Why is a hardness number asked to carry so much weight in a failure investigation?
Hardness numbers carry so much weight in failure investigations because hardness testing is fast, cheap and very nearly nondestructive, and that convenience is why a hardness number gets asked to carry more than it can. When a failed part comes back from the laboratory carrying a hardness number, that number does a great deal of work in the file: it is converted to a tensile strength, compared against a specification minimum, and offered as proof the material was not what the drawing called for.
The hardness test itself measures resistance to indentation at one small spot. Everything else is inference, and the governing hardness testing standards are unusually candid about how far that inference reaches.
How accurate are hardness conversion tables like ASTM E140?
Hardness conversions under ASTM E140 are estimates, and the standard says so in its own text. ASTM E140-12b(2019)e1, Standard Hardness Conversion Tables for Metals, turns a Rockwell reading into Brinell, Vickers, Knoop or Leeb, and it sets its own limits: "Since the various types of hardness tests do not all measure the same combination of material properties, conversion from one hardness scale to another is only an approximate process." The ASTM E140 tables are organized by material family, and the values "should only be considered valid for the specific materials indicated."
ASTM E140-12b(2019)e1 then goes further than most people who cite it expect, recommending "that hardness conversions be applied primarily to values such as specification limits ... and that the conversion of test data be avoided whenever possible." That recommendation is the reverse of ordinary practice, where the measured hardness value is converted and the specification is left alone.
How accurately can hardness predict the tensile strength of steel?
Hardness predicts the tensile strength of steel only approximately: the population correlation is strong, but the scatter around it is wider than many of the specification margins hardness readings get produced to settle. The underlying relationship between hardness and strength is real and well characterized. Pavlina and Van Tyne, in the Journal of Materials Engineering and Performance (2008, volume 17, pages 888 to 893), compiled hardness and strength data for more than 150 non-austenitic, hypoeutectoid steels spanning ferrite, pearlite, bainite, martensite and multiphase microstructures. Their regression gave tensile strength as 99.8 plus 3.734 times the Vickers value, at a coefficient of determination of 0.9347.
That is a strong correlation. The figures that matter in a dispute sit beside it: a standard error of 112 MPa on the Pavlina and Van Tyne tensile fit, and 102 MPa on the yield fit. Pavlina and Van Tyne drew the conclusion themselves, writing that the standard error “associated with the linear regressions makes them less than perfect for estimating strength from a bulk hardness measurement.”
Is tensile strength really about 3.45 times the Brinell hardness number?
Not reliably at high hardness: the field shortcut that tensile strength in megapascals is about 3.45 times the Brinell number tends to under-predict tensile strength at higher hardness. Pavlina and Van Tyne, in the Journal of Materials Engineering and Performance (2008), checked that three-and-a-half-times-Brinell rule against their compiled data for more than 150 non-austenitic, hypoeutectoid steels, and the paper calls the under-prediction significant for the highest strength steels evaluated.
The Pavlina and Van Tyne linear fits carry a separate boundary: they are reported as valid from 129 on the diamond pyramid scale upward, and the authors expect the correlation to become nonlinear at values below about 130. Neither the 3.45-times-Brinell shortcut nor the Pavlina and Van Tyne linear fits is at its best where high-strength parts fail.
Does one hardness reading represent the whole part?
Not necessarily, because a single hardness indentation is a local measurement. ASTM E18-25, Standard Test Methods for Rockwell Hardness of Metallic Materials, states the point directly: “Rockwell hardness testing at a specific location on a part may not represent the physical characteristics of the whole part or end product.” That is the reason microindentation testing exists as a separate method.
ASTM E384-22 is written around hardness variations over small distances: intentional ones from shot blasting, flame hardening and carburizing, and unintentional ones from decarburization or localized softening in service. ASTM E384-22 carries its own limits: forces at or below 25 gf give results that "should be considered to be qualitative in nature."
How does the location of a hardness indentation affect what the reading means?
Where the indenter lands decides what a hardness number means, because a reading is specific to the zone it was taken in. A hardness reading taken in metal the failure itself plastically deformed is not the as-supplied hardness; work hardening raises it. A hardness reading in a carburized case, a decarburized skin or a weld heat-affected zone measures that zone and nothing else.
Geometry matters too: ASTM E18-25, Standard Test Methods for Rockwell Hardness of Metallic Materials, carries annexes on minimum test-piece thickness and on correcting values read on convex cylindrical surfaces, both easy to overlook on a curved fragment. None of that makes a hardness number wrong; it makes the number specific to a location, which is why a defensible report says where each indentation was placed.
When is hardness a requirement in its own right rather than a stand-in for strength?
Hardness is a requirement in its own right under ISO 15156-2:2020, where it serves as an acceptance criterion for sulfide stress cracking resistance, and that use of hardness gets confused with using hardness as a stand-in for strength. ISO 15156-2:2020, covering cracking-resistant carbon and low-alloy steels for H2S-containing environments in oil and gas production, treats the hardness of parent material, welds and heat-affected zones as a control on sulfide stress cracking resistance. There hardness is not standing in for strength; it is the requirement.
A hardness limit written for cracking resistance is not evidence about tensile strength, and the reverse holds too.
What can a hardness survey reliably establish?
A hardness survey establishes relative differences, and it establishes them well. A hardness survey can show whether a failed part is harder or softer than an unused exemplar from the same lot, whether a case is present and how deep it runs, and whether a heat-affected zone hardened or softened relative to base metal, which a hardness traverse across a weld shows clearly.
A hardness survey is also a cheap screen: a reading well outside the expected band is the reason to run the tension test, the chemical analysis and the metallography that can settle the question. ASTM A370-26 covers tension, bend, hardness and impact testing of steel products in one document because the tests are meant to be read together.
What can’t a hardness number establish?
A hardness number does not fix the tensile or yield strength of a part to the resolution a specification argument usually demands. A hardness number does not identify an alloy; alloy identification is a question for composition analysis. A hardness number does not establish what the material was at delivery, since deformation, thermal history and service all move hardness afterward. And a hardness number never establishes that hardness caused anything.
The honest form of a hardness finding is narrow and still useful: at these locations, by this method, the material read this hard, consistent or inconsistent with the specified condition. Everything past that sentence has to be earned with another test.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.