Cleaning the System Before the New Boiler Goes On the Wall

Two winters later, on a photograph of a filter

The appliance is twenty-six months old, the secondary heat exchanger is blocked, and the manufacturer's engineer has been on site forty minutes, thirty of them taking photographs. One of the in-line filter, opened, with the magnet lifted out beside it. One of a jam jar of system water drawn from the drain cock on the return, held against a white wall. One of the commissioning record, where the section headed cleaning and water treatment carries a tick and nothing else. The claim is refused, and the argument that follows turns on what was done to the water two years ago — a question nobody can now answer.

This is not rare and it is not really a dispute about the boiler. Almost every wall-hung domestic appliance now carries a guarantee conditional on the system it is connected to: cleaned before the appliance goes on, dosed with a corrosion inhibitor at the concentration the manufacturer names, fitted with an in-line filtration device, and re-tested at every service. In the UK the instructions usually discharge that obligation by pointing at BS 7593, and at the appliance's own manual for whatever the code leaves to the manufacturer; on a commercial job BSRIA BG 29, pre-commission cleaning of pipework systems, sets out the same logic at a larger scale.

So the clean is a contractual step carrying an evidence requirement, and the evidence has to be made on the day. Two decisions hold it up: which cleaning operation this system needs, and what number the inhibitor dose is calculated against. The first, wrong, costs a day. The second costs the heat exchanger and the guarantee, silently, two seasons after anyone could still have fixed it for the price of a second bottle.

A jam jar, and the survey you do before quoting

Draw a sample before pricing the job, not on the morning you start it. A drain cock on the return, a clear jar, half a litre and five minutes of standing still tells you more about which operation this system needs than an hour of looking at radiators.

Black sediment that follows a magnet held to the glass is magnetite, the iron oxide produced when oxygen keeps reaching steel. A heavy load means an oxygen source running for years, and the useful question is which one: a feed and expansion tank open to the loft, a failed vessel letting the system go sub-atmospheric as it cools, barrier-free underfloor pipe, or a top-up habit. Brown material staying in suspension is finer corrosion product; water clear but strongly coloured is dissolved product, an old inhibitor or somebody's cleaner from a previous visit. Grey, gritty, non-magnetic solids are building debris, flux and jointing residue, and they matter because the magnetic filter everyone is about to fit will not hold them.

Then walk the system, because access rules a power flush in or out more often than chemistry does. Look for microbore tails, since a bore already half occupied will not pass mobilised debris and may not survive being asked to. Note cast iron columns, which hold a multiple of a steel panel of the same output and have been settling a bed in the bottom for decades. Establish whether an underfloor circuit carries an oxygen barrier to DIN 4726, because a barrier-free loop is a permanent oxygen source no clean will outlive. Two things then change the plan more than anything else: whether the system is open vented or sealed, since a swap frequently converts it, and whether a gravity primary or converted single-pipe circuit has left dead legs where nothing moves at any velocity you can generate.

1. Draw half a litre from the return drain point into a clear container and leave it standing.
2. Hold a magnet to the outside of the glass and watch whether the settled solids climb it.
3. Note the system type and the pressure or static head it runs at now.
4. Identify the smallest bore in the distribution, and any underfloor loops with the pipe's barrier specification.
5. Locate the drain points, the filling arrangement, and a discharge route for the volume you are about to move.
6. Open the existing filter; one nobody has serviced is a survey result in itself, and photograph what is in it.

Chemistry or a machine, and what each one genuinely moves

The trade argues this as a binary and it is not. Several operations are available, they do different work, and on most jobs the right answer is two of them in sequence. What separates them is not how aggressive they are — it is whether they supply velocity, because velocity is what carries loosened material out of a system rather than round it.

A chemical cleanse circulated by the system's own pump means dosing a cleaner, running hot with every emitter open for the dwell the product's data sheet gives, then draining and rinsing. On a system chemically dirty but not physically silted that is the correct answer, and its limit is arithmetic rather than chemical: the circulator was sized for design flow, design flow is a low velocity through the wide part of a radiator, and no chemistry alters the fact that a settled bed lies below the moving water rather than in it. A power flush answers that limit with an external machine supplying far more flow, reversing it, and letting you isolate and agitate one emitter at a time while dumping to waste. It is also the operation most likely to find every weeping compression joint in the building, which is a real cost and belongs in the conversation before the machine comes off the van.

The order of the day, and the one connection you never flush through

The sequencing matters more than the method argument and gets less attention. Whatever you do, you do it before the new appliance is in circuit — with the old boiler still connected, or through a temporary jumper across the flow and return where it was, but never through the heat exchanger you have just unwrapped. Its waterways are narrower than anything the old appliance had, the whole exercise exists to protect it, and pushing years of accumulated magnetite through it on day one is a self-inflicted version of the failure at the top of this page.

The same logic covers new pipework the swap brings. Copper and press fittings arrive carrying swarf, cutting oil and flux, so make that work up complete, flush it clear on its own, and connect it afterwards. A low-loss header or hydraulic separator goes in after the flush too: it is a low-velocity vessel by design and will collect precisely what you were trying to remove.

Dose last. Inhibitor goes in at the final fill, after the last drain-down anybody intends, and the reason is bookkeeping rather than chemistry — a partial drain to swap a valve afterwards removes an unrecorded share of the treatment and nobody adjusts for it. Where a further drain is unavoidable, recalculate and record the correction.

1. Clean with the old boiler in circuit, or through a temporary flow-and-return jumper.
2. Work emitter by emitter where the survey called for a machine, until each runs clear on its own.
3. Rinse until the discharge matches the incoming fill water, not until a set time has elapsed.
4. Fit the appliance, the filter and any separator, with new pipework flushed clear beforehand.
5. Refill to the cold fill pressure the height requires, and check the vessel before going further.
6. Bring the system to design flow temperature, then dose against the measured volume and circulate.
7. Draw the handover sample, test it, and have the appliance certificated by a person registered for the fuel.

Fill water is a material, and almost nobody specifies it

Every other material on the job arrives with a specification and this one arrives through a hose. What goes in at the final fill decides two failure modes over the appliance's life — scale, from hardness, and corrosion, from oxygen and dissolved salts — and where the clean addresses the history, the fill water addresses the future.

Scale is worse in a new appliance than in the old one, and not because the water changed. Deposition is fastest where heat flux is highest, and a condensing heat exchanger concentrates a large output into narrow waterways with a small water content — the reason it modulates well, and the reason a millimetre of scale in it is not equivalent to a millimetre in a cylinder. VDI 2035 Part 1 handles that by making permissible fill-water hardness depend on the installation rather than on the water alone: the total heating output and the system's specific volume in litres per kilowatt both enter the limit, and the direction of that catches people out. The more litres a system holds for each kilowatt of output, the stricter the permissible hardness becomes, because every one of those litres charges its own hardness through the same heat exchanger surface. Adding a buffer or a thermal store is what usually moves a domestic job into the band where the fill water has to be treated rather than simply run in off the main. Part 2 covers the corrosion side.

Where the main is hard, the two available answers are not equivalent and the difference is regularly missed. Demineralisation through a mixed-bed cartridge on the filling line removes dissolved salts and drops conductivity — the condition VDI 2035 calls low-salt operation, which several manufacturers now ask for by name. Base exchange softening swaps calcium and magnesium for sodium, preventing hardness scale but leaving dissolved solids and conductivity broadly where they were. Softened water is a scale answer, not a corrosion answer.

The chemistry you add has to suit the metals now in the system, and a swap often changes them. An aluminium-silicon heat exchanger is amphoteric and is attacked at high alkalinity, so appliances built with one specify a narrower pH window than an all-steel-and-copper system tolerates; that window is in the appliance's instructions, not a figure to carry over from another manufacturer's job. BS EN 14868 sets out how corrosion likelihood in a closed circulation system is assessed. Read the manual before choosing the cleaner as well as the inhibitor: an acidic cleanser left in contact beyond its stated dwell has ended more aluminium heat exchangers than dirty water has.

The clean has a finish line, and it is a sample rather than a clock

Nothing here is finished at a time. It is finished at a water condition, and the difference is why so many flushes stop early with everyone satisfied. Debris leaves the pipework quickly because velocity there is highest, so a sample at the boiler drain point runs clear long before the system is clean — the emitters and the dead legs give up their contents last, if at all, and they are where the deposit was.

Sample at the end of the system as well as at the plant, and compare what comes out against a jar of the incoming fill water standing alongside it; that comparison is what turns an opinion into a reading. Where the appliance instructions state parameters to be met at handover — appearance, pH, conductivity, chloride, inhibitor concentration — those are the acceptance criteria. Then photograph the sample, photograph the filter contents at first service, and keep both with the paperwork, because the scene at the top of this article is an evidential contest and evidence made on the day beats recollection two years later.

Systems that cannot be cleaned, and the value of saying so first

Some systems will not come clean and no operation on the table changes that. A microbore distribution already occluded to the point where a machine cannot establish flow is one. Barrier-free plastic underfloor pipe is the clearest case: oxygen diffuses through the pipe wall continuously, magnetite generation restarts the day you refill, and the only durable answer is hydraulic separation with a plate exchanger so the ferrous components sit on the other side. A corroded steel emitter holding water only because the sediment inside it is doing the sealing will discover that during the flush, in a room with a carpet in it.

Put this in the quotation rather than in an apology. A manufacturer will not warrant a heat exchanger against a system that cannot meet the water-quality condition, so a system in that state changes what the job is: separation, partial re-pipe, emitter replacement, or in the honest cases a conversation about the distribution being at the end of its life. All of those are easier to have while the old boiler is still on the wall and working than at four o'clock on a Tuesday with a hole in the plaster where it used to be.

The record that answers the engineer in two years' time

Everything above produces one artefact, and it is the artefact the opening scene was missing. The record needs the measured system volume and how it was arrived at; the cleaning operation, the product, and the contact time it actually had; the handover sample result against whatever parameters the appliance instructions state; the inhibitor product and the dose in litres against volume; the filter make, model and position; the vessel size, pre-charge and cold fill pressure; and the date with a name against it. In the UK the Benchmark commissioning checklist supplied with the appliance is the format the manufacturer expects, and its water treatment sections are the ones that decide a claim.

Register the appliance inside the manufacturer's window, and tell the customer in writing what the guarantee is conditional on: an annual service that includes opening the filter and testing the water, and no change that adds volume without a re-dose. A radiator added by somebody else in year three dilutes the treatment as effectively as a leak does, and the customer is the only person in a position to know it happened.

The last thing worth doing costs nothing. Keep the survey photographs, the sample photographs and the test results with the record. A refused claim is decided on what can be evidenced, and the day spent cleaning a system properly is worth exactly what you can later prove about it.