speed, steam-pressure and other conditions were held constant, except that the method of distribution of steam was varied, and the results of such a series of tests were then compared with those otherwise obtained. In still other instances the steam-pressure was made the variable element, or the ratio of expansion and point of cut-off, the indicated and dynamometric power being in each case compared as before to obtain a measure of the engine friction. By this systematic method it was anticipated that in time a correct theory and exact formulas might be produced. This expectation has not been wholly disappointed; but the results of the investigation, while eminently satisfactory, have proved to be quite opposed to the original assumptions of the older writers, and in most perfect accord with those of the practitioners.

The first of these series of experiments to be made in so satisfactory a manner as to justify publication, were those conducted under the supervision of the writer, in the winter and spring of 1883-4, by Messrs. Aldrich and Mitchell, and published in a paper read before the American Society of Mechanical Engineers in the autumn of 1884.* The engine employed was a Straight Line Engine, constructed under the eye of its inventor, Professor John E. Sweet, past president of the society, and representing well that type of engine. These experiments showed unmistakably the error of the older formulas, and revealed the unexpected fact that, in that class of engines at least, the internal friction does not vary with the addition of load, but remains constant, so far as could be detected, at all loads. The method of lubrication and its efficiency, the variations of steam pressure and of speed, slight as they were, were accidental causes of change of engine friction, having very much greater effect on the total than a variation of the power of the engine from that marking its resistance to motion, unloaded, up to the full rated power of the machine, and even far beyond the latter amount. The engine had been carefully designed with the special intent to make engine friction as low as possible, and the loss by friction at its rated power was but about six per cent. It came down to about five per cent. at the maximum power demanded of it, varying almost precisely in inverse proportion to the indicated power. The "friction card" was a measure of the friction of the engine at all loads.

This research was again undertaken at the request of the writer, * Trans. A. S. M. E., Vol. VIII., page 86.

in the winter and spring of 1885, by Messrs. Day and Riley, of Sibley College, Cornell University, employing a similar engine, built under the supervision of the inventor in the workshops at that school. The outcome of these investigations, which have also been recently fully reported and widely published in this country and in Europe, was thoroughly corroboratory of the previous conclusions. No measurable variation of the total internal friction. of the engine could be traced to the variation of engine power and load. Studying the effect of variation of steam pressure, it was found that some slight alteration was produced, the friction increasing very slowly as pressures were increased, but not in any important degree. These data have been since revised by Messrs. Carpenter and Preston, and it has been found that the change of friction with variation of steam pressure may be taken as insensible after passing the ordinary minimum working pressure of engines, the variation being observable only from about 50 or 60 pounds per square inch downward. It having been also suggested that the method of steam distribution might produce some change in the law exhibited by the types of engine having automatic adjustment of expansion by the action of the governor, Messrs. Gillis and Buchanan, in 1887, undertook, under the direction of the writer, to settle this question by experiments upon the engines of similar type, as employed in the mechanical laboratories of the Sibley College. These experiments fully confirmed those which had previously been made, and showed sensibly constant friction at all powers and loads, whether the engine was regulated by the automatic system, or by a governor operating the throttle valve in the steam pipe or at the steam chest.

We are now brought to the study of the latest, and as yet unpublished, experiments made to determine, with some degree of exactitude, the method of distribution of internal friction, aud, further, to ascertain whether all engines are subject to the same law as has been found to control the high speed engines previously employed in these researches. These last investigations were made with this object in view by Prof. R. C. Carpenter, of Lansing, Mich., and Mr. G. B. Preston, of Sibley College, as observers, experimenting first with the engines of the college laboratories, and later with other machines of various types in and near Lansing. Earlier experiments had shown the engine friction to be independent of the load, but to be a function of the characteristics of the engine itself, of the speed of piston and rotation, of the steam

pressure, and of the method of steam-distribution, the two lastnamed conditions having slight effect, the others being most important. The weight and design, and the character of the workmanship of the engine, primarily determine the amount of its internal friction; the resistance is also a direct function of its speed, and it is slightly and observably affected, within limits, by the steam-pressure variations, and by the character of valve-gear and of steam distribution and of regulation of engine. The speed and weight of the running parts of the engine may, so far as can now be ascertained, be taken as the elements controlling friction of the machine. The details of all this earlier work have been given at sufficient length in the earlier volumes of the Transactions of the American Society of Mechanical Engineers.

It now becomes an interesting and a vitally important problem to determine just how this friction of engine is distributed among its various moving parts, its journals and guides, stuffing-boxes and piston-rings. This has hitherto been regarded as a problem incapable of solution, since it was presumed that the total and the elements of the internal friction of engine would be so seriously variable with the alteration of load, that it would be impossible to measure the friction of the machine part by part, and to sum up the whole correctly. It having been found, however, that the internal friction of the engine is invariable in any measurable or important degree with variation of power, and that the socalled "friction-card" is a measure of the friction of the engine at all powers, the speed being constant, it is at once evident that we may now proceed to analyze the several parts of this total by analyzing the engine into its various friction-producing elements, and measuring up the several elements of the total friction, each by itself, and summing all for the total. The discovery of the sensible constancy of the total friction thus affords a new means and method of investigation. This accomplished, also, it becomes possible, knowing as we now do, the quantities of friction at each point of " pairing" of elements, as Reuleaux would say, and it becomes easy, to determine just where the most serious wastes of energy and power are met, and thence, just in what direction we are to study the design and construction of the machine with a view to the reduction of these wastes most promptly and effectively. The improvement of the efficiency of the steam-engine is to be now effected very largely by, its improvement as a piece of mechanism, and nearing, as we now are, the limit of the perfecti

bility of the engine thermodynamically, the engineer is compelled to look in this direction for further opportunity of advancement. The thermodynamic efficiency of the engine has attained, in the best of modern engines, very nearly its maximum under familiar working conditions; the efficiency of the engine as a machine. still offers some little chance of gaining upon the existing conditions of best work. The thermodynamic wastes are now by the best designers and constructors reduced to about ten per cent. in large engines, while the friction-wastes of the machine are usually considerably more in that class of engines, though less in the simpler and lighter engines of the recent high-speed type.

The plan adopted in the series of experiments to be described, in which Messrs. Carpenter and Preston proposed to endeavor to effect an analysis of the total internal friction of the steamengine, and to ascertain in what proportion it is distributed to piston and crosshead, crank-pin and shaft, valve-gear and guides, was to first determine the friction of the machine in the manner already practiced by them and by their predecessors in this work, then to dismantle the engine, part by part, driving the connected parts by a pulley and belt from the main line of shafting overhead, through a transmitting dynamometer carefully standardized, and thus to secure measurements of the resistance of part after part until all the rubbing parts having been thus examined, the sum of their resistances at the normal speed of the engine should give the total internal friction of the engine and the percentages of the whole due to the resistances of each point of connection or rubbing. In each experiment the endeavor was made to secure precisely the conditions of operation, so far as was practicable, which were usual in its regular working. For instance: the engine was always heated up by its own steam when the resistances of the piston and the valves were to be measured; the speed of engine was kept the same when testing the friction of journals as when it was doing its full work; the valve, balanced and unbalanced, was tested under the usual boiler pressures, as well as unloaded, and exactly as possible, and thus every precaution that could be devised was adopted to secure precisely the results that should be observed, were such observation possible, when the machine was at work. The engine was first driven by the shafting, and through the dynamometer, with everything connected and the cylinder heated up to its usual temperature by a run, immediately preceding, under steam, the cylinder heads and

steam chest cover only being removed to prevent any pump-like action of the engine while so driven. Next, the piston was disconnected, and the power demanded to give the engine its regular speed was observed with all other parts connected and moving; thus obtaining a measure of the friction of the piston alone, by differences. Then the next point of connection would be broken, and another observation would give the friction of the next successive piece, and so on until the whole engine had been gone over, when the machine was assembled again, part by part, and thus a check obtained on the previous determinations.

The first step of importance was to secure a good standardization of the transmitting dynamometer to be employed in the work. This method required the use of a transmitting dynamometer of

B

Fig. 18.

great accuracy. Sibley College possesses a number of such dynamometers, the accuracy of each of which was tested by comparison with a Prony brake, and also by lifting a known weight through a given space. The best result was given in each case by a dynamometer of the Morin type, built in the Sibley College shops. The principle governing its action is very simple, and is shown clearly by Fig. 18. A pulley of which the rim B is shown, is fitted loose on the shaft S. Four flat springs are securely bolted to the shaft S, and to the rim B. Now, if force be applied by a belt around B, to turn the pulley, and if resistance to its turning be produced by a fixed pulley on the shaft S, from which some machine is driven by the belting, the springs e will be deflected into new positions, c', an amount proportional to the force, and the fixed pulley will then revolve, thus driving the machine. To show the amount of power transmitted, and any variation that may occur in that power, a pencil is attached to the rim of the pulley, or to a post having an equivalent motion, and a recording apparatus, consisting of a series of gear wheels actuated by a spiral thread on a sleeve on the axis, causes a band of paper to move radially under the pencil. The recording apparatus can be stopped or started at will, without interfering with the motion of the machinery, by causing the loose sleeve to engage with a lug on the shaft. The results obtained with this dynamometer agreed closely with those results obtained by the Prony brake, and by moving a known