SHEET NON-FERROUS METALS 759 



hardness readings bear no relation to tensile strength due to the 

 anvil effect. 



The Rockwell hardness test is carried on in the same manner as 

 given in Appendix I of this paper. All of the Rockwell hardness 

 tests were made on the two grip ends of the tension test specimens. 

 This was done in order to have Rockwell hardness tests and tension 

 tests made on as nearly identical material as possible. Commercial 

 experience with the Rockwell machine in the routine testing of sheet 

 non-ferrous metals has indicated the importance of the following 

 precautions. 



Ball Penetrator. — Slight variations in the size and sphericity of 

 the ball penetrators are to be expected and must be guarded against. 

 The results of several hundred measurements made at various points 

 on the "B" scale show that a deviation of less than 0.00002 in. in the 

 diameter of the re-in. ball has no effect on the hardness readings. 

 Balls showing a greater variation give noticeable error in hardness 

 readings. Penetrator balls have been held within this limit. 



Anvil Surface. — It is desirable that the penetrator be perpen- 

 dicular to the testing surface of the anvil. It is assumed that the 

 penetrator is operating perpendicular to the seating surface and in line 

 with the capstan head. Any lack of perpendicularity would then be 

 due to lack of parallelism of the seating and testing surfaces of the 

 anvil. This is checked by a fixture designed to hold the anvil so that 

 the specimen supporting surface might be checked. A hardened flat, 

 2 in. square and | in. thick, having a \^-'m. hole through its center, 

 was lapped so that its two surfaces were flat and parallel to within 

 0.00005 in. By placing the fixture holding the anvil on the table of 

 an optimeter and sliding the supporting surface of the anvil under a 

 ball-pointed feeler gage, errors in flatness and parallelism may be 

 measured to 0.00005 in. Anvils whose surfaces were found to be 

 plane to within 0.0001 in. were considered satisfactory. 



Machine Errors. — Observations on four Rockwell machines at the 

 Western Electric Co., one of which handles approximately 10,000 

 tests per month and all of which were overhauled and calibrated for 

 their entire range once a week, showed that changes in calibration 

 occurring in this interval were less than one hardness number. Before 

 beginning a series of readings the correction to be applied was deter- 

 mined by taking readings on a standard block. The average change 

 on all three scales was about one hardness number. Variations from 

 calibration were rarely greater than 1.5 numbers unless the pene- 

 trator was damaged or the instrument was out of adjustment due 

 to misuse. 



