Furnace and oven temperature mapping

Most of what is written about temperature mapping assumes a cold room. The arithmetic transfers to a furnace or a drying oven without much change, but three things do not, and they are the three that decide whether a survey is worth anything: the number of set points, the sensor's contribution to uncertainty, and where the acceptance limit comes from.

Uniformity is a function of set point

A 2–8 °C cold room runs at one set point, so one mapping study describes it. A furnace qualified from 150 °C to 650 °C does not have one uniformity figure — it has a different one at each temperature. Radiant heat transfer rises with the fourth power of absolute temperature, so the balance between radiation, convection and conduction shifts across the range, and the pattern of hot and cold zones shifts with it. A survey at 200 °C tells you very little about 600 °C.

The working rule in heat treatment is to survey at the extremes of the qualified range and at any set point in routine use, then re-survey on a schedule. If you only ever run two recipes, two surveys describe the furnace. If the range is continuous, the extremes plus the middle is the usual minimum, and the standard you work to will say.

The sensor stops being free

In a cold room a calibrated logger contributes perhaps 0.2–0.3 °C (k = 2) and the chamber contributes degrees, so the sensor is a rounding error. At furnace temperatures that inverts.

The practical consequence: state the survey's measurement uncertainty and compare it against the tolerance before you start. A ±3 °C uniformity requirement measured with an instrument chain carrying ±4 °C expanded uncertainty cannot be demonstrated, whatever the readings say. That is the same decision-rule problem as any conformity statement — the tolerance belongs to the furnace and the uncertainty belongs to your measurement of it.

Empty, loaded, and the thermal mass problem

An empty furnace reaches set point quickly and holds it tightly. Load it and both change: the charge absorbs heat, the recovery after the door closes lengthens, and the coldest point usually moves into the middle of the load rather than staying at a wall. For a cold room the empty-versus-loaded argument is about air circulation. For a furnace it is about thermal mass, and the loaded survey is the one that describes production.

Soak time follows from this. The work zone is at temperature when the slowest-responding part of the load is at temperature, which is not when the controller says so. A survey that starts timing from the controller reaching set point will overstate uniformity, because the fast-responding sensors have stabilised and the load has not.

Where the acceptance criterion comes from

This is the part people get wrong, because the method looks universal and the pass mark is not.

• Aerospace heat treatment works to AMS2750, which defines a Temperature Uniformity Survey, a System Accuracy Test, instrumentation types and furnace classes graded by permitted uniformity. The class you claim sets the number.
• Automotive heat treatment works to CQI-9, which covers similar ground through a different assessment process.
• Pharmaceutical ovens and dry-heat tunnels work to Annex 15 for the qualification structure, with the lethality or depyrogenation requirement coming from the process rather than from a uniformity table.
• Laboratory ovens with no sector standard fall back on the manufacturer's specification and your own risk assessment — which is legitimate, provided the criterion is written down before the survey rather than chosen after it.

Do not take a uniformity tolerance from an article, this one included. Classes and tolerances are defined in the current revision of the standard you are audited against, and they change between revisions. Open the document.

What a hot survey has in common with a cold one

Everything else. Sensor count still scales with volume and with the number of known risk points; the geometric centre and the corners of the work zone are still mandatory positions; the sampling interval still has to be short enough to catch a recovery excursion; data quality still has to be checked before any statistic is believed. If you have run a cold-room study, you already know how to run this one — you are changing the sensor, the number of set points, and the document you take the pass mark from.