Heat Exchanger Inspection Methods

Purpose

This standing instruction governs HOW a heat exchanger gets inspected, so that two techs running the same furnace produce the same finding. Every method in common use has a characteristic blind spot, so the shop's rule is two independent indications before anyone declares a result in either direction.

The reason is that both errors are expensive and only one of them is visible. A missed crack sends a family into a winter of low-level exposure. A crack called on one ambiguous cue costs a customer, a neighborhood reputation and sometimes a complaint to the licensing board, and the tech never learns it was wrong. Method discipline is what separates a finding from an opinion.

Safety actions that gate the work

• Personal carbon monoxide monitor on before you enter the mechanical space, with the ambient reading recorded with the appliance off. Every method below involves standing beside a firing appliance you already suspect.
• If ambient carbon monoxide is elevated with occupants present, the appliance goes off and people leave the space before any further testing. Do not keep it firing to gather a better reading with a family in the house.
• Prove the control circuit dead before any panel, burner assembly or collector box comes off, proved on a known live source before and after per NFPA 70E-2021, 120.5, with work practices at 29 CFR 1910.333(b)(2), and close the appliance gas cock.
• The inspection itself liberates the deposit, and that is the act to control. Brushing, vacuuming or wire-wheeling scale inside a cell puts combustion particulate into your breathing zone, so use a HEPA-filtered vacuum at the deposit rather than compressed air, wear eye protection, and use respiratory protection under a written program per 29 CFR 1910.134 where the deposit clouds. Do not cut, grind or abrade unidentified cabinet liner or gasket to improve access; that is a separate scope under 29 CFR 1926.1101 for repair work, or 1910.1001 in general industry.

Scope

Covers the inspection methods themselves on residential and light-commercial gas furnaces and gas package units: as-found survey, combustion baseline, blower-transition testing, and visual and camera examination. Does not cover the evidence standard for condemning an exchanger, the tag, or the customer notification, all owned by the cracked heat exchanger condemnation and red-tag SOP; this procedure ends at the handoff to it. Does not cover boilers or oil equipment.

What each method can and cannot see

The two independent indications the shop requires must come from different rows. Two cameras of the same crack is one indication.

Roles and handoffs

Procedure

1. Record the as-found condition before the appliance is fired. Acceptance: ambient CO in ppm with the appliance off and the time, plus a written note of soot at the burner face, rust flakes in the burner box, discoloration on the cabinet, and the condition of the blower compartment seals and door gasket. Wrong looks like firing first to save time, which destroys the one reading describing what the occupants were in. Stop rule: elevated ambient with the appliance off converts the visit to the carbon monoxide alarm response SOP. Hazard: you are in a space that may still be accumulating, so the monitor is in your breathing zone.

2. Take a combustion baseline with the burner firing and the blower not yet running. Probe in the vent connector at the manufacturer's stated location, allow the readings to stabilize, and record CO, O2 and stack temperature together. Acceptance: a stable air-free CO value with the O2 it was corrected at, both written down, taken before the blower starts. Wrong looks like recording raw CO with no O2, which cannot be compared to anything later because dilution moves it. Stop rule: a baseline already above the appliance limit is a combustion fault to resolve first under the tuning SOP. Hazard: the appliance is firing with the probe in the vent, so the monitor stays on and the probe is handled by its handle, not the hot stem.

3. Run the blower-transition test and watch two things at once. Hold the probe in place through blower start and read the change in air-free CO, while a second look at the burner face records whether the flame lifts, rolls or shifts as the cabinet pressurizes. Acceptance: air-free CO changes by no more than the shop gate of 50 ppm across the transition, and the flame stays seated in the ports. Wrong looks like reading raw CO across the transition and calling the drop a good result, when dilution alone lowers raw CO. Stop rule: a change beyond the gate, or visible flame disturbance, is one indication and the appliance goes off pending a second. Hazard: this is the step with a firing appliance, an open door and a face near the burner, so keep the doors on except for the seconds the observation needs, never defeat the door switch longer than that, and keep the monitor in your breathing zone.

4. Inspect visually and by camera, cell by cell, with locations named. Acceptance: every accessible cell examined and its result recorded by cell number, with any indication photographed and located, and the areas you could not reach written down as not examined rather than left blank. Wrong looks like a single photo of a scale line labeled "cracked." Stop rule: an indication that cannot be located to a cell is inconclusive, not a finding. Hazard: cell edges cut through a thin glove, the metal holds heat long after the call, and any brushing or vacuuming inside a cell is done with the HEPA vacuum at the deposit and respiratory protection where it clouds.

5. Decide among three outcomes, and never a fourth. Acceptance: pass with two independent clean indications from different rows of the table, inconclusive with the reason and the method that would resolve it named, or hand off to the condemnation SOP with the appliance off and tagged. Wrong looks like an appliance left running with "possible crack, customer advised" on the ticket, which puts the decision on the person least able to make it. Stop rule: an inconclusive appliance stays off unless the tech can state, in writing, which two clean indications support leaving it on. Hazard: none at this step, it is a judgement made with the appliance off and the file open.

6. Where the outcome is pass, restore gas and power and prove nothing was left disturbed. Acceptance: all panels and the blower door refitted with gaskets seated, door switch confirmed to kill the unit when the door is pulled, one full call completed, and the post-repair air-free CO within 25 ppm of the burner-only baseline you took in step 2. Wrong looks like a blower door left unsealed, which by itself produces the same flame cue this SOP treats as an indication and will read as a crack on the next visit. Stop rule: a post-restore reading outside that band means something moved during the inspection, so find it before leaving. Hazard: this is where fuel and energy go back in with the family in the house, so stand to the side of the burner face and keep the monitor on until you leave the space.

The record this produces

One inspection record per visit, attached to the equipment:

• Ambient CO with the appliance off, the time, and the as-found visual notes
• Burner-only baseline: air-free CO, the O2 it was corrected at, stack temperature, and the probe location
• Blower-transition result: air-free CO before and after, the change, and the flame observation
• Cell-by-cell visual and camera results, with photographs located by cell, and the areas not examined
• Which two indications support the outcome, and from which methods
• The outcome recorded as pass, inconclusive or handoff, with the appliance state on departure

The next tech reads the cell-by-cell record so a second visit starts where this one stopped rather than repeating it. The manager reads the inconclusive rate by tech, because a tech who never records one is either unusually thorough or is resolving ambiguity by guessing.

Worked pass: 16-year-old induced-draft furnace, intermittent CO alarm history

Step 1: ambient reads 0 ppm with the appliance off at 10:14. Burner face carries light soot on burners two and three, and the blower door gasket is intact and seated.

Step 2: probe placed in the vent connector at the manufacturer's stated location. Burner-only readings stabilize at CO 55 ppm measured with O2 at 8.2 percent. Air-free is 55 multiplied by 20.9 divided by 12.7, which is 55 times 1.646, so 91 ppm air-free. Stack temperature recorded alongside.

Step 3 FAILS its acceptance and takes its stop rule. Across blower start, measured CO rises to 140 ppm while O2 rises to 9.6 percent, which is the dilution direction a leak into the flue path produces. Air-free is 140 multiplied by 20.9 divided by 11.3, which is 140 times 1.850, so 259 ppm air-free. The change is 259 minus 91, or 168 ppm air-free, against the shop gate of 50, so it is beyond the gate by 118 ppm air-free. The flame on burner three lifts and shifts toward the front within about four seconds of blower start. Appliance off. That is one indication, and the flame observation from the same transition is not counted as a second, because both come from the same test row pair and only the flame row is independent of the CO row.

Step 4: camera run cell by cell. Cells one, two, four and five show scale but no separation. Cell three shows an open crack at the upper bend, photographed and located. That is the second indication, from a different row of the table.

Step 5: outcome recorded as handoff. Appliance remains off, gas cock closed, tagged, and the cracked heat exchanger condemnation and red-tag SOP takes over for the evidence file, the tag and the customer notification. Step 6 is not run, because it applies only to a pass.

Self-audit, criterion by criterion: step 2 demanded an air-free value with its O2, and the run prints 55 ppm at 8.2 percent O2 with the correction shown inline as 55 times 1.646. Step 3 demanded a change of no more than 50 ppm air-free, and the run prints 91 and 259 ppm air-free, both corrected at the O2 measured at that same probe location, with the 168 ppm change subtracted inline rather than asserted. The raw figures, 55 and 140 ppm, are reported as raw and are not compared to the gate, which is stated in air-free. Step 4 demanded results by cell, and the run names five cells and locates the indication to cell three. The two indications counted are from different table rows, which is what step 5 requires.

References

• ANSI Z21 series listing limits for carbon monoxide in the flue products of listed vented appliances, at 400 ppm air-free, superseded by a lower manufacturer limit where stated
• The furnace service literature for the vent-connector probe location and for cell access, which are model-specific
• 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for proving dead; 29 CFR 1910.134 for respiratory protection under a written program; 29 CFR 1926.1101 or 1910.1001 for unidentified liner and gasket material
• See related: the cracked heat exchanger condemnation and red-tag SOP, which owns the evidence standard and the tag, and the gas furnace combustion analysis and tuning SOP