U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Bridges & Structures


Technical Advisory

Quality Control and Quality Assurance Inspections on Welded-Steel Fracture-Critical Members

November 27, 1979

Technical Advisory 5140.11


This Technical Advisory (TA) recommends certain practices and procedures for the nondestructive inspection (W) of welded-steel fracture-critical members in highway structures at the time of fabrication.


  1. The Problem in Terms of Fracture Mechanics
    1. There is a recurring problem with cracking in welded steel plane-girder bridges. The cracks range from a fraction of an inch at the time of discovery (stable, subcritical cracks) to several feet long (sometimes discovered only after a brittle fracture has occurred).
    2. In terms of the life of a bridge, the timing of the transition from a subcritical to a critical crack and the ensuing brittle fracture is a function of the fracture toughness of the steel (plate and/or weld) and the working stress in the cracked member. A "pop-in" crack is an unstable crack in its immediate environment, say in an embrittled weld metal or in an embrittle weld heat-affected zone. Whether the pop-in crack remains unstable after it leaves the localized zone of embrittlement and high residual stress depends on the strain-rate sensitivity and/ or the dynamic fracture-toughness of the steel plate, and the stress field at the crack tip.
    3. The present American Association of the State Highway and Transportation Officials (AASHTO) material toughness specification and the 1978 AASHTO fracture-control plan provides no protection against pop-in crack.
    4. Protection against a pop-in crack which originates either in a zone of local embrittlement or from an existing, undetected, crack which may grow to critical size in the stress field of the weld, requires:
      1. elimination of zones of local embrittlement (a workmanship problem which may be the result of human error); and
      2. better control over the nondestructive inspections carried out in the fabricator's quality-control (QC) operation and in the owner's quality-assurance (QA) procedures.
    5. This TA addresses paragraph 2a(4) (b), viz., improved QC/QA.
  2. The Problem in Terms of Quality Control
    1. Present practice usually either specifies radiographic testing (RT) as the NDI method for groove-weld butt splices or leaves the fabricator the freedom to choose between RT and ultrasonic testing (UT).
    2. Many fabricators are aware that there are two "problems" with the UT method. First, there is a "people problem" with UT; when the operator lacks the necessary training and/or skills he or she may either completely miss rejectable defects or find "rejectable" indications that are not real. Second, in the hands of a skilled conscientious person, the UT method will find cracks that will never be seen with RT. Either way the fabricator may incur unwanted expense and delays. On the other hand, RT provides a permanent record which can be read and reread at will. All factors considered, most fabricators are choosing RT rather than UT.
    3. On more than one occasion, RT has left the fabricator unaware of weld cracking that eventually was found to be so serious the fabricator and/or the owner suffered financially from the expense of repairs.
    4. Unfortunately (for the owner), all too often cracks are only discovered after the bridge has been erected. There have been several bridges found to have cracks in the 1970s; the cost of repair is very great once the bridge is under traffic. The cost is not only in dollars to retrofit/repair but also in time and inconvenience to the traveling public. Sometimes, there is litigation involving the owner and fabricator/erector; this also is very costly in terms of unproductive staffhours. Clearly, the best and cheapest solution is to utilize effective QC in the fabrication shop where necessary repairs can be made under favorable, least-cost conditions.
    5. Planar defects including cracks, incomplete penetration and lack of fusion are both dangerous and prevalent at the ends of welds at plate edges. Such defects are dangerous because the inherently high stress intensity and high residual stress make growth likely by fatigue and crack instability. Such defects apparently occur because run-off tabs are either too short or are not utilized for the full length of the tab; consequently, the start-up welding conditions may not have stabilized and planar defects are produced. The RT methods usually employed are not effective out to the edge of the plate nor is UT. The most effective NDI methods for ends of welds and plate edges are magnetic-particle testing (MT) and dye-penetrant testing (DT). However, DT will not detect subsurface defects and MT is only useful for defects in or very close to the surface. Unfortunately, it has been found that the MT method which uses copper prods can introduce surface cracks by depositing copper in the grain boundaries of the plate (or weld) at the point of contact between the copper prod and steel. An MT Yoke avoids this problem.
  3. The Problem in Terms of Quality Assurance
    1. Present practice often involves a passive approach to QA; i.e., the person(s) acting on behalf of the owner to assure the adequacy and effectiveness of the QC operation simply watches the fabricator's QC without any hands-on QA auditing. By hands-on auditing, the QA operation actually performs UT and/or RT to verify some designated percentage of the welds in the fracture critical member (FCM).
    2. Hands-on auditing, to be effective, must be a contractual requirement; i.e., the fabricator in signing the contract must agree to be responsible for correcting an rejectable indications found in the QA audit. Controversy sometimes develops over how much auditing should he done. Again, this must be clearly state in the contract special provisions. When the QC is based on RT and QA is based on UT, there is sometimes a lack of correlation between the two methods. This is generally a "people problem" where one or the other test method is conducted by people lacking in training or skills. When UT detects planar indications and RT shows nothing, this may be a question of a tight crack too fine to be detected with conventional RT techniques (as in American Welding Society (AWS) Structural Welding Code AWS D1.1, Section 6, Inspection). It should be recognized that RT is not as capable of revealing tight weld cracks as UT, unless much improved RT methods are used. A strong argument can be made for requiring UT in both QC and QA.
    3. Clearly, if delays in fabrication are to be avoided, QA must be performed on a scheduled basis, and as soon after QC NDI as possible. But here is another source of trouble. Much has been said and written about hydrogen cracking - a form of delayed cracking; nevertheless, fabrication procedures and consumables handling procedures continue to be a source of delayed cracking.
    4. Periodically, fabricators continue to experience the phenomenon of "cracks that aren't there at the time of QC" but are unmistakably present sometime later on QA auditing. There are cases on record where all parties agree that "delayed" cracking had occurred. This is NOT confined to high-strength 100 ksi yield steel; it can occur even in A36 steel, although not as likely. Choice (strength) of electrode is an important factor as well how the electrode protected from moisture pickup. The higher the strength of the weld deposit (and weld heat affected zone) the greater the propensity to delayed cracking. Clearly 24 hours between welding and inspection in insufficient and even 48 hours is questionable. Cracking should be expected to continue for up to three days after welding; researchers have recorded even longer periods of delayed cracking. If QC is done too soon, then QA is likely to find cracks that had not been formed when the original QC was done. If the contract has provisions for escalating the audit percentage when cracks are found, then the fabricator will be well advised to wait at least 48 hours before performing final QC ND1.
    5. The phenomenon of delayed cracking produces doubt on timing of crack development and responsibility when cracking is discovered in an in-service bridge. Indeed, the question properly arises as to when the cracks were initiated and whether the cracks are growing. Over and over again these important, if not vital, questions arise when in-service bridges are found to contain cracked welds, and the answers may affect the safety of structure. If non-rejectable UT indications had been recorded by QA/QC at the of fabrication, reappraisal of the designated locations may provide an answer to the all-important question of whether cracks are growing in the bridge.
    6. At the present time neither AWS D1.1 nor AASHTO require that non-rejectable indications be recorded in QC or QA. Because the recording of non-rejectable indications (even within specific limits) is a costly operation, the owner may wish to limit the recording of such non-rejectable indications to one or two per butt weld.


  1. Quality Control (QC) should be the responsibility of the contractor; Quality Assurance (QA) is the prerogative of the engineer. In fracture-critical members (FCMs) QC should be randomly verified by QA performing actual nondestructive testing of welds previously found acceptable by QC.
  2. If a weld is found to contain rejectable planar-type indications in verification sampling (auditing), the two consecutive welds of the same type (same welding specification) preceding the defective weld should be verified by QA. If two consecutive welds are found by QA to contain rejectable planar-type indications, four consecutive welds made to the same welding specification should be verified by QA, etc.
  3. The contractor should be required to use both RT and UT for the QC inspection of groove-weld butt splices in FCMs. The QA auditing should be done with UT. In the event that QA auditing discloses rejectable indications, repair and reinspection should be performed at the contractor's expense.
  4. Plate surfaces and plate edges at the ends of groove-weld butt splices should be MT by the dry-particle yoke method and/or DT. Likewise, wherever there are intersecting groove-weld butt splices, the MT (Yoke) and/or DT method should be used in addition to RT and UT.
  5. The MT by the prod method should be prohibited on FCMs.
  6. The UT report should record not only all rejectable indications but also selected non-rejectable indications with defect severity ratings within 5 db of being rejectable. The latter should be fully recorded as to indication rating, size, and location.


  1. While this TA is written primarily for fracture-critical members, it must be remembered that most of these recommended practices are equally applicable to the inspection of redundant members. The engineer in charge should utilize judgment on the application of the specific requirement to individual members depending on the degree of sensitivity involved.
  2. While critically in terms of safety is not so severe for redundant members, the cost implications still exist in such cases.


  1. Designs on future projects which utilize this type of detail should avoid bolting the tie plate to the longitudinal girder.
  2. Existing structures which have this detail should be carefully examined for evidence of cracking and the need for repairs. It is recommended that the bolts or rivets which connect the tie plate to the longitudinal girder be removed, and that the holes be protected against corrosion through use of smaller bolts and suitable washers or plates.

Rex C. Leathers, Director
Office of Engineering

Updated: 06/27/2017
Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000