AS 2193 grading explained: how a jack or load cell earns Class A

What does AS 2193 apply to?

AS 2193:2005, Calibration and classification of force-measuring systems of testing machines. In Australian practice it is the standard under which hydraulic pre-stress jacks, jacking systems and rams used in post-tensioning are calibrated, and load cells and proving devices classified. A jack calibration runs in the opposite direction to most: the jack's own gauge supplies the nominal in MPa and the reference load cell supplies the measurement in kN, so the deliverable is a conversion the crew applies on site, printed as an equation and a table, rather than a verdict.

What are the class limits?

Table 2.1 of the standard. Comparisons are inclusive: exactly 1.0 % is Class A. And clause 2.4 requires the calculated repeatability and mean error to be rounded to 0.1 % before the class is assigned, so 1.04 % is Class A and 1.05 % is the one case that depends on the rounding convention.

Readability: can the reading support the claim?

The applied force divided by the effective resolution of the scale, where the effective resolution is the smallest division divided by however finely a reader can interpolate between marks. A ratio of 400 or more is Class AA, 200 is Class A, 100 is Class B, 40 is Class C. Clause 3.3.2 states the same rule from the other end: an operating range should not be calibrated for classification below the force given by multiplying the scale resolution by the multiplier.

This expresses something physical. A point read too near the bottom of the scale cannot be graded well no matter how accurate the instrument is, because the reading itself is too coarse to support the claim. It is the sub-grade that most often limits a jack at its lowest pressure, and that is correct behaviour rather than a defect in the jack.

Repeatability: do the runs agree?

At each point, the largest reading minus the smallest, as a percentage of the mean of the runs, rounded to 0.1 % and classified against the limits. Three runs give one figure; the divisor is the mean, not the nominal.

Error: does the mean agree with the reference?

The signed difference between the mean indicated force and the force the reference measured, as a percentage, rounded and classified against the same limits. Where a laboratory cannot demonstrate traceability better than a given class above a certain force, because its own reference chain stops there, it declares that force and the error class is capped above it. That is a limitation of the laboratory's reference, declared by the laboratory, never a default in software.

How are the three combined?

The point's class is the worst of its three sub-grades, and the record keeps which one limited it, so a certificate can explain a downgrade rather than merely assert one. The instrument's class is its worst point. There is no averaging and no discarding of the worst point: a force-measuring system is classified by its weakest point because that is where somebody will use it.

A worked example from a pre-stress jack

At 20 MPa on the gauge, three ascending runs read 184.2, 184.9 and 184.5 kN on the reference, and the reference measured 184.5 kN at the mean. The jack's force scale is read to 0.5 kN.

The point is Class A, limited by readability. The jack repeated and agreed to Class AA; the scale was too coarse at this force to support the higher claim. At 60 MPa the same scale gives over a thousand divisions and readability stops being the limit, which is why a jack's class is so often decided at its first point.

What the grade does not say

A class is a classification of the measuring system, not a statement that it conforms to a specification. So no decision rule is required under ISO/IEC 17025 §7.8.6, and the certificate should not print PASS beside a class. The decision rules guide covers the instruments that do need one. The uncertainty is still evaluated and printed at every point; the uncertainty budget guide shows how, and why a jack's budget has to be assembled in kN rather than in the MPa the gauge reads.

The defects this arithmetic is most prone to in a spreadsheet were all found in real certificates: a grade cell spanning five rows so a later point contaminated an earlier one; a repeatability branch testing a raw force reading instead of the percentage, so any reading above 2 kN jumped to Class B; and a #REF! in the error grade cell, so the printed class meant nothing. None looked wrong on the page.