The following questions and comments have been asked by readers of our recent series of blogs on Hardness Testing (parts 1-8). Special thanks go to Debbie Aliya (Aliya Analytical,, George Vander Voort (Vander Voort consulting, and Alan Stone (Aston Metallurgical Services Company, for their contributions to this subject.

Comments & Questions 

Excessive plasticity: If the indenter pushes into the sample with more than 60% of the diameter penetrating into the surface, excessive material will dimple up around the indentation. This makes the reading less accurate and more difficult to read. Use a larger indenter or a smaller test piece.

Question 3: This is an interesting point about the 60%. But how does changing the test-piece size work? Maybe go to a lighter load. Some testing people use 500 kg instead of 3,000 for aluminum, for example. 

Response: Paragraph 7.1 of ASTM E10 states that “The diameter of the indentation shall be between 24 and 60% of the ball diameter.” Note 2, after 7.1, gives reasons. The upper limit is given due to reduced sensitivity in measuring the indent as it approaches the ball diameter. This would agree with your statement.

Conversions: Due to the nature of Brinell testing, conversion to the other hardness scales may be inaccurate.

Comment: In my experience, it is more of a problem going from other scales to Brinell. The main issue is inhomogeneity. 

Response: This is an extremely good point. We often “convert” between scales without taking into consideration factors such as inhomogeneity and the effect of indenter geometry on the results. 

All conversions between different hardness scales are approximations (i.e. they are not precise). There is a certain degree of imprecision in both measurements when made on the same specimen. The Brinell test, due to the large size of the indent compared to every other hardness test, is ideal for averaging out inhomogeneity due to local chemistry or microstructural differences, which is typical of as-cast, as-hot-rolled and as-annealed specimens. But if we do two different hardness tests on the same specimen, and do a Student-t test to evaluate the difference between the mean values, based on the mean, standard deviation and number of tests, I doubt that the differences between the means would be statistically significant at the 95% confidence level.

Actually, when I think about this more, I doubt that we could do a Student-t test here as the mean numbers from different tests could be, and probably would be, markedly different! But we could compare the test results to the E 140 correlations. This could be evaluated experimentally (we do not believe there is anything published on this topic).