Air classification performance for plant protein dry fractionation is heavily dependent on feed particle size distribution (PSD). In single-pass or multi-stage pulse protein separation, incomplete liberation, particle agglomeration and drifting milling output directly degrade enrichment factor, protein recovery and final yield. Real‑time particle size monitoring delivers continuous PSD data across the process stream, enabling closed-loop control and stabilizing protein fractionation performance.
1. Pre-classification feed quality control: guarantee adequate protein body liberation
The prerequisite for high protein enrichment is sufficient milling to liberate protein bodies from starch granules. Offline lab testing involves sampling, dilution and laser diffraction analysis, which creates significant time lag.
Real-time PSD sensors installed after the mill continuously track D50, D90 and particle size span of the incoming flour.
- If particles are too coarse: insufficient cell rupture, protein remains encapsulated inside starch agglomerates → enrichment factor drops. The system can automatically adjust mill rotational speed or feed rate to increase grinding intensity.
- If over-milling occurs: excessive starch fines are generated. Ultra-fine starch contaminates the protein-rich fine fraction, lowering protein purity and increasing powder agglomeration risk. The controller reduces milling energy immediately.
By locking the feed PSD window before material enters the classifier, real-time monitoring prevents bad feedstock from entering the classification stage, avoiding wasted throughput and lost protein yield.
2. Dynamic classifier rotor and airflow tuning for stable cut point
The air classifier’s cut point (d50) is highly sensitive to fluctuations in feed PSD, powder moisture and feed load. Without real-time particle feedback, operators rely on periodic offline samples, so drift is only detected after batches of off-spec product have already been produced.
On-stream particle measurement at both fine and coarse outlets provides instant particle distribution data:
- When coarse particles leak into the fine fraction: the cut point is too large. The controller raises classifier wheel speed or reduces primary airflow to tighten separation.
- When excessive fine protein particles are lost into the coarse stream: the cut point is too fine. The system lowers rotor speed or increases airflow to recover more protein fines.
This closed-loop control maintains a stable Tromp curve and sharpness index, preserving the target protein enrichment factor and minimizing protein loss into the coarse starch fraction, lifting overall protein recovery yield.
3. Early detection of particle agglomeration and process fouling
Powder agglomeration is a major hidden loss mechanism in dry protein fractionation. Fine protein particles stick together or attach onto coarse starch grains, causing liberated protein bodies to report incorrectly to the coarse discharge and reduce protein yield. Agglomeration can be triggered by rising moisture, frictional heat or filter fouling.
Real-time PSD monitoring detects agglomeration instantly by picking up unexpected shifts in the coarse-side particle population before protein content drops in lab tests.
Operators can then adjust air temperature, add dispersion air or clean rotor/fouled components at an early stage, rather than waiting for chemical protein test results to identify the problem. This avoids sustained low-efficiency operation and large yield losses.
4. Reduce transition waste during raw material batch changes
Pulse feedstock varies by harvest lot, bean variety and moisture, which changes milling behaviour and particle liberation. When switching batches, offline testing requires running large quantities of material to stabilize and sample, generating substantial off-spec waste.
Continuous real-time particle monitoring accelerates process setpoint tuning. The control system rapidly adjusts mill and classifier parameters as soon as PSD deviates, shortening transition time between raw material lots. Less out-of-spec intermediate material is produced, improving usable product yield and reducing downtime.
5. Enable data logging and process benchmarking for long-term yield optimisation
Real-time PSD data is logged alongside classifier operating parameters (rotor frequency, feed rate, airflow) and protein assay results. This creates a dataset linking particle characteristics to protein enrichment and recovery.
Process engineers can identify optimal operating windows, quantify how PSD variations affect protein shift and build predictive models. The database supports continuous process optimisation: refining target particle size setpoints to maximise the balance between fine mass yield and protein enrichment over long production runs.
Limitations to consider
Real-time particle sensors cannot directly measure protein content; PSD is an indirect marker. A good particle size profile does not guarantee protein liberation. It must be combined with periodic offline Kjeldahl protein testing to calibrate the correlation between particle size and actual protein separation performance. Dust buildup on sensor windows also requires regular maintenance to preserve measurement accuracy.
Real-time particle size monitoring turns discrete, delayed offline particle testing into continuous, actionable process feedback. It stabilises feed liberation, maintains a consistent classifier cut point, detects agglomeration/fouling early, cuts batch-transition waste, and enables long-term data-driven optimisation. By keeping particle characteristics within the designed operating window, the air classification system retains higher protein recovery yield while maintaining target protein enrichment.
