Design the line backwards from saleable kernel, never forwards from raw-nut intake. Reversing that order is the commonest cause of an unbalanced plant, and it is expensive to correct once concrete is poured.
A note on method, and on where these numbers come from
The published record for cashew plant design is thin. There are almost no floor areas, no square-metres-per-tonne ratios, no ceiling heights, no drainage specifications and no generator sizing guidance anywhere in the open literature. ComCashew’s own equipment guidebooks describe the gap by omission — a factory needs “more space, better and more reliable electrical systems, better ventilation and better design for a logical flow”, and then no numbers follow for any of it.
So this chapter states its basis openly. The stage capacity ranges below are drawn from published equipment literature. The cycle times that govern the batch stages — cooling after steaming, drying, humidification — are not published anywhere. ComCashew defers them to the machine builder: processors “should follow manufacturer’s specification on steaming duration, temperature and air pressure”, and give no drying duration, no humidifier time and no intermediate moisture target at all.
The cycle times used here therefore come from our own operating experience across lines we have specified, installed and commissioned. They are a starting point, not a specification, and they move with your nuts, your intake moisture and your vessel. A drying cycle that turns out to be sixteen hours rather than thirteen moves every batch stage below it and every buffer beside it. Measure your own the moment you can, and read what follows as the method rather than the answer.
Working backwards from annual tonnage
Design in this order:
- Target saleable kernel tonnage — from your market work, not your machine catalogue.
- ÷ production KOR (22–24%) = annual RCN requirement.
- ÷ operating days = daily RCN throughput. This is where the season assumption enters the engineering, and it is why two plants with the same nameplate can need different equipment.
- ÷ operating hours per day = hourly RCN feed rate at stage one.
- Apply the mass transformation. After shelling you are no longer moving RCN, you are moving kernel at roughly a quarter of the mass. Every downstream stage must be sized on that.
- Apply a rework multiplier at peeling and grading, because material goes round again.
Worked example — a 10 t/day plant. 3,000 t RCN a year over 300 days is 10 t/day. Over 16 operating hours (two shifts) that is 625 kg/h of RCN at intake. At 23% production KOR, kernel output is 690 t/year — 2.3 t/day, or 144 kg/h of kernel through drying, humidification, peeling, grading and packing.
The kernel side of the plant handles less than a quarter of the mass of the raw side. Sizing peeling on 625 kg/h is the single most expensive arithmetic error available to you.
Published stage capacities, side by side
Set the stages against each other and the bottleneck is obvious before you buy anything.
| Stage | Published capacity range | Notes |
|---|---|---|
| RCN calibration, continuous | 300–500 kg/h (3-screen) · 800–1,000 (6-screen) · 1,500–1,700 (9-screen) · 1,800–2,000 (12-screen) | Batch lines quoted at 5 t (35–45 min/batch) and 15–35 t (30–35 min/batch) are complete sorting lines, not single machines |
| Steaming | 250–1,000 kg per batch; 2–12 t/day per cooker | Batch process |
| Shelling / cutting | 140–1,200 kg/h depending on configuration | Head-machine and knife-machine families |
| Drying (borma) | 0.5–4.5 t per batch, 13–14 h/batch | Batch process |
| Humidification | 0.5–4.5 t per batch, 3–7 h | Batch process |
| Peeling | 120–480 kg/h of kernel | 2–3 passes typical; 20–40% manual finishing |
| Mechanical size grading | 80–100 kg/h | The structural bottleneck |
| Kernel grading (classify machines) | 150–180 (12-axis) · 210–270 (18-axis) · 280–360 (24-axis) · 420–540 kg/h (36-axis) | |
| Broken-kernel sorting | 150–200 kg/h | |
| Colour sorting | 500–1,000 kg/h (chute) · 1,000–3,000 kg/h (belt) | |
| Kernel sorting line before packing | 1,000–1,500 kg/h | A sorting line, not a packer |
| Final conditioning | No published capacity exists | A short low-temperature dry-back before sealing. Size on your own kernel throughput, not a catalogue rate |
| Vacuum packing | 50–70 packs/hour |
Every capacity above is quoted on a Size A input basis, and capacity varies with raw-nut size. A sheller’s real limit is nuts per minute rather than kilograms per hour, and nut size alone can swing rated capacity by around a fifth. Require every rate in every quotation to state the nut count and calibre band it was measured on, and reject the ones that do not.
Two conclusions fall straight out of the table.
Grading and peeling are the structural bottlenecks. A mechanical size grader at 80–100 kg/h sits in the same line as a vacuum packing station rated above a tonne an hour. Buy on nameplate and you build a plant whose most expensive stages idle behind its cheapest.
Calibration and packing are almost always oversized. They are cheap per kilogram of capacity, so buyers over-buy them, and the surplus does nothing but occupy floor.
One precaution worth taking seriously: ComCashew is explicit that mechanised whole-kernel grading remains unsolved, recording that processors “have consistently experienced problems with poor grading and breakage of kernels. More research is required.” Do not design a plant that depends on a grading machine hitting its published rate at your required quality. Put a hand-sorting bench alongside it, because you will use it — machines separate five to eight grades and the trade recognises more than twenty-four.
Set a plant size and watch which stage needs most units. It is not the expensive one.
One shift is 8, two shifts 16. The same annual tonnage over fewer hours needs bigger machines.
| Stage | Handles | Design rate | Top published rate, one unit | Units needed |
|---|---|---|---|---|
| Calibration | RCN | 750 kg/h | 2,000 kg/h | 1 |
| Steaming | RCN | 625 kg/h | 1,000 kg/h | 1 |
| Shelling / cutting | RCN | 719 kg/h | 1,200 kg/h | 1 |
| Peeling | Kernel | 359 kg/h | 480 kg/h | 1 |
| Mechanical size grading | Kernel | 187 kg/h | 100 kg/h | 2Binding stage |
| Colour sorting | Kernel | 187 kg/h | 3,000 kg/h | 1 |
| Packing | Kernel | 158 kg/h | 1,000 kg/h | 1 |
Design rates per stage
Applied to the 10 t/day worked plant, with every multiplier stated.
| Stage | Basis | Design rate | Multiplier applied |
|---|---|---|---|
| Calibration | RCN intake | 625 kg/h | ×1.2 for screen changeover between size fractions → 750 kg/h |
| Steaming | RCN | 625 kg/h average | Batch: 4 × 1,000 kg batches per 8 h shift, sized on cycle not flow |
| Shelling | RCN | 625 kg/h | ×1.15 for uncut recycling (under 5% uncut, re-fed) → 720 kg/h |
| Drying | Wet kernel | ~160 kg/h equivalent | Batch: 13–14 h cycle governs; size on daily volume ÷ batches per day |
| Humidification | Dried kernel | ~150 kg/h equivalent | Batch: 3–7 h cycle |
| Peeling | Kernel | 144 kg/h | ×2.5 for multiple passes, manual finishing and a first-season first-pass rate near 70% → 360 kg/h installed |
| Grading and colour sorting | Kernel | 144 kg/h | ×1.3 for rework and re-sorting → 190 kg/h |
| Final conditioning | Graded kernel | 144 kg/h | ×1.1 → 160 kg/h |
| Packing | Conditioned kernel | 144 kg/h | ×1.1 → 160 kg/h |
The peeling multiplier is the one that matters. The specification figure for first-pass peeling is 80–90%, and that is what you write into a machinery contract and test at acceptance. What factories actually report is nearer 70–80%, with 4–8% breakage and 20–40% of kernels going to manual finishing whatever the machine does. Specify against the higher figure; size the floor against the lower one. A peeling section sized at nameplate will be your bottleneck within a week of commissioning.
Final conditioning earns its own row because the moisture ladder does not otherwise close. Drying takes the kernel to 3–4%, humidification puts one to two points back so it can be peeled without shattering, and the packing specification is 3–5%. Moisture therefore has to come down again between peeling and packing. Something must do that, it handles graded product inside the finished-goods zone, and it is not the vacuum packer.
Batch against flow
Half the stages are continuous and half are batch, and the plant runs at the speed of the batch stages. Steaming, drying and humidification all hold material for hours; calibration, shelling, peeling and grading all move it in minutes.
That mismatch is not a fault to be engineered away — it is the shape of the process. What it demands is buffer volume between every batch stage and the continuous stage that follows it, sized on the batch cycle rather than the hourly rate. A layout that runs conveyors straight from a batch dryer into a continuous peeler will stall both.