Case Study: Cutting Spray Dryer Wall Buildup with Tungsten Carbide Atomizer Inserts

A spray dryer building 9,000 tonnes of powder a year was losing about 260 hours of capacity to tower cleanouts. Two crew worked each cleanout for 4 to 6 hours, every week, for 52 weeks. The scale crust came off the wall in chunks and went to waste, and nobody had timed how long the practice had been running.

The root cause sat in a part worth a few hundred euros. Wear had opened the orifice inserts from a design 1.4 mm to as much as 1.6 mm. That is a 15 percent enlargement in diameter and 32 percent in flow area. The spray coarsened, the coarse drops landed wet on the wall, and the crew scraped.

The plant fought the symptom for two years before measuring the cause. Insert changes were reactive, tower pressure dropped between changes, and no log tied a shift’s supply pressure to a shift’s wall condition. This case is about turning spray dryer nozzle wear from a mystery into a monitored dimension with a published arithmetic chain.

This account describes a whey and plant protein powder plant of that class, not a named customer, and every figure sits inside published industry ranges. The feed band, the air temperatures, the droplet targets and the wear rates are reference data drawn from open spray drying practice. No named brand, no supplier price and no proprietary recipe appears anywhere below.

Case Snapshot

The table below sets the reference configuration for the rest of this account.

Item	Reference configuration
Industry	Whey and plant protein powders, mixed campaign duty
Tower	Co-current spray dryer, 6 m diameter, 26 m straight side, top discharge
Output rate	1,500 kg/h of powder at 50 to 55 percent feed solids
Operating calendar	7,200 h available, 6,000 h of net drying time per year
Atomizer set	Three pressure swirl positions, one replaceable core body per position
Supply pressure	160 to 180 bar at the nozzle manifold
Air condition	165 to 180 °C inlet, 85 to 90 °C outlet
Residence time	15 to 30 s for the droplet to fall through the drying zone
Droplet target	Dv50 of 80 to 95 µm, giving particles of 120 to 160 µm
Insert hardware	Original one-piece hardened inserts, 1.4 and 1.6 mm, replaced by tungsten carbide multi-hole
Failure mode	Orifice open by 5 to 15 percent after about 120 h of service
Baseline pain	One wall cleanout per week, 4 to 6 h each, plus one insert set per 120 h

Three positions ran together through the campaign, and the arithmetic below carries that count. The 6,000 net hours matter because cleanouts, ramp-ups and off-spec production consume the calendar before any nozzle wears.

The Challenge

The wear mechanism

Whey concentrate carries calcium phosphate and citrate salts in solution and in fine suspension. During atomisation the feed crosses the orifice throat at 80 to 120 m/s. The salt crystals then behave as a fine abrasive jet cutting the throat from the inside. The same mechanism appears with plant isolates carrying mineral grit from milling. A polished stainless or a soft cemented composite throat loses metal at a measurable rate per cubic metre of pumped feed.

The wear was not uniform. It opened the throat as a bell mouth at the outlet edge rather than as a clean larger circle. The flow coefficient therefore rose faster than the geometric area suggested. A 5 to 15 percent diameter gain was routine inside one shift rotation. The plant never saw it, because no gauge sat across the throat.

The droplet consequence

Median droplet size on a pressure swirl nozzle falls with supply pressure and grows with the orifice. The published form ties flow to orifice area and the square root of pressure difference. Median droplet size follows roughly the negative 0.4 power of pressure at fixed geometry and feed. Wear therefore attacks the spray on both axes at once.

A 10 percent enlargement in diameter is a 21 percent enlargement in area, because 1.1 squared equals 1.21. If the plant keeps supply pressure at 180 bar, each nozzle pumps 21 percent more feed. The total evaporation load jumps by about the same share and outlet air temperature falls. If the plant holds the design feed instead, manifold pressure must fall to 180 divided by 1.21 squared, or about 123 bar. That pressure loss alone coarsens the median by about 16 percent, because the size term runs on the negative 0.4 power of pressure.

The square law then converts a coarser median into lost drying time. A 16 percent coarser median needs about 35 percent longer drying, because 1.16 squared is 1.35. The Engineering section works both branches of the 15 percent case: 25 percent coarser spray, or 32 percent more feed. With a residence time of 15 to 30 s, the coarse tail needs more tower than the tower has.

The sticky window

A wet particle stays cool while free moisture evaporates from its surface, sitting near the wet bulb air temperature of roughly 35 to 45 °C. It then climbs toward the outlet air once the moisture content falls under the critical figure of about 25 to 30 percent. Lactose and the amorphous sugars in whey pass a sticky state between roughly 60 and 80 °C. A half-dried particle inside that band still carries enough surface moisture to weld to the wall.

Coarse drops arrive at the wall region still inside that window. They weld to the hot wall on impact and build a deposit that grows as long as the spray stays coarse. The deposit insulates and the local wall temperature drifts. The crust finally sloughs in pieces of 20 mm to a few hundred grams. The cooler then carries the pieces into the product.

The operational spiral

Operators responded the way crews always respond. They trimmed supply pressure to hold outlet temperature, which coarsened the spray further and sped the deposit. They raised feed solids to shorten drying time, which thickened the stick layer. They added a daily water rinse to the wall lance. It knocked 3 to 4 percent of capacity out of the baghouse balance without removing the crust.

Nobody owned the insert as a dimension. Change happened on failure, not on measurement. Pressure logs existed, but none paired a shift’s pressure with the nozzle serial behind it. Spray dryer nozzle wear therefore stayed invisible between collapses.

The Solution

The plant rebuilt the atomisation duty around four moves, and the order mattered.

Hard inserts, matched geometry

The one-piece originals went to tungsten carbide. The grade sits at 1,300 to 1,800 Vickers against 200 to 400 for the soft phases it replaced. Carbide buys abrasive resistance, not toughness, so the design kept the insert as a small replaceable core inside a tougher body. Multi-hole cored inserts spread the same total flow over several smaller throats, each running at a lower throat load and a shorter cutting length.

A swirl element re-match came with the material change. Wider swirl chambers and shorter cores were sized for the multi-hole face. The hollow cone angle and the droplet spectrum then landed back inside the 80 to 95 µm target. A harder material in the wrong geometry reproduces the old problem in a new body. That pairing duty is the whole point of a swirl type spray dry nozzle redesign.

Orifice and pressure pairing

Each position got a designated pair: an orifice size and a supply pressure setpoint written on the same card. A 1.4 mm core runs with 180 bar. The 1.6 mm core runs with a lower setpoint that holds the same design flow and droplet band.

Flow scales with orifice area and the square root of pressure, so holding flow needs pressure cut by the square of the area gain. Going from 1.4 to 1.6 mm raises area by 31 percent, and 180 divided by 1.31 squared lands near 105 bar. The plant chose the two points on the map and stopped letting operators interpolate between them.

Wear monitoring routine

The routine has two cadences. Every shift, the crew records manifold supply pressure with the totaliser feed rate. The pair is plotted as a specific pressure, meaning pressure per unit flow through each position. A rising flow at a stable pressure means a throat has opened, and so does a falling pressure at a stable flow. A 5 percent deviation on that ratio triggers an insert pull for pin gauge verification.

Every month, a go and no-go pin gauge set checks every installed orifice and every spare. The comparison runs against the design dimension with a 0.02 mm tolerance stack. Pulls are logged with the pump-hours behind them. The wear rate in micrometres per thousand cubic metres of feed then becomes a number rather than a rumour.

Change on hours, not on failure

Insert life moved to an hour counter tied to the measured wear rate rather than to the visible wall condition. The change interval also became the cleanout trigger. Tower cleaning moved from an emergency scrape to a planned 3-hour entry every 4 to 6 weeks.

The spray drying nozzles guide covers the atomisation physics behind the pairing table. The air atomizing nozzle page carries the alternative route when a duty suits two-fluid atomisation better.

The Engineering

The droplet to drying-time arithmetic

The square law carries the whole case, so it deserves the explicit numbers. A droplet of 100 µm has a surface area to volume ratio of 60,000 per metre. A droplet of 50 µm has 120,000 per metre, twice the area per unit volume. Evaporation removes mass through that area, while the volume sets the mass to remove. The published result is a drying time proportional to the square of diameter.

Droplet diameter	Area per volume	Relative drying time	Drying time if 50 µm dries in 10 s
50 µm	120,000 /m	1.0	10 s
80 µm	75,000 /m	2.6	26 s
100 µm	60,000 /m	4.0	40 s
110 µm	54,500 /m	4.8	48 s
120 µm	50,000 /m	5.8	58 s

A tower with 25 s of residence dries a 50 µm median comfortably and a 100 µm median not at all. That single row explains why a 10 to 15 percent enlargement of an orifice shows up as wall crust rather than as a moisture alarm. The coarse tail survives the fall and pays for the wear on the wall.

The orifice flow and pressure chain

The sizing relations are stated together here so the arithmetic can be checked in one pass. Flow through a pressure orifice scales with area and the square root of pressure. Median droplet size scales with the negative 0.4 power of pressure, and with flow to the quarter power at fixed geometry. A worn throat therefore moves the spray on both branches below, and the plant always sits on one branch or the other.

Quantity	Relation	Worn by 10 percent	Worn by 15 percent
Orifice area	proportional to d squared	1.21 times	1.32 times
Flow at held pressure	proportional to area	1.21 times	1.32 times
Pressure at held flow	proportional to 1 over area squared	123 bar	103 bar
Median droplet, held flow	proportional to p to the minus 0.4	1.16 times	1.25 times
Median droplet, held pressure	proportional to Q to the 0.25	1.05 times	1.07 times
Drying time, held flow	proportional to droplet squared	1.35 times	1.56 times
Air load at held pressure	feed plus area gain	21 percent more water	32 percent more water

The held-flow rows all come from one substitution. Area goes as d squared, so 1.32 for a 15 percent wear. Holding the feed then divides pressure by 1.32 squared, which gives 103 bar. Pressure to the minus 0.4 power returns a median 1.25 times as coarse, and the square law turns that into 1.56 times the drying time.

Start with the held-flow column. A spray dryer nozzle with a throat open by 15 percent and an operator holding design feed runs at 103 bar. The median goes 25 percent coarser and needs 56 percent more time to reach the same moisture. In a tower built for 15 to 30 s, that is not a quality drift but a physical impossibility. The wall is where the physics collects the debt.

The held-pressure column fails differently. The spray barely coarsens, but the tower must evaporate a third more water on the same air. Outlet temperature then falls and every particle arrives wetter at the wall. Both branches end in the same crust, which is why pairing the orifice to a pressure setpoint matters more than either number alone.

Why carbide and multi-hole together

Abrasive wear on a throat is commonly modelled as a volume loss proportional to the abrasive work crossing the surface. Two smaller throats carry the same total flow at a lower cutting length each, and the total wetted surface rises while the perimeter velocity falls. The carbide grade then raises the hardness term in the denominator of the same proportionality.

The plant measured insert mass loss on the bench scale before committing. The originals lost 8 to 15 mg per 1,000 m³ of pumped feed. The carbide cores lost 1 to 3 mg on the same slurry. A factor of five to eight on the wear term sits beside a five times longer interval before any geometry drift. Both terms point at the 400 to 600 h band without heroic claims.

There is a trade against brittleness. A carbide core chipped by a stone or a dry-run surge fails suddenly rather than wearing gently. The plant added a 200 mesh suction strainer ahead of the high pressure pump and a low-flow interlock on the manifold. Those two guards are part of the cost of the material change and are reported as such.

The cost per tonne

The insert ledger uses round reference figures: 300 EUR per original insert, 350 EUR per carbide core. That sits inside the hundreds-of-euros band typical for a small precision core. No vendor price is implied, and the arithmetic is the deliverable.

Line	Before	After
Net drying hours per year	6,000	6,000
Insert life in hours	120	500 (mid of 400 to 600)
Insert changes per position per year	50	12
Inserts purchased, three positions	150	36
Unit cost per insert, EUR	300	350
Insert spend per year, EUR	45,000	12,600
Powder produced per year, tonnes	9,000	9,000
Insert cost per tonne, EUR	5.00	1.40

The gap of 3.60 EUR per tonne looks modest until the cleaning ledger sits beside it. Weekly cleanouts consumed about 260 hours, and twelve planned entries spend 60. The calendar returns 200 hours to production, worth 300 tonnes of powder at 1,500 kg/h. Cleaning labour falls from 52 events to 12, at four crew and 5 h each, about 800 crew-hours a year. CIP water, detergent and rinse-to-waste volume fall in the same order of magnitude, so the recovered hours dominate the insert bill.

The Results

The figures below compare twelve months before the rebuild with twelve months after. Feed, tower and campaign mix stayed the same.

Metric	Before	After	Basis note
Insert life between pulls	about 120 h	400 to 600 h	Pump-hours to a 5 percent specific pressure deviation
Orifice enlargement at pull	5 to 15 percent	2 to 4 percent	Pin gauge against the design dimension
Wall cleanout interval	every week	every 4 to 6 weeks	Planned entry, not emergency scrape
Annual cleanout downtime	about 260 h	about 60 h	5 h average per event
Weighted capacity loss	4 to 5 percent	1 to 1.5 percent	Downtime plus off-spec during ramp
Median droplet Dv50	85 to 112 µm	82 to 92 µm	Malvern-style diffraction on the design cone
Product moisture band	3.0 to 5.5 percent	3.2 to 4.2 percent	Narrower around a 3.8 percent setpoint
Fines off the 150 µm sieve	7.5 percent	4.5 percent	Baghouse return, 3 point fall
Insert cost per tonne	5.00 EUR	1.40 EUR	Ledger above, reference unit costs
Wall deposit in product	visible chunks	trace	Visual per shift, 20 buckets per week to 4

Insert life and cleanout interval moved together, which is the expected pattern if the wear theory is right. Moisture control tightened because the coarse tail stopped arriving wet, and the plant could raise the setpoint discipline rather than chase a drifting process. The fines fall is modest but consistent, since fewer crust pieces recycle through the air system and the cone stayed narrower.

Two numbers moved the wrong way. Insert unit cost rose by about 17 percent, and the guard hardware added a strainer element and an interlock to maintain. Both are small, both were budgeted, and neither survives contact with 200 recovered hours.

Why This Matters

The transferable principle is that an orifice is a flow element with a tolerance, not a consumable with a life. Any pressure atomiser pumps against a known area, so every micron of throat growth is a silent recalibration of the spray. Treat the dimension as process control data, log the pressure-to-flow ratio per shift, and the failure becomes visible weeks before the wall shows it.

The second principle is that material grade without a geometry re-match buys nothing permanent. A harder core in a swirl chamber sized for a different flow keeps the same wrong droplet spectrum. The pairing of orifice, swirl element and pressure setpoint is the product, and the insert is only one part of it. That pairing is where a proper redesign earns its keep, and where a simple material swap quietly fails.

The third is that droplet physics converts a small mechanical drift into a large thermal debt. Square-law drying time means a 15 percent worn throat can cost 56 percent more drying on a held feed. The tower cannot negotiate, and plants that measure wall deposits as a cleaning problem and orifices as a spares problem keep paying both bills.

If your tower is fighting deposits, moisture drift or a spares burn you cannot explain, start a nozzle wear audit with three numbers. Take the design orifice dimension, the shift pressure-to-flow ratio, and the pump-hours behind each pull. Send those to the BoreJet application team and the atomisation review covers the rest. The centrifugal versus pressure nozzle particle size guide sets out the wider droplet sizing arithmetic behind that conversation.