The certificate is not the complaint
Here is the mess in one paragraph. You send a lot of machined brackets or gear blanks to a heat treater in your cluster. The parts come back with a hardness certificate that says 58 HRC on five sample points, and a third of the lot will not sit flat on the surface plate. You call the treater. He says the hardness is in spec, the certificate is attached, and the distortion is your machining stress. You have no record of how the parts were supported in the furnace, no agreed fixture, no pre-dispatch dimension sheet, and no certificate format that tells you where those five points were taken. The argument goes nowhere because you are comparing two different documents: a result sheet against a dimensional complaint. Fix the paperwork and the load plan before the next lot, not after.
The hardness certificate answers one question: did the sample points reach the specified hardness on the specified scale? It does not answer whether the part moved, whether it was supported at temperature, or whether the quench flow reached all sides evenly. Those are process questions, and in job work they are yours to specify because you own the drawing and the purchase order.
What actually drives distortion in a job-work load
Distortion is not a single fault. A published account from a commercial heat treater lists the factors as material chemistry, prior thermal and machining history, geometry, fixturing, case uniformity, and the quenching process, and notes that the heat treater controls the last three (Stack Metallurgical Services, Thermal Processing, “Vacuum Carburizing Large Gears”). Read that list carefully. Chemistry and prior history come from your side — your mill certificate, your normalising, your machining stresses. Fixturing, case uniformity, and quench are the treater’s side. A distortion dispute that does not separate these two halves is a shouting match.
Two mechanisms matter most at SME lot sizes:
- Support at temperature. The same source states that parts must be arranged and supported to minimise movement or creep during the cycle and to allow rapid cooling. At carburising temperatures and long cycles — the article cites cycles in excess of 100 hours for some case depths — unsupported sections creep and ovality increases. Even at shorter hardening cycles, a long thin bracket lying flat on a tray with no support at its ends will sag.
- Quench symmetry. The article is explicit: if heat removal is not symmetric across the cross-section, the stress from martensite formation causes distortion, and fixtures must let the part enter the quenchant with minimum turbulence and even flow on all sides. A part that enters the oil edge-first, or sits in a basket where neighbouring parts shield one face, cools unevenly and bends toward the slow-cooling side.
Fixture design guidance from a fixture supplier’s technical presentation (Sajjan Group, “Optimization of Heat Treatment Fixtures”) lists the criteria plainly: a good ratio between fixture weight and payload, low fixture weight for handling, good oil draining, good flow of gases and quench media, and good support of parts. The same presentation notes that above 700°C, heat-resistant fixture alloys are no longer treated as flexible and load calculation is done on creep stress. In plain terms: the fixture itself sags if it is under-designed, and a sagging fixture takes the parts with it.
None of this means you need a metallurgist on staff. It means the purchase order has to carry three things it probably does not carry today.
The three attachments that belong on every heat-treatment PO
Treat the heat-treater’s load plan, fixture choice, and certificate as three separate specifications. If any one is missing, the dispute is unwinnable regardless of who is right.
1. The drawing with critical dimensions and pre-heat-treatment values
Not the customer’s drawing alone. A working sheet that names the two or three dimensions that matter after heat treatment — flatness on the mounting face, bore size, centre distance, runout — and records the measured value before dispatch. Measure on the same setup you will use on receipt. If you measure bore size with a plug gauge before dispatch and a bore gauge after receipt, you have introduced your own variable into the comparison.
This sheet does two jobs. It tells the treater which features you will check, so he can plan support around them. And it gives you a baseline, so a dimensional complaint becomes a delta rather than an opinion.
2. The agreed support and fixture method, in writing
You are not asking for the treater’s proprietary cycle recipe. You are asking for a description of how the part sits in the furnace and how quench flow reaches it. A workable format is three lines on the PO or an attached sketch:
- Orientation: part standing on its flange face, or hanging, or nested in a cast tray with support at the boss.
- Support points: where the fixture contacts the part, and whether any unsupported span exceeds a stated length.
- Quench path: how the part enters the quenchant and which faces are shielded by neighbouring parts or by the basket.
This converts “distortion problem” into a checkable claim. If the agreed method says the part stands on its flange and the lot comes back with the flange bowed, you have something to discuss. If there is no agreed method, the treater can always say the part was loaded the way it always is.
For thin sections and long spans, ask specifically whether the part can be supported at both ends or hung rather than laid flat. The Thermal Processing article describes a hanging fixture arrangement in which the load is suspended from an overhead trolley, with no grid assembly, so parts contact the oil without interference and heat removal is more symmetrical. That is one treater’s setup for large gears, not a universal rule — but the principle, that support and quench access are designed together, applies to a 200-gram bracket as much as to a gear.
3. The certificate format
A hardness certificate that says “58 HRC” and nothing else is close to useless for a dispute. Specify the format before the first lot. At minimum:
- Batch identification that matches your PO and your incoming lot number, not just the treater’s internal job number.
- Hardness scale and method stated explicitly — HRC, HV, or HB, and the instrument used. A conversion between scales is not the same as a measurement on that scale.
- Sample location described on a sketch or by named features: “three points on the flange face, one at the bore, one at the web.” A certificate with five points and no locations cannot be compared with your drawing requirement.
- Number of parts sampled against the lot size, and whether sampling was per piece or per batch.
- Any post-treatment operation — tempering temperature range, stress relief — recorded, because a hardness reading after temper is not comparable to one before.
If your customer’s drawing or purchase order specifies a standard, test method, or acceptance criterion, that specification governs and should be quoted on your PO to the treater. Do not substitute a general practice for a contractual requirement. Where no standard is cited, agree the format with the treater and write it down; a written agreement is enforceable in a way that a phone call is not.
Incoming inspection: measure the delta per batch, not per part
Per-part records are noisy. One bracket out of tolerance tells you little; the same bracket out of tolerance in every lot tells you the process is not stable. Record the delta — post-treatment minus pre-treatment — for each critical dimension, per batch, and plot it over ten lots. You are looking for two things: the average movement, and the spread. A stable process with a consistent 0.08 mm bore growth can be compensated in machining. A process that swings between 0.02 mm and 0.15 mm cannot, and that is the conversation to have with the treater.
Keep the two failure modes separate in your records: dimensional movement and hardness shortfall. They have different causes and different remedies. A lot that is dimensionally out but hard is a support and quench question. A lot that is dimensionally fine but soft is a cycle, atmosphere, or material question. Mixing them in one complaint lets the treater answer the easy one and close the file.
When a distortion dispute arises, do not argue metallurgy
You will not win a technical argument with a heat treater about martensite transformation over chai. Run a commercial control instead.
- Count parts in and count conforming parts out. A joint trial on a representative lot, with both parties present and the counts written down and signed. The Thermal Processing article describes the same logic for validating fallout: run a representative batch, count input and output conforming parts, and document the result with the treater’s certification.
- Serialise or batch-code a sample lot so that a part found out of tolerance on receipt can be traced to its position in the load. If the parts at the basket edge are the ones that moved, you have found the quench symmetry problem without a microscope.
- Agree a written concession or rework route before the trial. If the lot comes back with 12 out of 500 out of tolerance, what happens to those 12? Rework, scrap with a credit note, or acceptance under deviation with the customer’s written approval. Decide this in advance so the trial produces a decision, not another meeting.
This is a commercial control, not a technical one, and it works with any treater in any cluster. It also gives you a number to take to your customer if the deviation needs to be declared.
What not to put in your own acceptance criteria
You will find published fallout percentages for different furnace types — figures like 1–1.5% for gas batch furnaces or fractions of a percent for vacuum furnaces. Those come from individual heat treaters’ experience and vendor knowledge bases, not from a standard, and they should not be written into your PO as an industry norm. Your acceptance criterion is whatever your customer’s drawing, purchase order, or the applicable standard says. If none of those specifies a distortion allowance, agree one with your customer in writing before you agree it with the treater.
The same discipline applies to furnace-type claims. A published account of one heat treater’s experience with large gears reports better distortion results from vacuum carburising with a hanging fixture than from pit furnaces, and notes that when distortion is consistently lower, pre-heat-treatment machining can be adjusted to reduce post-heat-treatment grinding. That is a useful direction to explore with your treater if your volumes justify it. It is not a rule that vacuum is better for your part, and it is not a reason to change your PO language.
The short version
Distortion in job-work heat treatment is a support and quench problem before it is a metallurgy problem. You control the drawing, the pre-dispatch measurement, the agreed support method, and the certificate format. Put all four on the purchase order, measure the delta per batch, and settle disputes with a counted joint trial rather than an argument. The hardness certificate tells you the part is hard. Only your own records tell you whether it is still the shape you machined.
FAQ
Does the hardness certificate cover distortion?
No. A hardness certificate records the hardness result at the sampled points on the stated scale. It does not record dimensional movement, support method, or quench conditions. Distortion has to be controlled through the drawing, the agreed fixture and support method, and pre- and post-treatment measurement.
What should I send with the parts on the first lot?
The drawing or a working sheet naming the critical dimensions and their pre-heat-treatment measured values, a written description or sketch of the agreed support and orientation, and the certificate format you require — batch identification, hardness scale and method, sample locations, number of parts sampled, and any post-treatment operation.
My treater says the distortion is from my machining stress. How do I check?
Separate the variables. Record the pre-treatment dimensions on the same setup you use after treatment, and look at whether the movement is consistent across lots or erratic. Consistent movement points to a stable process you can compensate for in machining; erratic movement points to a process control problem. If you suspect residual stress from machining, the Thermal Processing article notes that a pre-machining thermal treatment — normalising, sub-annealing, quench and temper, or full annealing — is the usual route to address it. That is a conversation to have with your own process engineer, not a conclusion to assert in a dispute.
Do I need to cite a BIS or ISO standard on the purchase order?
Only if your customer’s drawing or purchase order requires it, or if you have agreed a standard with the treater. Do not cite a standard you have not read and cannot verify applies to your part family and material. Where no standard is specified, a written certificate format agreed with the treater is the practical control.
How many parts should be in a joint trial lot?
Enough to be representative of a normal production load, and small enough that both parties can count and inspect it without disrupting the schedule. The principle is a counted, witnessed batch with input and output numbers documented and signed — not a statistically designed experiment. Agree the lot size and the concession route before the trial runs.