322. The causes of failure in this stage most frequently derive their origin from errors in the preceding one; and it is sufficient merely to indicate a few of their sources. They frequently arise from having neglected to take into consideration that metals are not perfectly rigid but elastic. A steel cylinder of small diameter must not be regarded as an inflexible rod; but in order to ensure its perfect action as an axis, it must be supported at proper intervals. Again, the strength and stiffness of the framing which supports the mechanism must be carefully attended to. It should always be recollected, that the addition of superfluous matter to the immovable parts of a machine produces no additional momentum, and therefore is not accompanied with the same evil that arises when the moving parts are increased in weight. The stiffness of the framing in a machine produces an important advantage. If the bearings of the axis (those places at which they are supported) are once placed in a straight line, they will remain so, if the framing be immovable; whereas if the framework changes its form, though ever so slightly, considerable friction is immediately produced. This effect is so well understood in the districts where spinning factories are numerous, that, in estimating the expense of working a new factory, it is allowed that five per cent on the power of the steam-engine will be saved if the building is fireproof: for the greater strength and rigidity of a fireproof building prevents the movement of the long shafts or axes which drive the machinery, from being impeded by the friction that would arise from the slightest deviation in any of the bearings. 323. In conducting experiments upon machinery, it is quite a mistake to suppose that any imperfect mechanical work is good enough for such a purpose. If the experiment is worth making, it ought to be tried with all the advantages of which the state of mechanical art admits; for an imperfect trial may cause an idea to be given up, which better workmanship might have proved to be practicable. On the other hand, when once the efficiency of a contrivance has been established, with good workmanship it will be easy afterwards to ascertain the degree of perfection which will suffice for its due action.