How to Flush a System So the Flush Removes Something

Why this matters

Most flushes fail, and they fail invisibly. Water is put through a system for an hour, it runs clear at the hose, everybody signs off, and the deposit that the flush was supposed to remove is still sitting exactly where it was, because clear water running past a settled solid is what a failed flush looks like. A flush is not a duration. It is a velocity high enough to lift what you are removing, a path that forces the water where the deposit is, a number of system volumes sufficient to dilute what is dissolved, and an endpoint you measure rather than judge. Get those four right and the flush produces evidence. Get them wrong and it produces a signature on a form.

Before water moves

Protect the supply connection. A hose from building water into a system holding cleaner is a cross-connection, and a cross-connection between a chemical-bearing system and potable water is a public health event rather than a plumbing detail. Fit the backflow assembly class the adopted plumbing code requires for the fluid category involved; where a chemical cleaner is or has been present, that is a reduced-pressure principle assembly, not a check valve.

Protect the discharge. Terminate the flush line to a drain receptor with a physical air gap, sized for the flow you are about to put down it, and confirm the drain will take that rate before you open a valve, because a flush that overwhelms a floor drain puts the water into the building.

Know what is in the water you are discharging. Flush water carrying cleaner, biocide, glycol or heavy solids is a discharge with a chemistry, and the local sewer authority sets what may go to drain and at what pH. Neutralise and dilute per the safety data sheet and the authority's limits, or collect for disposal. Take glove class, eye and face protection and any respiratory control from the safety data sheet for the specific product, and never combine a residual acid cleaner with a chlorine-bearing product in a drum or a drain line, because that combination generates chlorine gas.

Treat a long-stagnant system as biologically active. Water that has stood warm and still for months or years grows organisms that become airborne when the line is disturbed. Route the discharge into a drain rather than spraying it into open air, do not use a high-pressure spray to clear deposits from a basin or fill while people are in the space, and where the system is known or suspected to be contaminated, respiratory protection selected under a programme meeting 29 CFR 1910.134 is the control for the inhalation route.

Do not shock a hot system with cold water. Let a boiler or a hot vessel cool to the temperature the manufacturer allows before introducing cold fill, since thermal shock cracks sections and heat exchangers. Relieve pressure through a drain valve and confirm zero on a gauge before opening any joint, and confirm the fluid is below scald temperature, which burn-prevention literature puts in the region of a few seconds of contact at 140 F.

Count the building. Taking a system down means no heat or no cooling for the duration, and draining an outdoor or roof-mounted coil in freezing weather means the water left behind can freeze in the low points. Plan the outage window with the occupier, and drain low points fully or keep the section warm.

Step 1: Name what you are removing and pick the tracer that proves it left

You are removing one of three things and they need different endpoints.

Loose solids need a velocity and a catch. The proof is what you collect at the discharge and a turbidity or patch test that falls to source level.

A dissolved chemical, most often a cleaner after a chemical clean, needs dilution and a chemical endpoint. Conductivity is the best tracer because it is instant, cheap and responds to the dissolved load rather than to appearance; pH works where the product moved the pH away from the fill water.

A biological load needs disinfection followed by a residual check, and flushing alone is not the treatment.

Write the tracer and the target down before you start, and take the source reading first. A flush endpoint expressed as "matches the makeup water" is only meaningful if you know what the makeup water reads today.

Step 2: Set the velocity, not the duration

Scour is a wall shear problem. In turbulent flow the shear at the wall rises with roughly the square of velocity, a little under it because the friction factor falls slowly as Reynolds number rises, so doubling the velocity multiplies the scouring force by something in the region of three and a half. That relationship is for turbulent flow in a full pipe; it does not describe a laminar trickle through a small tube.

Which is why flushing at design flow accomplishes almost nothing. Design flow is the velocity the deposit already survived. Distribution practice commonly targets a flushing velocity in the range of 2.5 to 5 feet per second, and unidirectional flushing achieves more scour from the same supply because it forces the whole flow down one path at a time.

Convert your target velocity to a flow rate for the pipe you have. Flow in gallons per minute is about 2.45 times the velocity in feet per second times the square of the inside diameter in inches. A 2 inch branch at 4 feet per second therefore needs about 39 gallons per minute, and if the flush supply cannot deliver that, the honest options are a bigger supply, a temporary pump, or flushing a smaller section at a time. Flushing the whole system at once with a garden hose is not a slower version of the right answer, it is a different and ineffective one.

Step 3: Force the path

Water takes the easiest route, so a flush of an open system flushes the shortest path and leaves the rest untouched. Close everything except one leg, flush it to its endpoint, then move to the next. It is slower on paper and faster in practice, because the alternative is flushing the whole system three times and still missing a branch.

Bypass what you must not flush through. A plate heat exchanger, a pump, a control valve with a soft seat and any small-orifice device is a place for a dislodged slug to lodge, and a component already thinned by erosion can be perforated by a high-velocity slurry. Route around them, or fit a temporary strainer immediately upstream of them and inspect it after each leg.

Open and flush the legs that have no flow in normal service, because that is where the deposit is. Dead legs, capped stubs, bypass lines and isolated branches are the whole reason a flush is being done, and they are the parts a lazy flush never reaches.

Step 4: Treat volumes as a floor, and the measurement as the answer

You cannot flush a volume with a volume. In a well-mixed vessel the concentration of what you are displacing falls exponentially with the number of system volumes passed through: about 63 percent gone after one volume, 90 percent after 2.3 volumes, 99 percent after 4.6 volumes. A straight pipe run behaves closer to plug flow and clears in a little over one volume; those two ideals bracket a system with no bypassed volume, and only such a system. Add a closed balancing valve, a dead leg or an unequal parallel path and the real result falls outside the bracket, worse than the well-mixed case, because part of the volume is not being flushed at all. That is why a count of volumes is a floor and the tracer is the endpoint.

So compute the volumes as a floor to know whether you are asking for one hour or six, then flush to the tracer. The calculation tells you what is possible; only the measurement tells you what happened, and the gap between them is diagnostic in itself.

Step 5: Catch what comes out

Fit a bag or basket strainer at the discharge, or run the last part of each leg through a clear section. At the end of the leg you should be able to hold up what the flush removed. A flush that produces nothing from a system that was demonstrably dirty did not clean the system, it rinsed the parts that were already clean, and the correct response is more velocity or a different path rather than more time.

Step 6: Restore, re-treat, record

A flushed system is an unprotected system. Bare metal that has just had its deposits and its old chemistry removed corrodes quickly in aerated water, so refill and dose to the treatment target promptly rather than leaving a system standing full of raw water over a weekend. Record the volumes flushed, the tracer readings at start and finish, the endpoint reached per leg, and the makeup meter reading, because the next person needs to know what was actually achieved rather than that a flush occurred.

Worked example: a 900 gallon hydronic loop after a chemical clean

The loop holds about 900 gallons. After the cleaner has done its work, the loop water reads 1,800 microsiemens per centimetre against makeup at 640. The target is to get the loop within 10 percent of makeup, so the excess above makeup has to fall from 1,160 to about 64.

Take the ratio: 64 divided by 1,160 is 0.055, and the number of well-mixed volumes needed to reach that is the natural log of 1 over 0.055, about 2.9 volumes. That is 2,610 gallons. At a flush rate of 60 gallons per minute through one 2 inch branch at a time, which is about 6 feet per second in that branch and well above the transport threshold, the arithmetic says roughly 44 minutes of flowing water.

They flushed three full volumes, more than the calculation asked for, and measured 780 microsiemens per centimetre. Excess above makeup was 140, against a predicted 58 at that point. Measured excess was about 2.4 times the model.

The gap is the finding, not an error. The exponential model assumes a well-mixed volume, and a hydronic loop is not one: two coil branches were sitting behind balancing valves that had not been opened, holding several hundred gallons of cleaner-laden water that the flush had been going around. They isolated the main loop, flushed each coil branch individually to its own endpoint, noting that a bypassed branch discharges at close to full cleaner strength rather than at the loop's diluted reading, so pH and the sewer limit were re-checked on that slug specifically and it was throttled to a rate the receptor could take, and the next full-loop reading came back at 660, an excess of 20 and comfortably inside target.

Read what that sequence proves. The calculation set a floor and was right about the floor. The measurement caught the short circuit, and the size of the discrepancy pointed at a bypassed volume rather than at a slow flush, because a slow flush converges toward the model rather than plateauing above it. Had they trusted the model and stopped at three volumes, the loop would have been re-dosed with treatment on top of a residual cleaner, which in most programmes means the inhibitor is consumed neutralising a chemical that should have left.

How to verify you got this right

Compare the closing tracer reading against the source, taken the same day, not against the reading you wrote down last week. Supply chemistry moves.

Read every leg individually rather than the return header, because the header reads as the average and the average hides the leg that did not clear.

Confirm the differential pressure and flow through each heat exchanger and coil against its clean curve after refilling. A flush that succeeded chemically and left a slug parked in a coil shows up here and nowhere else.

Keep the catch. The bag or basket contents, photographed against the leg it came from, is the record that the flush removed something, and it is the difference between a report and a claim.

References

• AWWA guidance on distribution system flushing, including unidirectional flushing practice and target flushing velocities
• ASHRAE Standard 188, Legionellosis: Risk Management for Building Water Systems, where a stagnant or biologically active building water system is being returned to service
• OSHA 29 CFR 1910.134, respiratory protection, and 29 CFR 1910.1200, hazard communication, for the controls and safety data sheets governing cleaner residues and aerosol exposure during flushing
• The adopted plumbing code in your jurisdiction for backflow assembly class at the supply connection and air gap at the drain termination, and the local sewer authority for discharge limits
• See related: What Suspended Solids Do Once They Are Inside a System; Why a New System Needs Cleaning Before It Needs Treating