A recent Machine Design article, “The basics of fatigue in welded steel structure,” made reference to the fact that extensive research has been conducted by various organizations in a number of countries over the last several decades. Here’s a brief overview of that research work.

A brief history

Due to the severe nature of the fatigue failure mode, much research has been conducted over the past few decades in an effort to both predict service life and minimize failures as much as possible. In September of 1986, the Transportation Research Board’s (TRB) National Cooperative Highway Research Program (NCHRP) issued Report 286: "Evaluation of Fatigue Tests and Design Criteria on Welded Details."

This report was sponsored by the American Association of State Highway and Transportation Officials (AASHTO) and authored by two of the leading experts in this area, P. B. Keating and J. W. Fisher. It is arguably the most significant and comprehensive single document on fatigue design of welded structures. Here is a brief synopsis of the major research summarized by this document:

Data from 1966 to 1974

A large scale research program was conducted over this eight year period. (For more information see NCHRP Report 102: "Effect of Weldments on the Fatigue Strength of Steel Beams" issued in 1970 and NCHRP Report 147: "Fatigue Strength of Steel Beams with Welded Stiffeners and Attachments" issued in 1974. Results from a database of over 800 tests were statistically analyzed and used to develop the original AASHTO specifications for fatigue design of welded steel bridge structures. These tests consisted of various welded joint details subjected to cyclic stresses on a test fixture until fatigue failure was imminent. The main goals of this program were:

• Reveal the parameters that were significant in describing fatigue behavior (Report 286).

• Quantify the fatigue strength of welded bridge details (Report 286).

Amazingly, only two major parameters emerged:

• Stress range, and

• Joint detail.

As will be discussed in the quantification details later, this is a rare case in which an extremely complex problem lends itself to a relatively simple analytical solution.

The study did not, however, attempt to “evaluate the adequacy of weld improvement techniques.” The main intent was to “define the lower bound fatigue resistance for as-welded details, or the minimum level of fatigue strength that would be obtained provided that standard fabrication and inspection procedures were employed.” (Report 286.)

Although it is beyond the scope of this article, much research has also been conducted on several weld-improvement techniques that have been shown to improve fatigue life, such as shot-peening and other cold working techniques, and thermal stress relieving.

Three types of steels were used during the test phase (Report 286):

ASTM A-36 "Standard Specification for Carbon Structural Steel."

ASTM A-441 "Specification for High-Strength Low-Alloy Structural Manganese Vanadium Steel.” (Note that this standard was withdrawn in 1988 and replaced with A-572 "Standard Specification for High-Strength Low-Alloy Columbium-Vanadium Structural Steel."

ASTM A-514 "Standard Specification for High-Yield-Strength, Quenched and Tempered Alloy Steel Plate, Suitable for Welding."

It is important to remember that the fatigue life of welded structures is independent of material strength. In fact, in some cases, the higher-strength class A-514 proved to be less fatigue resistant than lower-strength grades (Report 286). This most likely stems from the relative weldabilty of these materials.

“The tests demonstrated that all fatigue cracks commenced at some initial discontinuity in the weldment… and grew perpendicular to the applied stress… These discontinuities are always present, independent of the welding process and techniques used during fabrication.” (Report 286.)

Data from 1974-1986

During this period, even more testing was done through the NCHRP and other American organizations (such as International Union of Railroads.) Also, similar work was done in a number of other countries, including Japan, England, Switzerland, East Germany, West Germany, and Canada. Also, some of the major standards organizations did their own studies which agreed with these findings, including the ECCS (European Convention for Constructional Steelwork) and ISO (International Standards Organization).

As all of this new data was carefully compared to the fatigue relationships that had already been developed, substantial agreement and consistency was found. This, of course, precipitated even more confidence in the fatigue curves that had been originally adopted by the AASHTO. Altogether, a total of about 1,500 (Report 286) more test results were incorporated into the fatigue database in addition to the aforementioned 800 data points. A wide range of standard components, structures, and welds were tested, including:

• Plain rolled beams.

• Longitudinal welds.

• Welded beams

• Flat plates.

• Flange splices.

• A514/A517 straight transitions.

• Box girder longitudinal welds.

• Transverse stiffeners.

• Web attachments.

• Web gusset plates.

• Flange tip attachments.

• Flange surface attachments.

• Cover-plated beams.

Data from 1987 to 2006

During this period, the American Welding Society (AWS) gradually incorporated many of the findings summarized by the NCHRP Report 286 into their D1.1 Structural Welding Code for Steel. Some of the highlights include:

• The information on “Fatigue Stress Design Parameters” was expanded. Quantitatively, this information is basically identical to the recommendations of Report 286.

• By the time the 2006 revision of AWS D1.1 was released, the distinction between redundant and non-redundant members was no longer recognized. This is a departure from NCHRP Report 286, which originally recommended a 20% decrease in allowable stress range for non-redundant structural members. The Commentary (C-2.13.4) of AWS D1.1-2006 gives the rational for this change as follows:

•  “The concept of recognizing a distinction between redundant and non-redundant members and details is not based upon consideration of any difference in the fatigue performance of any given member or detail, but rather upon the consequences of failure.”

• “The reduced allowable stress range curves, designated for non-redundant structures were derived by arbitrarily limiting the fatigue stress ranges to approximately 80% of the stress range curves for redundant members and details.”

•  “… it was decided that specifying allowable stress ranges that were only 80% of the mean minus 2 standard deviation curves for fatigue detail test data… constituted a double conservatism.” (This is evident from a cursory examination of the original fatigue data curves. The vast majority of the fatigue failure data points are well above the lower bound curves recommended by the NCHRP.)