How to read a temperature uniformity survey report

A uniformity survey is one of the few documents where the reader is usually not the person who produced it. An auditor, a customer’s quality department, or you in eighteen months looking at a furnace that has started drifting — all reading a PDF and deciding whether to trust it.

This is about what has to be in that document for the answer to be yes, and about three specific ways a survey can look finished while being unusable.

The two numbers people confuse

Almost every argument about a survey report is really an argument about spread versus maximum deviation, and they answer different questions.

Spread is the difference between the hottest and coldest sensor at a given moment. It describes the work zone: how much variation exists inside the furnace.

Maximum deviation is the largest difference between any sensor and the setpoint. It describes the furnace’s relationship to the temperature you asked for.

They are not interchangeable, and a survey can legitimately have a spread larger than its tolerance while passing comfortably. Consider five sensors at 903, 902, 898, 897 and 901 °F against a 900 °F setpoint with a ±10 °F tolerance. The spread is 6 °F. The maximum deviation is 3 °F. Now shift every sensor 4 °F warmer: the spread is unchanged at 6 °F, but the maximum deviation is 7 °F. Same uniformity, different conformance — because uniformity is about the relationship between sensors, and conformance is about the relationship to setpoint.

The practical consequence for reading a report: if the summary table shows a spread next to a tolerance, check which quantity the tolerance actually governs. A report that puts spread beside the tolerance and prints PASS is not necessarily wrong — but it cannot be reconciled without knowing that the comparison being made is against a number in a different column.

What the record has to show

A survey report is a claim that a furnace held temperature within tolerance across its work zone. Everything below exists so a reader can check that claim rather than accept it.

The furnace, identified. Make, model, serial number, and the work-zone dimensions the survey covers. A survey is only valid for the volume it was performed in — if the usable zone was later changed, the survey describes something that no longer exists.

The qualified temperature range, and the temperatures actually surveyed. These are two different things and the gap between them matters. A furnace qualified from 600 to 1800 °F but surveyed only at 900 and 1400 °F has no evidence at its extremes, which is where a furnace is most likely to fail.

Every sensor, with its calibration status. Type, serial number, calibration date, calibration expiry, and the correction factor applied. A survey performed with an out-of-calibration sensor is not a survey — and the reader can only establish that if the dates are printed.

Correction factors, and the corrected values. If a factor was applied, both the raw and corrected readings should be visible. A single column of numbers with no indication of whether they are raw or corrected is not auditable, and a reader cannot tell which by looking.

Sensor placement. A diagram or a position table. “Nine sensors” tells you nothing; nine sensors clustered in the middle of the work zone tells you the survey was designed to pass.

The recording window, stated explicitly. When the temperature was reached, when stabilization was judged complete, and when data collection began and ended.

The samples themselves. An appendix of the readings the statistics were computed from. This is the part most often omitted and the only part that lets a reader recompute anything.

Three checks that catch an unusable survey

These are not exotic. Each one takes seconds and each catches a report that would otherwise be filed and relied on.

1. Does the chart’s time axis agree with the recording window?

Take the recording start and end printed in the temperature table, then look at the trace chart’s x-axis. They should match.

If the axis starts before the trace does, the statistics may have been computed over a window that includes the ramp — and uniformity is evaluated on the soak, after all sensors are within tolerance, not while the furnace is still climbing. A survey that includes ramp data can fail a furnace that passed, or, more insidiously, report a hot spot that is really an overshoot on approach.

If the axis and the appendix disagree by a constant offset — four hours, five hours — you are looking at a timezone problem, and it means at least one number in the document is in a different clock from the rest.

2. Does the appendix’s first reading fall inside the stated window?

Open the sample appendix and compare its first timestamp against the recording window on the summary page. The record should start where the window starts.

If the appendix begins before the window, the statistics may include data the survey should have excluded. If it begins well after, part of the window was not recorded.

3. Do the statistics and the chart tell the same story?

If the report states a hot spot, a cold spot, a spread and a maximum deviation, those numbers came from somewhere. Find them on the chart. The hot-spot sensor should visibly be the highest trace; the maximum deviation should be visible as the largest excursion from setpoint.

A statistics table beside a chart that does not show what it describes — or beside no chart at all, or beside a line saying no data was recorded — means the two halves of the document came from different code paths, and at most one of them is right.

Stabilization is not the same as reaching temperature

The most common structural mistake in a survey record is treating “the furnace reached setpoint” and “the survey may begin” as the same moment.

They are not. Sensors reach tolerance at different times — that is the entire point of a uniformity survey — so the window opens when the slowest sensor has settled, not the first. Data collected before that describes a furnace still equalising, and it will make the work zone look less uniform than it is.

A report should therefore show three instants and not two: when the temperature was reached, when stabilization completed, and when recording ran. If a report shows stabilization and start as the same timestamp, either the furnace stabilised instantly or the record is not distinguishing them.

What a survey does not tell you

Worth stating, because surveys get over-read.

A survey describes one furnace, in one configuration, on one day, empty. It does not describe the furnace loaded, and load changes airflow. It does not describe the furnace after a heating element is replaced. And a passing survey says nothing about whether the process running inside it is capable — a perfectly uniform furnace will happily hold the wrong temperature.

ProcessView+ produces these reports from recorded sensor data rather than from typed-in values, which is the difference between a survey record and a survey transcript. If you want to see the report structure before committing to anything, the free trial generates one against a built-in virtual controller.