Dimensional and Metallurgical Verification of Replacement Drive Units

2. What verification is for

Three distinct questions are commonly conflated under the heading of quality, and separating them clarifies what inspection can and cannot deliver.

Will it fit? A dimensional question, answerable by measuring interface geometry, and the only one of the three that a fitter can partly answer on the machine.

Will it function? A question about kinematics, displacement, ratio and porting, answerable by measurement and partly by test.

Will it last? A question about material condition, surface integrity and load capacity, answerable only by destructive metallurgy on a sample and by process control on the population — never by inspecting the individual unit being shipped.

The third is where the difference between two outwardly identical drives resides. A gear of the correct geometry in the wrong steel, or in the right steel with an inadequate case, or with the correct case and an untempered martensite layer left by an aggressive grind, will fit and function indistinguishably and fail at a fraction of the expected life. This is the technical reason why supply chain traceability is a quality attribute rather than a bureaucratic one: for the property that matters most, the evidence cannot be regenerated downstream.

3. Reference geometry and datums

3.1 The datum problem

A measurement is meaningful only relative to a stated reference. Geometric tolerances — runout, perpendicularity, position, concentricity — are those of ISO 1101¹ and are defined with respect to datums established under ISO 5459², and a value quoted without its datum is not interpretable. Two inspectors measuring the same housing bore runout against different references will legitimately obtain different numbers, and neither is wrong.

For a final drive the datum structure normally begins at the mounting interface: the flange face establishes a primary datum plane, the spigot or pilot diameter a secondary datum axis, and one bolt hole or a keyway a tertiary datum to fix rotation. Every interface feature that matters to the machine — sprocket mounting face, bolt circle position, shaft projection, seal running surface — is then controlled relative to that frame.

3.2 Why the frame must match the original

A replacement unit measured against a self-consistent but different datum frame can be entirely within its own tolerances and still not fit, because the tolerances of the machine's structure were allocated against the original frame. The commonest manifestation is a sprocket face that is square to the wrong reference, giving a drive that bolts up and runs with the sprocket a fraction of a degree out of the track plane — a small angular error that translates into an appreciable lateral offset at the track and shows up as accelerated track and sprocket wear rather than as a drive fault.

3.3 Splines

The spline connecting motor to sun gear, or carrier to sprocket hub, is a fit rather than a dimension, and it is specified under ISO 4156-2⁹ as a class with defined tolerances on effective and actual tooth thickness, not merely on major and minor diameter. Verification is by composite gauge — a GO gauge for the effective fit and measurement of actual space width for the minimum-material condition — rather than by micrometer. A spline that passes on diameter and fails on effective thickness will assemble and will carry load on fewer teeth than intended.

Where the spline is one of the floating connections discussed in DYCO-TP-107, this matters twice: the fit must transmit torque and must simultaneously permit the radial float on which planet load sharing depends.

4. Dimensional inspection

4.1 Methods and their limits

Coordinate measuring machines are the general instrument for housing and carrier geometry, and their accuracy is itself specified and verified under ISO 10360⁵. Linear size tolerances follow ISO 286-1³. Their weakness is sampling: a CMM measures a bore by taking a finite number of points, and a bore with a three-lobed form error sampled at three points can return a perfect circle. Point density and distribution are part of the method, not an operator preference.

Hand instruments remain appropriate for many features and carry their own constraints — a bore micrometer measures a diameter along one line and will not find lobing at all, and a two-point measurement of a three-lobed bore returns a constant reading in every orientation.

4.2 Measurement uncertainty and the decision rule

Every measured value carries an uncertainty, and the comparison of a value with a limit is therefore a decision under uncertainty rather than a simple comparison. The established convention, set out in ISO 14253-1⁴, is that the uncertainty is taken from the party making the claim: conformance is proven only when the measured value lies inside the tolerance by at least the measurement uncertainty, and non-conformance only when it lies outside by the same margin. Between those bands lies an ambiguous zone in which neither claim is established.

The practical consequence for inspection of drivetrain components is that a measurement capability substantially better than the tolerance is required for the result to be useful. Where the uncertainty is a large fraction of the tolerance, most measured values fall in the ambiguous zone and the inspection decides very little. This is the technical content of the rule of thumb that measurement uncertainty should be a small fraction — commonly quoted as a tenth, sometimes a quarter — of the tolerance being verified.

5. Gear inspection

5.1 Elemental and functional

Two families of gear inspection exist and they answer different questions. Elemental inspection measures individual geometric deviations — single and cumulative pitch deviation, profile form and slope, helix form and slope, runout — each against its own tolerance, on a dedicated gear measuring machine. Functional inspection rolls the gear in mesh with a master and records the composite variation, which lumps all deviations into one trace.

Elemental inspection diagnoses: a profile slope error points at the cutter or the grinding wheel dress, a helix slope error at machine alignment, runout at the workholding. Functional inspection screens: it is fast, it correlates reasonably with noise, and it can pass a gear whose individual deviations happen to compensate one another. Both are legitimate, but only the elemental result supports a claim about the grade of individual deviations; a composite result, however it is toleranced, is a statement about the composite variation alone.

5.2 What an accuracy grade means

The flank tolerance classification of ISO 1328-1⁷ assigns numbered grades in which a lower number is more accurate, with each grade step corresponding to a fixed ratio in allowable deviation. A stated grade is a statement about the deviations that were measured and tolerated — and it is specific to the parameters named. A gear qualified to a grade on pitch and profile is not thereby qualified on helix, and a gear number quoted without the parameters and the standard edition is incomplete.

The grade also says nothing about the surface or the metallurgy. A gear can be accurate and soft, accurate and shallow-cased, accurate and burnt. Accuracy governs the dynamic and load-distribution factors and the noise, and through those factors it affects the stress the teeth carry; it does not govern the permissible stress, which is set chiefly by the material condition of section 7.

6. Surface texture

Surface texture is specified under ISO 21920-2⁶ by parameters derived from a filtered profile, and the filter is part of the specification: the same surface reports different roughness values under different cut-off wavelengths, so a texture value without its sampling length and filter is not comparable with another.

Two applications dominate in a drive unit. Gear flanks: roughness enters the surface durability calculation through the lubricant film parameter, since it is the ratio of film thickness to combined roughness that determines whether asperities interact, and a rougher flank on the same lubricant operates at a lower specific film thickness and is more prone to the micropitting discussed in DYCO-TP-102.

Seal running surfaces: here the requirement is not simply "smooth". A radial lip seal counterface has an optimum roughness band — too rough abrades the lip, too smooth impairs the lubricating film the lip needs — and, critically, a specified lay. A surface finished by plunge grinding has a circumferential lay with no helical component; a surface finished by traverse grinding or by any process leaving a lead angle acts as a screw thread and will pump lubricant along the shaft in one direction of rotation. Lead is a defined defect on seal counterfaces, it is not detectable by a roughness parameter, and it is verified by a dedicated lead measurement or a thread-check test. It is a recurring cause of seal leakage on components that pass every dimensional check.

7. Metallurgical verification

7.1 Hardness

Surface hardness is measured by Rockwell under ISO 6508-1¹¹ on components robust enough to take the indentation, and by Vickers microindentation under ISO 6507-1¹⁰ where the region of interest is small — a case, a tooth flank, a weld heat-affected zone. Conversion between scales is possible but approximate, and the tables of ISO 18265¹² carry the explicit caveat that they are material-dependent; a converted value is weaker evidence than a value measured on the scale specified.

Surface hardness alone is a poor discriminator between a good gear and a bad one. Two gears can share a surface hardness of 60 HRC and differ by a factor of two in case depth, which is the parameter that governs whether subsurface shear stress is contained within the hardened layer.

7.2 Case depth

Case depth is determined on a sectioned, mounted and polished sample by a microindentation hardness traverse from the surface inward, at a defined load and spacing. Two definitions are in use and they are not interchangeable. Effective case depth is the perpendicular distance from the surface to the point at which hardness falls to a defined limiting value — conventionally 550 HV for carburized steels, though the limit is properly stated with the result. Total case depth is the distance to the point at which the case is metallurgically indistinguishable from the core.

Total case depth is the larger number, and quoting it where effective depth is specified overstates the component. The depths of a carburized case and of an induction- or flame-hardened layer are both determined under ISO 18203¹³, which replaced the earlier separate standards for each. The limiting hardness differs, reflecting the different hardness profiles the two processes produce: the effective depth of a carburized case is read at the conventional limit given above, and that of an induction- or flame-hardened layer at a limit tied to its minimum specified surface hardness. Slew ring raceways, discussed in DYCO-TP-105, are almost always induction hardened and fall under the latter.

7.3 Microstructural features

Several conditions that reduce load capacity are invisible to hardness testing. Retained austenite in a carburized case is soft, unstable and can transform in service with a volume change. Intergranular oxidation — oxygen penetration along prior austenite grain boundaries in gas carburizing — produces a shallow non-martensitic surface layer at the tooth root, precisely where bending fatigue initiates. A continuous carbide network at the surface embrittles it. Inclusion content is rated micrographically under ISO 4967¹⁴ and apparent grain size under ISO 643¹⁵. Each requires a sectioned and etched sample; none is inferable from a hardness number or from the appearance of the finished part.

7.4 Grinding burn

Of all the defects in this section, grinding burn deserves particular emphasis because it is common, invisible and directly damaging. Excessive heat at the grinding contact re-austenitizes and re-quenches a thin surface layer, producing untempered martensite, or over-tempers the layer below it. Either changes the residual stress state at the surface from the compressive condition that heat treatment was performed to produce into tension — the condition most favorable to crack initiation. A burnt flank is dimensionally perfect, correctly hardened at the depths a traverse would sample, and materially compromised at the surface where contact fatigue starts.

The established detection method is surface temper etch inspection to ISO 14104¹⁶, equivalently ANSI/AGMA 2007¹⁷: a controlled sequence of acid etch, rinse and neutralization that develops visible contrast between correctly tempered material, over-tempered material and re-hardened material. It is a qualitative, whole-surface, non-destructive method covering the entire flank rather than a sampled point, which is exactly what is needed for a defect whose location is unpredictable. It is also a hydrogen-charging process, so a post-etch bake is part of the procedure on high-strength steel.

7.5 Non-destructive testing

Magnetic particle inspection to ISO 9934-1¹⁸ detects surface and near-surface discontinuities in ferromagnetic components and is the standard method for cracks in gears, shafts and carriers, including the quench cracks that can follow heat treatment. Penetrant inspection to ISO 3452-1¹⁹ serves the same purpose on non-ferromagnetic parts. Ultrasonic inspection reaches subsurface discontinuities in section thicknesses that surface methods cannot address, and is applied to forgings and to large rings.

None of these establishes material properties. They establish the absence of detectable discontinuities of a given size, which is a different and narrower claim.

8. Sampling, and what a certificate certifies

8.1 Destructive tests are sample tests

Case depth, microstructure and inclusion rating require a sectioned specimen, so the component tested is by definition not the component shipped. Verification of these properties is therefore a claim about a population supported by a sample and by control of the process that produced it — heat treatment furnace records, load thermocouples, batch identity — rather than a claim about an individual unit. Sampling plans — ISO 2859-1²⁰ for inspection by attributes — define the relationship between sample size, acceptance quality limit (AQL) and the confidence the result supports.

This is the structural reason why traceability to a heat treatment batch is not paperwork for its own sake. Without batch identity, a destructive test result attaches to nothing, and the inference from sample to shipped part cannot be made at all.

8.2 Inspection documents

EN 10204²¹ provides the vocabulary generally used in this trade, and the distinctions it draws are worth stating plainly because they are frequently blurred in commercial use:

• Type 2.1 — a declaration of compliance with the order, with no test results. It states that the supplier believes the goods conform.
• Type 2.2 — a test report with results of non-specific inspection: tests on material representative of the production, not necessarily on the delivered items.
• Type 3.1 — an inspection certificate with results of specific inspection on the delivered items or on the test unit of which they form part, such as a heat treatment batch, validated by the manufacturer's own inspection representative, who must be independent of the manufacturing department.
• Type 3.2 — as 3.1, but countersigned by the purchaser's representative or an independent inspector.

The distinction that matters most is between 2.2 and 3.1: whether the numbers on the document were obtained from the items in the crate or the test unit they belong to, or from material merely representative of them. Both are legitimate documents. For a destructively determined property such as case depth the tested piece is never the shipped piece, so a 3.1 result is only as good as the batch identity tying the sectioned sample to the delivery, and without that link a 2.2-type result is often the only kind that can exist — but a document should be read for which claim it is making. Where testing is performed by an external laboratory, the laboratory's competence is itself the subject of ISO/IEC 17025²², and an accredited result carries a defined scope naming the specific methods it covers.

9. Summary

Verification of a replacement drive answers three separable questions — will it fit, will it function, will it last — of which only the first two can be established by inspecting the unit being shipped. Fit is a matter of interface geometry referenced to the same datum frame the original drawing used, and a self-consistent measurement against a different frame can be simultaneously correct and useless. Gear accuracy grades govern dynamic load, load distribution and noise, and say nothing about the material condition that chiefly sets the permissible stress. Case depth is a sectioned measurement with two non-interchangeable definitions; retained austenite, intergranular oxidation and carbide networks are invisible to hardness testing; grinding burn is invisible to dimensional inspection and to a routine hardness traverse, is found in practice by a temper etch, and converts the surface residual stress from the compressive state heat treatment exists to produce into the tensile state that initiates cracks. Because the material properties that determine service life require destructive testing, they are claims about a population supported by batch traceability and process control, and the grade of inspection document distinguishes a test performed on the delivered items or their test unit from one performed on material merely representative of them. A part that fits and turns has satisfied the two questions that can be checked on arrival, and neither of them is the one that determines how long it lasts.