ANCIENT AND MODERN WORKS. 19 



of being riveted to the top of the girder on both sides of the web. 

 (3) The girder was changed from a continuous web for the whole 

 circumference to construction in segments; riveted together at 

 the ends by means of vertical angles. (4) The tank was anchored 

 to the tower by eye-rods. (5) The butt-straps covering the radial 

 joints were placed on the outside instead of on the inside of the 

 tank bottom. The initial rupture was assumed as having taken 

 place at the juncture between the spherical- shaped plate and 

 the inner ring of plates connected with it. In opposition to this 

 theory, reference is made to a full discussion by Prof. Marsden, 

 p. 179. The designing engineer offers the following explana- 

 tion of the disaster: The failure to rivet the bottom angle securely 

 to the head of the girder as well as to the side angle. A movement 

 was caused by the pressure of the water normal to the bottom, 

 which movement brought an outward pressure upon this girder, 

 which was unsupported except by tensile strength in itself. That 

 is to say, there was an outward pressure on this girder similar 

 to the pressure of the side walls of the stand-pipe and possibly 

 to as great a degree. This pressure would, of course, as in the 

 walls of the stand-pipe, be concentrated at a weak joint, sufficient 

 finally to overcome the comparatively small resistance of one 

 of the joints where the girder was riveted together. This joint 

 failed and the girder was pushed out, or moved from under the 

 bottom at that point, and the weight of the water forced the 

 bottom away from the angle iron by pulling the counter-sunk 

 rivets through the plate. At the same time, the whole tower 

 being weakened at the top, took a twisting motion; the bottom 

 plate fell down and tore off the centre, and the entire structure 

 collapsed. It was believed by the designer that the initial rupture 

 occurred at the sides and not in the centre, as it seemed incredible 

 that the centre plate, even of poor material, could have failed, 

 as the pressure at that point was a minimum and the thickness 

 of the material was ample for the calculated stress. (Abstract 



