerances. If any indication attribute was reported to be outside these tolerances that indication was determined to be either a missed or false indication. If an indication was reported in the general vicinity of a known indication but the attribute values were not within range, it was considered a missed indication. Once all the indications were classified into one of the three categories, they were then plugged into the FEMA recommended formula for scoring. This formula contains a "confidence factor" that takes into account the number of discontinuities included in the test. The test candidate must receive a score of 80 or above in order to pass the examination. The scoring formula with confidence factor is as follows:

where R = UT technician rating, n = total known flaws, D = total detected flaws/n, F = total false indications/total reported Indications (Confidence Factor: n–2/2(n–1)).

Disappointing Results
The results of the initial examination were disappointing. None of the 12 passed the practical examination using the established attribute tolerances. As intended, NPI analyzed all the data. One basic question needed to be answered: Was it the examination or the technician? It was obvious some test candidates lacked basic UT fundamentals; however, the wide variation in the attribute values reported by the technicians suggested the examination parameters needed to be evaluated.

The next examination group (Group 2) was scheduled to be examined in November 2002, so the following procedure modifications were made:

  • An eight–hour class was conducted the day prior to the examination that reviewed basic AWS UT techniques and acceptance criteria.
  • Attribute tolerances remained unchanged but one of the five attributes per indication was allowed to be out of tolerance for each indication reported by the technician (Table 1).

  • Six additional test samples were acquired that fully represented all moment–frame configurations, including backing bars.
  • Each test candidate was required to use a controlled set of transducers, cables, calibration blocks, and couplant that the examiner supplied specifically for the examination.

Group 2, which consisted of 12 test candidates, was examined to the new test parameters. The results indicated only 17% passed the exam.

Evaluating the Group 2 Data

Of the 12 test candidates, 2 passed the practical examination with a score of 80 or above. The evaluation of the test data revealed some interesting statistics.

It is clear the examination failure rate of 67% is a cause for concern

    Only 17% of the test candidates reported the laminar–type discontinuities present.
  1. Thirty–three percent reported a discontinuity in the general area of a known transverse crack, but only 8% identified and correctly reported the transverse–type discontinuity.
  2. Twenty percent of the known discontinuities had a low detection rate, where less than 66% of the test candidates correctly identified and reported these discontinuities.
  3. One of the seven test specimens was included in the examination in order to obtain a consensus evaluation of backing bar root–type discontinuities. Only one individual located the known discontinuity in this sample within the newly recommended tolerances. This sample was not included in the examination grading.

The biggest technical error observed while proctoring the examination was the reporting of discontinuities that were not fully "peaked" in the first leg —Fig. 4. The lack of signal peaking was due, primarily, to the interference of the weld crown. This signal amplitude peaking problem causes errors in discontinuity location (depth) and indication rating (dB). Most of the indications may have been properly reported from the second leg (Fig. 5) or, in the case of butt joints, from the other side. As a result of the first leg peaking problem, it is believed that a lot of single indications may have been reported two or three times as separate, discrete indications, resulting in a high incidence of false indications.

Fig. 4 — Weld crown inhibits movement of search unit. Centerline of sound beam is unable to reach point of maximum reflection from slag inclusion. Signal is not fully peaked in first leg.
The FEMA formula includes a statistical confidence factor based on specific sample size. The "n" in the formula is the number of "known flaws." If n is 20 flaws. the factorial (n–2/2(n–1)) equals 0.474. If an inspector performed the "perfect" test by properly detecting and evaluating all 20 flaws, with no indications missed and no false indications, the inspector's score, as determined by [(n–2/2(n–1)]X(1+D–F), would be 95. To achieve a score of 100, the inspector would need a "perfect" test on a minimum of 201 flaws as determined by conventional rounding to the nearest percent.

Test Samples
The backing bar used to make structural welds can be a source of nonrelevant indications that are extremely difficult to interpret and evaluate. The AWS D1.1–2002 Code includes additional guidance for scanning groove welds containing steel backing. These special techniques involve increased inspection time and often involve grinding the weld and/or full or partial removal of the backing bar. This is not a practical scenario when performing UT in the field. It is not practical to allow technicians to modify samples used in a practical examination. This makes it necessary to provide the technicians with test samples that can, in fact, be evaluated without the use of extraordinary means. Therefore, backing bar welds used for examining technicians require even greater validation than those without backing. It may require the company preparing the examinations to weed out many samples before they have a valid set. This was the case with one NPI backing bar test sample that was determined to be invalid for examination purposes.

Fig. 5 — Point of maximum reflection is reached in second leg.
Acquiring, validating, and maintaining an adequate number and variety of test samples for examinations/reexaminations can be very costly; however, they are necessary to ensure examination integrity.

Although preliminary testing results are inconclusive, it is clear the examination failure rate of 67% is a cause for concern. Even utilizing "controlled" test equipment, the indication rating, depth, and location varied significantly from technician to technician.

The improved passing rate with Group 2 can be mainly attributed to the modified indication attribute tolerances, but the consistent variation of principle indication attribute values can be critical with regard to accept/reject criteria. For example, in AWS D1.1, Part F, the acceptance of a discontinuity is based on a calculated decibel (dB) rating system and the length of the discontinuity. The difference between a Class A and a Class D indication on a typical moment–frame connection, with a thickness of 112 inches or less, is 4 dB. There is only a 1–dB difference between each of the Classes A, B, C, and D, with A being a reject weld indication and D an acceptable weld no matter what the length of the indications. Having a tolerance of "+4 to no limit" would indicate the length of the indication might not really be a factor. However, the tolerances were arrived at as a compromise based on the current state of the art and what is thought to be reasonable and achievable results.

Based on the damage created by the North Ridge earthquake and the fact that flaws that were thought to be detectable by UT initiated some of the damage, it certainly is time for change. However, change is slow in coming and requires a process consisting of baby steps. Having reviewed the final tolerances with the NWCEL, NPI believes the approach presented here is one step in the right direction.

Preexamination training definitely showed some benefit. While many technicians exhibited a lack of UT basics, others lacked confidence, and still others were obviously intimidated by the examination itself. Training is considered an essential part of future examinations. The errors in the application of the UT technique observed can be corrected through training.

The current UT mythology has inherent problems that can be improved on by moving the industry toward UT techniques that improve reliability of the test such as the FEMA recommended AWS D1.1, Annex K, for UT flaw detection and sizing. This change may require increased inspection cost, as some of these techniques are more labor intensive.

Designers, engineers, and contractors can help by providing welds that can be more reliably inspected. The primary source of nonrelevant indications and errors in interpretation of indications is the steel backing bar. When steel backing bars are employed in making welds that are to be UT examined, the removal of the backing bar will significantly improve inspection results.

Steel construction has come to rely on ultrasonic testing and UT remains the preferred NDE method in testing complete–joint–penetration welds. It is the responsibility of the NDE community to provide confidence in the test method and to ensure test personnel are qualified. Addressing the problems and investigating possible solutions is the only way to accomplish this goal.

LEE GARRISON — LEE@QNPI.COM — is with NDE Professionals, Inc., Portland, Oreg.