Stable D90 in continuous ultrafine grinding + air classification for plant protein dry fractionation relies on closed-loop process control, stable raw material quality, steady equipment condition, and systematic in-line monitoring.
Stabilise raw material properties first
Raw material variation is the largest source of PSD drift.
Pre-condition feedstock to a fixed moisture window (8–11% for peas/chickpeas) with homogenisation bins; blend batches to reduce variance in bean maturity, variety and hardness.
Remove foreign material and broken seeds. Hard impurities cause erratic particle fracture and mill wear.
Keep feed temperature consistent; cold or warm beans alter brittleness and fracture behaviour.
Constant and uniform feed rate
Fluctuating feed directly shifts residence time and energy input.
Use loss-in-weight (LIW) feeders instead of volumetric feeders for high precision.
Install mass-flow hoppers with anti-bridging features (vibration or air assist) to avoid surging, ratholing or intermittent feeding.
Lock the set feed rate once the process window is validated; avoid manual random adjustments.
Tightly control the core classifier parameters (wheel speed & air flow)
D90 is highly sensitive to the ratio between classifier rotor speed and process air volume.
Maintain classifier wheel speed with variable-frequency drive (VFD) closed-loop speed control.
Stabilise total system airflow: use variable-speed fans and air pressure sensors to compensate for filter clogging or air leakage.
Keep the wheel speed / airflow ratio constant. If one parameter is changed, adjust the other proportionally, rather than modifying only one variable.
Avoid gradual drift from air filter fouling; schedule regular filter pressure monitoring and replacement.
In-line particle size monitoring with PLC closed-loop feedback
Deploy real-time PSD sensors at the fines outlet to continuously track D90 and D50.
Link measurement signals to the control system:
If D90 rises (particles too coarse): automatically increase classifier wheel speed or slightly reduce airflow.
If D90 falls below target (over-fine): reduce wheel speed or raise airflow.
Apply slow, small incremental adjustments (not large jumps) to prevent oscillation in PSD.
Monitor mill health and wear
Rotor, pins, impact liners and classifier blades gradually wear over continuous operation, which changes grinding intensity and cut point.
Track motor current, vibration and chamber temperature as early warning indicators.
Schedule periodic inspection of wear parts. Worn components reduce impact efficiency and cause D90 to drift coarser over time.
Maintain proper mill cooling to avoid heat-induced protein agglomeration that falsely shifts measured PSD.
Manage recirculated middling stream
The recycled mixed protein–starch fraction returning to the mill affects feed composition and particle loading.
Keep recycle flow rate and PSD stable; avoid sudden spikes of coarse material re-entering the grinding zone.
Monitor temperature of recycled middlings; hot recycle material raises mill chamber temperature and causes agglomeration.
Standardise sampling and quality verification
Define fixed sampling frequency and sampling point. Do not rely only on in-line sensors; perform off-line PSD validation with laser diffraction to calibrate the online instrument regularly.
Record D90, moisture, powder temperature, feed rate, wheel speed and airflow in a production log for trend analysis.
Aggressive closed-loop correction can create oscillating PSD. Slow, restrained parameter tuning gives more stable D90, but requires a wider quality tolerance band. Also, maintaining fixed D90 may slightly vary protein purity or yield when raw material properties shift.
