intermittent bracing is used, thermal effects will induce
large stresses in the runway. Unless the building is free to expand from the braced bay there is a tendency for the structure to " buckle" lengthwise because of the large induced forces, thus throwing the runway out of alignment.
Paper presented by JOHN E. MUELLER (January, 1965, Issue)
Discussion by J. J. MURRAY M R . JOHN E. MUELLER'S article in the January, 1965,
J. J. Murray is Staff Civil Engineer, Jones & Laughlin Steel Corp.
Pittsburgh, Pa.
issue of the Engineering Journal was most interesting.
The writer would agree in practically every respect with the conclusions cited. The writer cannot agree, however, with the suggestions submitted in the Discussions in the April issue of the Engineering Journal. Mr. V. R. P. Saxe suggests the attachment of the crane girder web directly to the flange of the lower shaft of the crane columns. We have a large number of buildings in which a similar practice was used. We have had serious troubles with these structures. The crane girder, when designed as a simple span, will rotate at the supports under live load. This rotation, unless permitted, will shear off the connection or produce web failure. We have had both phenomena occur. O n our buildings the connections were riveted. However, whether riveted or welded the end connection, as illustrated by M r . Saxe's Figs. 1 a n d 2, will attempt to induce negative moment in the crane girder. Unless designed as a continuous girder, high stresses will be induced at the supports. O u r practice has been to outlaw this type of connection. A further weakness in the illustrated design relates to t h e fact that there is no provision for transfer of the horizontal thrusts of the crane into the column shaft. Even on relatively lightly loaded runways, carrying 15 or 20 ton cranes, we have had trouble with this detail. Under service conditions we have experienced horizontal cracking in the web as the horizontal thrust effects have bent the top flange back and forth. This detail we also do not allow in our work. In the case of M r . T. Klayton's Discussion in the April issue, the writer would again disagree (with his comment (b)). We have had trouble with long runways braced in the manner he suggests. In one case it was necessary to eliminate all longitudinal bracing except the bracing in the center of the building " r u n " . When
Failure of Simply-Supported Flat Roofs
by Ponding of Rain Paper presented by JAMES GHINN (April, 1965, Issue) Discussion by MARTIN P. WALSH, JR., P.E. T H E ARTICLE BY Prof. Chinn, while a complete treatment of an isolated simple beam, fails to caution the reader that deflection of subframing members must also be considered in the ponding problem. T h e typical framing situation encountered is often considerably more serious due to the additional deflection of joists, subpurlins and other framing members which serve to transfer the roof loads to the main girders. The sample problems in the paper assume zero deflection in the members between adjacent girders; this requires infinitely rigid subframing members. In practice these subframing members often have deflections equal in magnitude to those of the main girders. This office has found this cumulative deflection ponding effect neglected in many designs. Also, the last equation on page 39 is inconsistent unit-wise. T o make the answer 0.344 (instead of 344), the unit weight of water must be expressed as 0.0624 kips per cubic ft instead of 62.4 pounds per cubic ft.
Martin P. Walsh, Jr., is Deputy Building Commissioner, City of