In each test given in this paper the value of the change has been secured through the actual lubricating virtues of the oil which has been used to secure the reduction in friction. The power reductions are not in any way due to changes made in the mechanical running of the shops and mills operated upon, such as cleaning of bearings and relining of shafts. Work of this character is never allowed to take place during the period between the two tests which are required for the comparative results. The utmost care is exercised so that the results will show nothing except the absolute difference between the lubricating character of the oils which were found in use and the oils which are substituted.

When testing under this system was begun one of the first problems to be solved was how to nullify, offset, or record the influence upon the total power of the plant tested upon by the stopping and starting of the various machines. Assuming that a total reduction in horse power of five per cent. was secured through a change in lubricants it was necessary to determine that this reduction was not due to any other feature, and it was also important that some ouside influence did not entirely absorb this reduction. The solving of this problem has made it necessary to establish a rather elaborate system of indicator work and to have the schedules of such tests arranged so that the best average load can be secured.

To make a proper comparative power test each plant must be taken individually and consideration given to the effects of weather and the class of work the plant is engaged upon. The possibilities must also be considered of the same conditions existing at some future time when all changes in lubrication which are to take place have arrived at a good basis. This requires a knowledge of the effect of temperature, humidity, and the change in the material which the plant manufactures, as well as a general knowledge of all the conditions surrounding the particular class of plant operated upon.

The work load factors in the various tests are determined by means of frequent reports made by the overseers of each depart

ment. These reports show the number and kind of machines that are not in operation for periods exceeding fifteen minutes. The horse power for all machines requiring correction is ascertained and a debit and credit balance is made so that both tests will be upon a comparative basis.

The engineering shop tests which are given in this paper, with one exception, have been run from electric motors and therefore are necessarily only sections of larger shops. The one exception is a plant which is operated by a Willan's highspeed engine. The comparative tests on engineering shops are made by having all the machines run at their regular speed but without any work in the lathes or on the cutting tables. In this way the work factor is entirely eliminated and comparative conditions can be duplicated.

All the plants operated upon were working under ordinary conditions, and they continued to work in the same condition as long as the test was being made.

DETAILS OF ACTUAL TESTS.

The straight test for one condition usually requires three days. About forty sets of indicator diagrams are taken each day while the engine is carrying full load, a set being all the cards taken from all indicators in operation at one time. A day's record is also made of the horse power required to operate the engine and transmission. This load is usually taken at noon, and not less than ten sets of diagrams are taken for each test. Where a subdivisional test showing the load of each department is being made, the schedule is so arranged that different departments are thrown out of operation for periods not exceeding one half hour, during which time never less than ten sets of diagrams are taken to secure the average load for the balance of the machinery. On a simple test, with four indicators, a full load and transmission test requires 200 diagrams per day. When eight indicators are used in a subdivisional test, the number of diagrams per day sometimes reaches 2,000, which would make 8,000 indicator diagams for a four day comparative test. Never has their

been a business that has allowed men to become so expert in indicator work as that which requires the conducting of these tests for showing the comparative efficiency of lubricating oils, and the tests given in the present paper, which are only a small selection from those that have been made, could not have been conducted unless there has been an equal opportunity to organize a highly trained staff of engineers native to each country and to equip these men with every instrument necessary for their work.

METHOD OF CONDUCTING A TEST.

It has been found from experience that Monday is a very unsatisfactory day for securing the power of the textile mills and of any other plant that has been shut down over Sunday. This day, therefore, is usually given to setting up instruments and laying out the test. Schedules are prepared and the different men individual to each plant are instructed in regard to their various duties. The test starts Tuesday morning and is usually of eight hour's duration, schedules being arranged to secure an average load which will be quite away from the influences of the breakfast hour in England, the rest hours on the Continent, and noon hours in America, each country having some peculiarity which interferes with a uniform schedule. Mills in Japan run day and night for the entire week which makes it impossible to secure transmission and subdivisional data.

The diagrams are all taken at regular intervals prescribed by the schedule and as fast as they are taken they are stamped and worked out by the Planimeter engineer. The mean effective pressures are then written upon log sheets and at the conclusion of the day's test, the constant of each end of each cylinder having been obtained, the horse power for each card is worked out either with a twenty inch slide rule or a multyplying machine and the result put on the log sheets. These log sheets are so arranged that by totalling the horse power vertically and horizontally the result will be self-checking. The tests are repeated for two or three days in order that a fair average of all surrounding condi