Page 398 - Forensic Structural Engineering Handbook

STEEL STRUCTURES 11.29

Almost no code or specification requires testing at the component or whole-structure
level, with the notable exception of proof testing of pressure vessels and storage tanks. The
pressure vessel procedure typically over-pressurizes the vessel by 10 percent; and, in addition
to providing some assurance of integrity, if performed at a temperature high enough to
ensure that the vessel plates and weldments are in ductile range of toughness, it will blunt
any cracks or other discontinuities that may be present. Proof tests of this type performed
below the ductile-to-brittle transition temperature are potentially quite dangerous, especially
in vessels made of steels of marginal quality. No such proof tests are normally employed in
bridge or building codes (although some load tests have been done on bridges). For bridges,
erection procedures sometimes inadvertently perform the same function.
Many bridge engineers go by the rule of thumb that if the bridge can be erected, it will
perform satisfactorily in service. This is just recognizing for many structures, that the erec-
tion loads are more severe than the service design loads when the structure is not complete.
This also reflects the comment offered in the previous section on “Methods of Analysis”
where it was noted that the designer should recognize and take into account the fabrication
and erection loads. Occasionally, both in forensic analysis and when new designs are con-
templated, large-scale component or whole-structure tests are undertaken. It is essential
(and may seem obvious) that they be carefully planned to duplicate the structure being
studied, they be instrumented to provide the critical information required, and that the con-
ditions of loading (and thus, the resulting fracture mode, if a full failure test is contem-
plated) be compatible with those known to be present in the structure. Because large-scale
and full-structure tests are expensive and often provide only a single test result (success or
failure), careful analysis of the structure prior to testing, installing good instrumentation,
and gathering as much data as possible are a better investment than post-testing explana-
tions. Despite the difficulties in performing them, in post-failure investigations, in demon-
strating the viability of new designs, and in certifying the success of retrofitting procedures
to prevent failures, full-size or large-component tests are sometimes the only way to pro-
vide information on how a complex structure will behave.

TYPES AND CAUSES OF NONPERFORMANCE
AND FAILURE

Design Deficiencies
Dangerous structural deficiencies may result from designs in which a basic assumption is
not valid. An example was the nodal point supports of the Hartford Coliseum space frame
that collapsed in January, 1978. The connection at mid-length of the top chord of the space-
frame did not provide lateral support for the assumed column length of that top chord. As
illustrated in Fig. 11.12, the horizontal chord was connected to flexible gusset plates that
were unable to prevent lateral displacement. Hence, the effective column length was nearly
twice as great as assumed in the design. 41,42 As a result, many members were highly over-
loaded and deformed, which redistributed the loads and resulted in total collapse of the
structure, as illustrated in Fig. 11.13.
Another example of failure originating from design was the brittle fractures of welded
steel moment frame (WSMF) connections during the Northridge earthquake in 1994 where
the use of low-toughness material, in this case the groove weld metal, became the principal
reason for essentially elastic fractures of the connections. 43,44 In this case the discontinuities
were the steel backing bars that were used at the column face and provided an unfused edge
lack-of-fusion condition perpendicular to the bending stresses in the beam flanges, as illus-
trated in Fig. 11.14. This initial geometric condition provides an edge crack equal to the