The chief trouble with iron castings is their liability to have internal strains put upon them in cooling, in consequence of their shrinking. The amount of this shrinkage varies with the quality of the metal, and with the size of the casting and its comparative thickness. Thus locomotive cylinders shrink only about 1/16- in. per ft. (1-192 = .0052), while heavy pipe castings and girders shrink 1/10 in. per ft. (1-120 = .0083), or even 1/8 in. per ft. (1-96 = .0104). While small wheels shrink only 1/25 in. per ft. (1-300 = .0033), large and heavy ones contract 1/10 in. per ft. (1-120 = .0083). The "shrink-rule" is employed by pattern-makers to relieve them of the labour of calculating these excesses, the scales being graduated to inches, etc, which are .0052, .0083, etc, too long. Now, if thick metal proportionately shrinks more than thin, we must expect any casting not absolutely symmetrical in every direction to change its form or proportion. A cubic or spheric mould yields a cube or a sphere as a casting; but a mould, say of the proportions of 100 X 5 X 1, shrinking differently according to dimensions, gives a casting not only less in size but in somewhat different proportion. In many cases we still find them strained and twisted.

Those parts which cool first get their final proportions, and the later cooling portions strain the earlier, the resistance of which to deformation puts strains on those cooling. This initial strain may of itself break the casting, and, if not, will weaken it. Castings of excessive or varying thickness, and of complicated form, are most in danger from internal strain. This strain is gradually lessened in time by the molecules "giving." In a casting such as a (Fig. 11), say a thick press cylinder, the outer layers solidify and shrink first, and as the inner layers contract after the outer ones have "set," there is compression of the outer layers and tension of the inner. Such a cylinder will, if subjected to internal pressure, be weak, because there is already in the inner layers a force tending to expand them. The cylinder would be stronger if these layers were braced to resist extension, or, in other words, were already in compression. If we cool the interior first, by artificial means, while delaying the cooling of the exterior layers, we have these layers braced to receive gradual or sudden pressure, and this is especially desirable in cannon. In a panel like b, with a thin but rigid flange, the diagonals shrink more slowly than the rim, and a crack is likely to appear.

A casting like that in c would solidify on the thin side first, and when the thick side shrank, it would curve the bar and compress the thick part, and put the thin in tension. Wheel and pulley castings d are especially troublesome. The latter have a thin rigid rim, which cools before the arms, and when the latter cool they are very apt to break by tension. If the arms set first, they are apt to break the rim, as they make a rigid abutment which resists the rim-contraction, bending the rim and breaking it from within outwards. In the cooling of casting?, the particles range themselves in crystals perpendicular to the cooling surface; hence we may expect to find weak points at sharp corners, as in e. The remedy for this is to round off all angles.