May 28, 1908] 



NA TURE 



81 



constant, ranging; from o'0032 for his largest p' "-t to 

 o'oo285 for his sm.illest plate. The plotted results of 

 M. Eiffel's observations on square plates and those 



Fit;. 2. — Curve showing the dimensional variation in the air-resistance of square plates. 



made at the National Physical Laboratory are shown 

 in Fig. 2. There appears, therefore, to be a purely 

 dimensional factor in the resistance of plates, which 

 for the case of square plates has the effect of increas- 

 ing this resistance up to an area of approximately 10 

 square feet, when it becomes practically constant. 



In the small-scale experiments made in a uniform 

 current of air at the National Physical Laboratory, it 

 was found that although the resistance per unit area 

 of combinations of plates, such as lattice-work, differed 

 considerably from that of square or circular plates, the 

 resistances of similar combinations of plates were 

 .ipproximately the same. To test the possibility of 

 predicting the resistance of a complex structure in the 

 wind from observations on a small-scale model in a 

 current, a model lattice girder was constructed of 

 wood, with a span of 30 feet, and a depth of 3 

 feet 6 inches. This was placed on the tower (Fig. 3), 

 and a set of observations made on it. A small-scale 

 model of this was made in brass, the linear dimen- 

 sions being reduced in the ratio of i to 42. The re- 

 sistance of this was determined in the current. On 

 comparing these resistances they were found to have 

 precisely the same ratio as that of the resistances of 

 the large square plates in the wind and the small 

 square plates in the current, that is, the resistance 

 of the large girder was 18 per cent, greater than that 

 of the small one. The conclusion was that the re- 

 sistance of any structure, however complicated, can 

 b-3 predicted with considerable accuracy from observ- 

 ations on a small model of it, as in the similar 

 problem of the resistance of ships. 



The itnportant case of the resultant wind pressure 

 on roofs is more difficult to treat experimentallv. 

 owing to the oblique impingement of the wind, which 

 renders the position of the centre of pressure uncertain. 



L'ntil recent years it has been customary to treat 

 the forces on a roof due to wind pressure as pressures 

 affecting the windward side only, but from experiments 

 on small models in a current of air, Mr. Irminger, of 

 Copenhagen, has shown that there is a considerable 

 suction effect on the leeward side of the roof, due to 

 the eddies from the ridge. It does not necessarily 

 follow from this that the resultant effect on the leeward 



side of a roof must be a suction, as this will depend 

 on the pressure inside the building also. 



In the National Physical Laboratory experiments a 

 roof model was erected on the 

 tower, having sides each of 56 

 square feet in area. The re- 

 sults of the observations of re- 

 sultant pressure on the leeward 

 side showed widely different 

 values, according as the con- 

 ditions were those of a roof 

 supported on columns through 

 which the wind could pass 

 freely or on walls. In the 

 former case it was found that 

 the reduction of pressure inside 

 the roof due to the eddy from 

 the eaves of the windward side 

 was approximately of the same 

 magnitude as the reduction of 

 pressure outside due to the eddy 

 from the ridge, so that the re- 

 sultant pressure on the leeward 

 side was practically zero. When 

 the conditions were those of a 

 roof supported on walls, the 

 maximum wind forces were 

 found to exist (a) when the doors 

 and windows on the windward 

 side of the building were open 

 and those on the leeward side closed, and (6) vue 

 versa. In case (a) the maximum wind force was on 

 the leeward side of the roof outwards, and in case (fe). 



LENGTH OF SIDE IN FEET 



Fig. 3. — Wind Observation 'l\'\\i:i- with jiiudel ; 

 ill position. 



30 feet by 3"5 feet 



it was on the windw.ird side inwards. It follows, 

 therefore, that in such a b'jilding the roof shauld be 

 designed so as to be equally strong in each direction. 



NO. 2013, VOL. 78] 



