Fire Test Headers: Sight Glass, Turbine Meters & Closing Time

A fire pump test header fails in three quiet ways: a sight glass window that was sized by thickness alone, a flow meter that reads a few percent wrong because of its pipe runs, and a valve that closes faster than the pipe can tolerate. Each one can be checked with arithmetic before the header is built. Here are the sums we use at LEDE, with some common figures corrected.

• Window stress rises with the square of aperture over thickness, so a 40 mm window on 12 mm glass carries about 2.1 MPa at a conservative allowable stress
• A meter quoted as ±0.5% of full scale reads up to 2.5% off at one fifth of range
• Closing a DN150 test valve in 20 seconds instead of instantly cuts the pressure rise from about 6 MPa to about 50 kPa

How Thick Should the Glass in Sight Flow Indicators Be?

Thick enough for its aperture, and the aperture matters more than most datasheets admit. A flat glass window held in a soft gasket behaves like a simply supported circular plate. Under pressure p, the peak tensile stress at the centre of the outer face is about σ = 1.2 × p × (a ÷ t)², where a is the clear radius and t is the thickness. The 1.2 comes from a glass Poisson's ratio of about 0.2.

Work an example. Take a 40 mm clear aperture, so a = 20 mm, on 12 mm glass. Then (a ÷ t)² = 2.78. At 2.5 MPa the stress is 1.2 × 2.5 × 2.78 ≈ 8.3 MPa. Glass fails in tension, and its safe working stress is much lower than its lab strength. A figure near 7 MPa is often used for annealed borosilicate, and thermally toughened glass is allowed several times more. Ask your sight glass maker for the figure and the standard it comes from. At 7 MPa the same window is good for about 2.1 MPa. Shrink the aperture to 30 mm and (a ÷ t)² falls to 1.56, which lifts the allowable pressure to about 3.7 MPa. A rating of 7.5 MPa burst on 12 mm glass would put the centre stress near 25 MPa, which only toughened glass could survive. So we treat a burst figure as a target to confirm with a test report, and we size from the working-stress calculation.

Now the material. The common story is that soda-lime glass shatters in water hammer and borosilicate does not. The calculation says something different. Borosilicate 3.3 has a low expansion coefficient of about 3.3 × 10⁻⁶ per K, against about 9 × 10⁻⁶ for soda-lime. A cold test flush hitting a window warmed by sun or hot plant air makes thermal stress of about E × α × ΔT ÷ (1 − ν). For a 50 K shock that is roughly 13 MPa in borosilicate and 41 MPa in soda-lime, as an upper bound. This thermal shock is where borosilicate earns its place. Mechanical strength depends on thickness, diameter and edge support, not on glass type alone.

Finally, can an inspector see the flow? A sprinkler with K = 5.6 at 7 psi gives 5.6 × √7 ≈ 14.8 gpm, about 56 L/min. In a DN32 sight flow indicator with a 35 mm bore, that is about 0.97 m/s. A vane that turns at 0.15 m/s, which is our target, will respond to about one sixth of that flow. We ask suppliers to state the minimum visible flow, not just the pressure rating.

How Do Turbine Flow Meter Runs and XGQT1 Couplings Protect Accuracy?

A turbine flow meter wants a steady, swirl-free profile. Swirl and uneven profiles push the rotor speed away from the calibration. The usual cure is straight pipe: 10 diameters upstream and 5 downstream. In DN150 with a 154 mm bore, that is 1.54 m and 0.77 m. At 150 percent of a 1,000 gpm pump, flow is 94.6 L/s and velocity is about 5.1 m/s, so the profile has little time to recover from a bend.

Alignment helps keep those runs straight. The XGQT1 rigid coupling locks the pipe ends with angled pads. Our target is to hold angular play to a fraction of a degree, so the run does not kink at a joint. We have no test report we can quote for 0.05°, and it should be treated as a design target.

Put numbers on accuracy with a simple root-sum-square. Assume the meter is ±0.5% of reading, the readout adds 0.1%, and installation effects add 0.5% when the runs are right. The total is √(0.5² + 0.1² + 0.5²) ≈ 0.7%. Assume installation effects of 2% when the runs are wrong, and it becomes 2.1%. The straight pipe is worth more than a better meter.

Read the accuracy statement carefully. A figure of ±0.5% of full scale is not ±0.5% of reading. At 20 percent of range, ±0.5% FS means ±2.5% of the reading. Ask which one the datasheet means.

An orifice plate wears in a different way. As the sharp edge rounds off, its discharge coefficient rises. If Cd goes up by 5 percent while the calculation still uses the old value, the indicated flow is low by 1 ÷ 1.05, which is about 4.8 percent. This is why sight glass and a turbine meter are paired: one is a quick visual check, the other is the number.

Bolting decides whether the coupling stays put. Follow the bolt tightening method for grooved fittings in three passes, at about 30, 60 and 100 percent of the rated torque, alternating from side to side. There is a geometric reason for alternating, beyond friction. An M16 bolt has a 2.0 mm thread pitch, so each full turn on one nut closes that side by 2 mm. If a crew runs one nut down three turns while the other is snug, the other pad can stand open by up to about 6 mm. The gasket lip is squeezed into that opening. Alternating in half-turn steps keeps the unequal opening to about 1 mm.

The tilt also bends the bolt. Bending stress is roughly E × d × φ ÷ (2L), where φ is the tilt in radians and L is the free length. With d = 16 mm, L = 40 mm and a tilt of 0.25°, that is about 180 MPa. At 0.5° it is about 370 MPa. Add roughly 320 MPa of axial stress at a preload of 50 kN, and the total passes the 640 MPa yield of an 8.8 bolt. Delayed fracture in grade 8.8 bolts is rare. The more likely failure is yield or fatigue under pressure pulses. We state these as worked examples and ask installers to follow the coupling maker's torque.

Fire Test Header Comparison by Build

Property	LEDE Sight Glass, Turbine Meter & XGQT1 Test Header (Targets)	Single-Pane Soda-Lime Viewport Line	Welded Flange and Orifice Header	Basic Threaded Test Line
Window sizing basis	Working stress from aperture and thickness	Thickness only	No window	No window
Thermal shock margin	Borosilicate 3.3, about 13 MPa at 50 K	About 41 MPa at 50 K	Not applicable	Not applicable
Flow reading	Turbine meter, ±0.5% of reading (target)	Visual only	Orifice, drifts as the edge wears	None
Straight-run support	10D upstream and 5D downstream, rigid joints	Not defined	Welded, rigid	Threads, can sag
Bolt control	Three-pass alternating sequence	Wrench feel	Star-pattern flange torque	Not applicable
Valve closing control	Programmable actuator, S-curve	Manual handwheel	Manual handwheel	Manual handwheel

"A test header is only as honest as its slowest check. Size the glass from stress, keep the meter run straight, and close the valve slowly. Then the number on the screen and the picture in the window say the same thing."

— Guo Wei, Chief Metallurgical & Critical Infrastructure Piping Specialist, LEDE

How Slowly Must a Test Valve Close, and What Do Rotork Controls Do About It?

Slowly enough that the water column decelerates gently. The worst case is sudden closure in less than 2L ÷ a, where L is the pipe length to the next open point and a is the wave speed. For a steel DN150 line with a 4.5 mm wall, a is near 1,250 m/s. With L = 100 m, 2L ÷ a is 0.16 s. A closure faster than that gives the Joukowsky rise, Δp = ρ × a × v = 1,000 × 1,250 × 5.08 ≈ 6.4 MPa. That is far above a window rated near 2 MPa.

Slow the closure and the rise falls fast. For a closure time T well above 2L ÷ a, the pressure rise is about ρ × L × v ÷ T. With ρLv = 508,000 Pa·s, T = 2 s gives about 250 kPa and T = 20 s gives about 25 kPa. A smooth S-curve has a peak deceleration near twice the average, so we plan for about 50 kPa at 20 s. A butterfly valve cuts most of its flow in the last third of travel, so the slow part of the curve belongs there.

This is the job of an electric actuator. Smart actuators from the major makers, such as Rotork controls, accept a 4–20 mA or Modbus setpoint and follow a programmed travel profile. LEDE does not make the actuator. We specify the profile and the closing time, and the buyer confirms the actuator's ingress rating and the conditions it was tested under. A handwheel can do the same job, but only if the operator counts the turns.

Three-step acceptance checklist for a fire test header:

1. Check the window aperture and glass thickness against the working-stress sum, and ask for the minimum visible flow of the indicator
2. Measure the 10D and 5D straight runs at the turbine flow meter, then follow the three-pass alternating bolt sequence and check pad closure on both sides
3. Time the valve closing against ρLv ÷ T, then pressure test the header and compare the meter reading with the sight glass response

ledefittings.com, LEDE grooved pipe fittings and valves, supplies XGQT1 rigid couplings, sight flow indicators and grooved fittings for fire protection pipework. All figures here are design targets or worked examples unless a test report says otherwise; documentation is available on request.