Pilot-scale NCM cathode material processing bridges lab R&D and full commercial manufacturing. The primary goal is to stabilise PSD (particle size distribution), including D50, D90, D99 and span, while preserving NCM crystal integrity and avoiding metal contamination. For pilot jet mill and air classifier circuits referenced from ncm-mill.com, particle size control relies on feed pre-treatment, jet mill process parameters, closed-loop classification, atmosphere control, inline monitoring and structured test runs.
1. Stabilise feed material before entering the jet mill
Unstable feed PSD, moisture and agglomerates are the most common root causes of inconsistent pilot particle size results.
- Pre-screen the incoming calcined NCM powder to remove oversized hard agglomerates and foreign lumps. Feed D50 and maximum particle size must be kept within a fixed window before milling.
- Control moisture strictly. NCM readily absorbs moisture; humid powder forms sticky agglomerates that resist breakup and distort PSD. Maintain dew point below -40°C in feed hopper and feeding section.
- Use continuous, low-pulsation screw feeding instead of batch dumping. Fluctuating feed mass flow changes particle concentration inside the grinding chamber and shifts cut point.
2. Tune core jet mill operating parameters
For pilot fluidized bed jet mill or spiral jet mill systems, three main knobs adjust fineness: grinding gas pressure, feed rate and classifier wheel speed.
- Grinding nitrogen pressure: Higher gas pressure increases particle-particle collision energy and produces finer powder. Lower pressure gives coarser particles. Pilot testing should define a stable pressure setpoint with narrow tolerance (±0.05 MPa). Do not over-pressure high-nickel NCM; excessive collision energy may damage crystal structure and degrade electrochemical performance.
- Feed rate: Reducing feed rate extends particle residence time and delivers finer PSD. Overfeeding causes particle crowding, reduces collision efficiency and shifts D50 upward. In pilot trials, run staged feed rate sweeps to map PSD vs throughput.
- Classifier wheel speed (most sensitive for top-cut): Increasing the VFD speed of the ceramic classifier wheel raises centrifugal force. Only smaller particles can pass through the wheel as finished product; coarser particles return to the grinding loop. This directly controls D99 and prevents “killer coarse particles” that can pierce battery separators. Higher classifier speed = finer cut point; lower speed = coarser product.
All material-contact surfaces in pilot circuit use full ceramic lining to eliminate metal wear debris and avoid introducing impurities while maintaining consistent powder flow.
3. Optimise closed-loop classification and circulating load
Pilot NCM lines use closed-circuit grinding: oversized particles are recirculated back to the jet mill instead of being discarded.
- Balance circulating load. Too high circulation increases fine powder buildup and broadens PSD; too low circulation wastes capacity and leaves coarse tails.
- Minimise dead zones in piping, elbows and cyclone bodies. Residual trapped fines will intermittently release and cause PSD drift between pilot batches.
- Install secondary inline sieve after collection as a safety guard to remove rare oversized agglomerates for battery-grade NCM.
4. Control gas environment and prevent re-agglomeration
NCM fines tend to re-agglomerate after milling, which falsely shifts laser PSD test results.
- Use dry inert nitrogen as grinding gas, not compressed air, to avoid oxidation and moisture-induced caking, especially for high-nickel NCM.
- Keep gas-to-solid ratio stable. Too much gas dilutes particles and reduces collision frequency; insufficient gas causes poor powder transport and agglomeration.
- Maintain low temperature during pilot runs. Excessive heat can alter NCM surface properties and increase fine particle cohesion.
5. Implement inline PSD monitoring and closed-loop feedback
Pilot scale requires real-time feedback rather than only off-line lab testing.
- Deploy online laser particle size analyser at the finished powder outlet. Continuously track D50, D90 and span.
- Set up PLC interlock logic: if D50 drifts outside target window, automatically adjust classifier wheel speed or fine-tune feed rate within safe limits.
- Keep consistent sampling protocol. Grab samples at fixed time intervals and sample points; avoid sampling only at start or end of batch, which will misrepresent average PSD.
- Maintain record of all parameters: feed PSD, moisture, gas pressure, classifier frequency, feed rate, nitrogen dew point and PSD results for traceability.
6. Use structured pilot test matrix and minimise cross-contamination
Pilot work is designed for parameter mapping, but cross-contamination will invalidate PSD data.
- Run single-variable tests: change only one parameter at a time (e.g. classifier speed) while holding all others constant. This isolates the effect on PSD.
- Clean the full circuit thoroughly before switching NCM grades. Residual powder from previous pilot batches will mix in and distort particle size and impurity data.
- When target PSD is achieved, run a confirmation stability batch at fixed parameters to verify repeatability before scaling up.
7. Common troubleshooting for pilot PSD drift
- D50 suddenly becomes coarser: check feed agglomeration, drop in grinding pressure, classifier wheel fouling or overfeeding.
- Excess fines and wide span: classifier speed too high, gas pressure too high, or powder re-agglomeration from moisture.
- Unstable D99: classifier wheel wear or powder buildup on rotor blades, which changes effective cut geometry.
Controlling particle size in pilot NCM production is a multi-variable optimisation. The primary control hierarchy is: stabilise feed quality → tune classifier wheel speed for top-cut → adjust jet mill nitrogen pressure and feed rate → stabilise closed-loop circulation and inert atmosphere → apply real-time PSD monitoring and parameter logging. Pilot tests should preserve NCM crystal structure while locking D50, D99 and span within specification, and generate reproducible operating data that can be transferred to full-scale production lines.
