Furnace Acceptance Testing: What to Check Before Shipment and After Installation

Accepting an atmosphere furnace is not a formality. One machine carries five systems — heating, atmosphere, control, drive and safety — and if any of them is signed off without a real test, the problem waits until mass production. By then most of the money has been paid and your negotiating position is gone.

Two kinds of bad acceptance are common. One is a walk around the furnace at the factory, a judgement that the workmanship looks good, and a signature. The other is powering up at site, reaching set temperature, and calling it passed. Both miss the measurements that actually determine whether you can produce.

The workable approach is two stages: factory acceptance (FAT) to confirm the machine was built to the drawings, and site acceptance (SAT) to confirm it still performs in your building. Each stage has its own list.

Factory acceptance: was it built correctly

FAT cannot be done in your own workshop. Go to the supplier's site, or at minimum require the complete test records and video. The purpose here is consistency between design and build.

Structure and appearance

• shell welds continuous and undistorted; insulation packed properly, which you can sample through an inspection opening
• refractory free of cracks and spalling, anchors in place
• door, observation ports, thermocouple ports and gas connections in the positions shown on the drawings
• nameplate data — power, voltage, phases, atmosphere type, maximum temperature — matching the contract

Heating and electrical

• heating element type, specification, quantity and connection matching the drawings
• each zone independently controllable, thermocouple type and calibration correct
• components in the electrical cabinet matching the contracted bill of materials, terminals numbered
• insulation resistance and earth resistance test records

Gas train and sealing

• routing, valve types (manual, pneumatic, motorised), flow meter and gauge ranges per drawing
• leak test record: test pressure, hold time, pressure drop
• safety valve, explosion relief and flame arrestor installed

Functional tests — the part most often skipped

• heat the empty furnace to maximum temperature and record the curve and time to reach it
• control accuracy per zone and temperature fluctuation during soak
• test every safety interlock individually: simulate high oxygen, low pressure, power loss, door not closed, and confirm the hydrogen or atmosphere valve acts as designed
• trigger every alarm and confirm the corresponding message appears on the HMI

Interlocks are the priority. Many suppliers will demonstrate normal operation but will not volunteer a demonstration of failure behaviour. Failure behaviour is what you need. Require each item to be simulated and sign the record.

Site acceptance: does it still work in your plant

Once the machine is in your building, the variables change: voltage, gas purity, water quality, ambient temperature, exhaust conditions. Any one of them can degrade performance. SAT is about performance under load.

Installation checks

• foundation level, anchor fixing, flexible connections between furnace and ducting
• phase sequence, voltage deviation, earthing
• gas supply purity and pressure against specification — most often overlooked, and water or oil in the line will ruin every atmosphere figure downstream
• exhaust routing, discharge location, flashback protection

No-load tests

• heat to working temperature and record time and energy
• measure temperature uniformity once stable with an empty chamber
• measure purge time and gas consumption to reach the set atmosphere

Loaded tests — the decisive ones

• load with the real load weight and the real tray arrangement, and run the full process curve
• measure temperature uniformity under load, at least five points covering top, middle, bottom and both ends
• measure atmosphere under load — oxygen or dew point — sampling where the gas is hardest to displace
• run three to five consecutive cycles, or 24–72 hours continuously, and watch stability

One point deserves emphasis: empty-chamber uniformity is not loaded uniformity. With an empty chamber the gas flows freely and radiation is even, so the numbers look excellent. Once loaded, temperature differences appear inside the stack and at the edges. If the contract says "uniformity ±5 °C" without stating test conditions, the argument is guaranteed. Write it into the contract: loaded, full load, how many points, how long after stabilisation, what tolerance.

Safety and interlock retest

Interlocks tested at the factory must be retested at site. Transport vibration and field wiring errors can defeat them. The power-loss test matters most: pull the main breaker and confirm the atmosphere valve closes, alarms trigger, and chamber pressure holds.

Documents to collect

Acceptance is not only a signature; it is a set of documents that will support maintenance for the next decade. Without them, replacing a heating element in three years becomes a research project.

• equipment certificate and certificates for major bought-in components
• general arrangement drawing, chamber section, electrical schematic, wiring diagram, gas diagram (paper and electronic)
• operation manual, maintenance manual, fault code list
• calibration certificates for instruments and sensors
• factory test records: heat-up curve, leak test, interlock tests
• spare parts list and recommended spares, with model numbers and suppliers
• drawings or specifications for wear parts: elements, thermocouples, seals, rollers

Put this in the contract: incomplete documentation means no acceptance. Otherwise the machine works and you remain dependent.

The four most disputed points

1. How uniformity is measured. Empty chamber gives ±3 °C, loaded gives ±10 °C, and the supplier says the contract did not mention load. The test condition must be in the contract.
2. Heating rate. A stated "degrees per hour" rarely says empty or loaded, or from what starting temperature. Slower heating under load is normal; what matters is which condition the contract names.
3. Atmosphere figures. At what state was 50 ppm measured — empty and static, or loaded and dynamic? Where is the sampling point? Unclear answers guarantee a dispute.
4. Capacity. The contract says "X tonnes per day" without specifying load arrangement or soak time. When output falls short, the answer will be that you loaded it wrong.

A practical measure: put all of these into the technical agreement and state that site testing under load is the governing measurement. Agree the test method in writing before testing starts, so interpretation is not negotiated afterwards.

Retention and warranty

The usual practice is to hold around 10% as retention, released after acceptance and finalised at warranty expiry. Warranty is typically 12 months, and it should run from the acceptance date, not the shipment date — make that explicit.

Also agree separately: the rectification period for failed items, and what happens if rectification still fails. For custom equipment, returning the machine is rarely realistic, so "rectify, deduct, extend warranty" is the workable clause.

ZKR tests leak tightness, heat-up curves and interlock functions before shipment and delivers the records with the machine. For hydrogen and other flammable atmospheres, the interlock list and purge sequence are written into the technical agreement at quotation stage, and acceptance is checked against that same list. Power range is 100–2000 kW, configured per project.

Frequently asked questions

Do I have to travel to the factory for FAT?

Ideally yes, particularly for custom or first-of-type equipment. If you cannot, require complete test video and records, and state that factory records form part of the acceptance documentation — no records, no acceptance.

How many points for temperature uniformity?

Five-point and nine-point methods are both common. What matters is that points cover top, middle, bottom and both ends, and that the test is under load. Too few points is unrepresentative, too many is impractical on site; five to nine is a realistic range.

How long should loaded testing run?

At least three complete cycles, or 72 hours continuous. The purpose is stability, not a single good batch. A first cycle that passes and a third that drifts is a common pattern.

If minor issues appear, should I sign?

Separate functional defects from cosmetic ones. Functional defects — non-uniform temperature, atmosphere out of specification, interlocks not acting — should not be signed off. Cosmetic and labelling issues can be signed with a rectification list, a deadline, and corresponding money withheld.