Particle Characterisation Techniques for Industrial Slurries

Start with the particle problem: size range, concentration and why offline sampling often misleads Compare the main techniques - laser diffraction, imaging and in-situ probes - against real slurry conditions Build a selection checklist and validation steps so the chosen method holds up in production

20 October 2026 8 min read

Summary of this article:

Start with the particle problem: size range, concentration and why offline sampling often misleads.
Compare laser diffraction, imaging and in-situ probes against real slurry conditions.
Build a selection checklist and validation steps so the chosen method holds up in production.

Particle characterisation of an industrial slurry starts with an awkward fact: the slurry has no single true particle size. Laser diffraction, imaging and an in-line probe will give three different medians for the same sample, and all three will be correct, because each method measures a different thing. The useful question is which measurement tracks the equipment you care about, at the concentration your solids see in the process, in a form able to survive scrutiny when something goes wrong.

This guide to slurry particle size analysis follows the logic we use on real multiphase systems: define what you need to measure, respect the sampling, match the method, then validate against pass marks.

Start with the particle problem, not the instrument

Every reported size carries three choices, and the method makes them for you.

The equivalent diameter. Laser diffraction reports the sphere with the same scattering pattern. Imaging reports a projected-area circle or a Feret length. A sieve passes a particle on roughly its second-smallest dimension. FBRM reports a chord length.
The weighting. Imaging counts by number, laser diffraction reports volume and sieving reports mass.
The dispersion state. You measure flocs and agglomerates as they exist in the process, or primary particles after deliberate dispersion.

Weighting alone settles most arguments between instruments. Take 99 particles of 5 µm and one 50 µm agglomerate. By number, the agglomerate is 1% of the population. By volume, the agglomerate holds 91% of the solids. The number median is 5 µm and the volume median sits near 50 µm.

Share of the sample	99 particles at 5 µm	1 agglomerate at 50 µm
By number (imaging)	99%	1%
By volume (laser diffraction)	9%	91%

Next, pick the size your equipment responds to. Thickeners and hydrocyclones respond to settling velocity, so a Stokes diameter or a sieve-based P80 fits. Filtration responds to fines and surface area. Pump and screen wear respond to the coarse tail by mass. A specification or a dispute needs a standard method with a written method statement. Process control needs speed and repeatability and tolerates a known, stable bias.

Why offline sampling so often misleads

Many particle size errors start before the instrument runs. Pierre Gy’s theory of sampling shows sampling error grows with the cube of the top particle size: double the top size and you need about eight times the sample mass. The coarse tail is thin, too. If 2% of a 1 g quartz sample sits above 500 µm, counting alone puts ±9% error on that fraction.

Slurry samples settle fast. Quartz at 50 µm clears the top 100 mm of an unstirred tank in about 45 seconds. Take samples from vertical pipe sections or with a cross-stream cutter, and split them with a rotary riffler as ISO 14488 describes.

Dilution is fine once tested. Dilute in filtered mother liquor, run a 2×, 5× and 10× series, and check D10, D50 and D90 stay within the repeatability limit. Keep samples wet where possible, since drying fuses particles together.

Comparing the main particle characterisation techniques

Each method answers a different question. On elongated or platy solids, expect disagreement, and build a known offset between methods instead of chasing identical numbers.

Laser diffraction

Laser diffraction (ISO 13320:2020) fits scattered light to an optical model and reports volume-weighted equivalent spheres from about 0.1 µm to 3 mm. The standard sets method repeatability limits of 3% CV at D10, 2.5% at D50 and 4% at D90, which makes laser diffraction the default for specifications.

Watch for four failure modes.

Fraunhofer theory holds only above about 50 µm. Below 50 µm, transparent particles produce phantom fines, so use Mie theory with the correct refractive index.
High concentration reads fine. In a HORIBA test on a sulphide ore, D50 fell from 83.5 µm to 74.4 µm as transmission dropped from 89% to 31%.
Needles and plates read broad and coarse. In work by Agimelen and co-workers, a population of identical 100 µm needles, inverted as spheres, produced false modes from 0.8 µm to 140 µm.
Air bubbles show up as a false coarse mode.

Imaging and microscopy

Image analysis (ISO 13322-1 and 13322-2) records 2D projections and reports size and shape for each particle. Use imaging when shape drives behaviour, such as needle crystals blinding a filter cloth.

Dynamic instruments count hundreds of thousands of particles per run, so throughput is rarely the limit. Resolution, focus and overlap are. Guidance drawn from ISO 13322 asks for at least 3 pixels across a particle for size and 9 for shape. Counts decide the tail. Sympatec, citing ISO 14488, puts a volume distribution at more than 1,000,000 particles for 1% error, and an Entegris test produced a convincing but wrong D90 from 5,213 particles.

AI segmentation separates touching particles and bubbles far better than thresholding, and loses accuracy as images get crowded. In one study, Mask R-CNN scored an AP50 of 0.98 at low gas fraction and 0.57 on synthetic images at 5% void fraction. Report precision and recall on your own material before trusting the output.

In-situ probes

In-line probes remove sampling error and add model error. None gives ground truth.

FBRM reports chord lengths, and converting chords to sizes needs an assumed shape. Count rate also rises with flow past the window. Read FBRM as a trend: nucleation onset, breakage, flocculation endpoint.
Ultrasonic attenuation (ISO 20998) measures without dilution, up to 70% solids by volume on Sympatec’s OPUS. The method needs physical property data for both phases and struggles in heavily aerated streams.
In-line laser diffraction still passes high-solids slurries through diluters.

Each probe sees a small region near its window, so calibrate against an offline reference across the operating window. CiDRA reports its CYCLONEtrac cyclone sensor within 5 percentage points of lab results (one standard deviation) across seven plants. A CFD model of the pipe or vessel shows whether a probe position represents the bulk before you install anything.

A selection checklist for your slurry

Size range: does the method cover the tail you care about? Above 1 to 2 mm, sieve.
Concentration: if the method dilutes, have you shown the dilution leaves the distribution unchanged?
Measurand: do the reported diameter and weighting match what your equipment responds to?
Shape: needles, plates and fibres need imaging alongside any size-only method.
Gas: bubbles corrupt diffraction, FBRM and ultrasonic signals. Air in water has a relative refractive index near 0.75 and scatters strongly. Confirm with imaging how many of your “particles” are bubbles.
Timing: control needs an in-line trend, and a specification needs an offline standard method.
Method statement: record the optical model, refractive index, diluent, obscuration and ultrasound time with every result.

Validation steps before you commit

Qualify the instrument with a certified reference material such as glass beads.
Measure one suspension repeatedly, then fresh subsamples. The gap between the two shows how much error comes from sampling.
Spike 1 to 2% by mass of a sieved coarse fraction and confirm the method recovers the spike.
Run a gauge R&R study across operators.
Reconcile against the process, for example by closing a size-by-size mass balance around a hydrocyclone.
Revalidate when feed, chemistry, temperature or flow regime moves outside the tested window.

Frequently asked questions

Why do laser diffraction and image analysis give different results?

Laser diffraction reports a volume-weighted distribution, and imaging counts by number. One 50 µm agglomerate among 99 particles of 5 µm is 1% of the sample by number and 91% by volume.

Is FBRM a particle size measurement?

No. FBRM reports chord lengths, and converting chords to sizes needs an assumed particle shape. Use FBRM for trends such as nucleation, breakage and flocculation.

Does diluting a slurry change the particle size distribution?

Sometimes. Dilute in filtered mother liquor, run a 2×, 5× and 10× series, and check D10, D50 and D90 stay within the repeatability limit.

How many particles does image analysis need to count?

For a volume distribution with 1% error, Sympatec, citing ISO 14488, puts the figure above 1,000,000 particles. Low counts hurt the coarse tail and D90 most.

How do bubbles affect particle size results?

Air in water has a relative refractive index near 0.75 and scatters strongly, so bubbles show up as a false coarse mode in laser diffraction and as extra chords in FBRM. Confirm the gas fraction with imaging before trusting either result.

Where to go next

If your instruments disagree on a slurry, or you need to choose a method before buying one, we run a method-comparison study on your own material. You get the results on a common basis, the source of each offset, a recommended method and a validation plan. Where gas or flow segregation is in play, we add imaging and CFD to show what each sensor sees.

Talk to an engineer at innovexis.co.uk to scope a study on your slurry, or email us at enquiry@innovexis.co.uk.