Two Numbers Have to Exist Before the Question Can Be Answered
A blend is uniform when it meets a stated criterion. That sounds circular until you notice that most plants never state one, and then spend meetings disagreeing about whether a sample "looks right". Two figures remove the argument.
| Figure | What it fixes | Where it should come from |
|---|---|---|
| Mixing index | The maximum variation you will accept between samples. | The product specification, agreed with the customer or set from what the formulation can tolerate. |
| Sample size | The smallest mass that can still be considered representative. | The segregation scale of the blend, which is a function of particle size and density difference, not of the sampler. |
Both have to be fixed before the trial, not after seeing the numbers. A criterion chosen after the data arrives is a description of the data, not a test of the process.
The Mixing Index: Measuring Variation Instead of Averaging It Away
The commonest mixing index is the relative standard deviation of a key component across a set of samples. Take several samples from different positions, measure the concentration of the component you care about in each, and compare the spread of those readings to their mean.
The comparison that matters is against the unmixed feed, not against zero. A feed that already varies widely has more room to improve, and the same absolute spread means something different in the two cases. This is why a mixing index is usually quoted as a fraction of the initial variation rather than as a bare number.
| Measure | What it tells you | What it hides |
|---|---|---|
| Mean concentration | Whether the right amount of the component is present. | Everything. A perfectly dosed blend and a badly segregated one can share a mean. |
| Spread between samples | Whether the composition varies from place to place. | Whether the sampler could see the variation at all. |
| Relative standard deviation | The spread as a share of the mean, so blends of different dose can be compared. | Nothing by itself, but only if the sample mass is stated alongside it. |
| Index against unmixed feed | How much of the available improvement has been achieved. | Whether the remaining variation matters for the product. |
The last row is the one to keep in view. An index can look excellent and still leave variation that matters, if the component is a minor additive whose effect is concentrated. The criterion belongs to the product, and the index is only the tool that measures it.
Sample Size Has to Be Larger Than the Segregation Scale
This is the part most often got wrong, and it fails silently. A sample that is too small reports a variation that is either far too large or far too small, depending on whether it happens to catch the coarse particles or miss them.
The scale that matters is set by the powder, not the sampler. Two powders that differ in particle size or in density will separate over a characteristic distance, and that distance is the scale the sample has to straddle. A sample smaller than it can sit entirely inside one band and report a misleadingly even result in a bed that is stratified.
| What varies | Why it segregates | What it does to the sample size |
|---|---|---|
| Particle size | Vibration and flow move small particles down and coarse particles up. | Wide size distribution means a larger scale, so larger samples and more of them. |
| Particle density | Dense particles sink through a fluidized or rolling bed. | Density difference sets how far apart the bands form. |
| Particle shape | Flakes and needles pack differently and roll differently from spheres. | Irregular shapes bridge and hold voids, which makes the pattern patchier and harder to sample. |
| Moisture and static | Damp or charged fines cling to surfaces and to coarse particles. | Creates short-scale variation that a small sample can easily miss entirely. |
The practical rule that follows: state the sample mass whenever you report a mixing index. An index without a sample mass is not a measurement, because the same bed will produce different numbers at different sample sizes.
The Mixing Curve Is Not Monotonic
Plot the mixing index against time and you get a curve that falls, flattens, and on a free-flowing powder can begin to rise again. The flat region is the operating window. Mixing past it is not extra safety, it is a different process.
| Region of the curve | What is happening in the bed | What to do |
|---|---|---|
| Falling | The ribbon is turning the bed over and the components are being distributed. | Keep going. Sampling here only confirms the trend. |
| Flat | Further turning no longer changes the distribution measurably. | This is the window. Set the time from the middle of the flat region, not its left edge. |
| Rising | Energy is going into the bed as motion and the coarse fraction is migrating to the surface or the edges. | Stop. More time is now degrading the blend. |
The middle of the flat region rather than its left edge, because the left edge is where the improvement is only just becoming unmeasurable, and small variations in feed or filling will put you back on the slope. Setting the time from the middle buys tolerance without paying for the rising region.
Where the Blend Actually Fails Is Downstream
A blender can deliver a uniform bed and still ship an ununiform product. Every movement after the mixer is an opportunity to separate particles again, and the operations that do it are the ordinary ones.
| Operation | What separates | How to reduce it |
|---|---|---|
| Discharging into a drum or bag | The falling stream sorts by size; fines drift, coarse drops straight. | Keep the drop short, use a full-width discharge rather than a single point, and avoid free fall. |
| Pneumatic conveying | Air velocity carries fines further than coarse particles. | Lower velocity where the layout allows, and avoid long runs and sharp bends. |
| Storage in a hopper | Filling builds a cone that sorts by size; the center and the wall differ. | Sample the hopper outlet during discharge rather than the top surface. |
| Vibration in transit | Sustained vibration percolates fines downward. | Treat transport as part of the process, and sample on arrival if the product is sensitive. |
This is the reason the acceptance sample often belongs at the packing end. A sample taken from the mixer tests the state the powder was in a moment before the most segregating steps in the whole line. That is a real measurement, but it is not the measurement the customer receives.
Setting Up a Trial That Produces a Decision
A mixing trial that ends in a number and a time is worth a day. One that ends in an impression is worth nothing. The sequence below is short and each step removes an argument later.
- Fix the index and the sample mass first, before anything runs. Write both on the trial sheet, with the acceptance threshold
- Sample the unmixed feed, this is the baseline the index is measured against, and skipping it makes the result uninterpretable
- Take samples across time, not just at the end, five or six points across the expected window is enough to show where the flat region begins
- Sample from several positions and depths each time, a single position measures that position, not the blend
- Confirm at the packing end on the confirmatory run, the second batch should be sampled where the product leaves, not where it was mixed
Two batches are enough if both are sampled the same way and both land in the flat region. One batch cannot distinguish a good process from a lucky one, which is why the repeat is part of the trial rather than an optional check.
How the Machine Geometry Enters the Argument
None of the above depends on the machine type, but two machine properties decide whether the flat region exists in a useful place. The first is whether the agitator turns the whole bed over or only the lower part of it. The second is whether the discharge spreads the outflow or concentrates it.
On a horizontal ribbon mixer the inner and outer ribbons move powder in opposite directions along the trough, which is what turns the bed over rather than circulating it in one path. That is why the flat region on this design is usually reached quickly and holds over a useful range of fill level.
For agricultural powders the same geometry has to cope with a wider particle size distribution, which sets a coarser segregation scale and a larger sample. A ribbon blender for agricultural powders is laid out around that, with a full-width discharge to shorten the drop.
Where the powder has to be coated or lightly wetted rather than only blended, the same trial logic applies but the criterion changes, because the target is coverage rather than concentration. That case is closer to the dispersion logic used for emulsion paint, where wetted surface area rather than average composition is the quantity being measured.
Tell Us the Blend and the Criterion
Send the components, their particle size and density, the batch mass and the uniformity you have to reach. Our engineers will recommend the machine and work through the trial sequence with you, quoted directly from the Wuxi factory.
Discuss the mixing trialChat on WhatsApp
WhatsApp +86 138 1677 4633 export@polycmill.com
Where These Blenders Are Built
Trough shells, ribbon flights, shaft assemblies and the discharge arrangements are fabricated and assembled in the same plant in Wuxi. That matters to a uniformity claim because the ribbon-to-shell clearance and the symmetry of the two ribbon handings decide whether the bed turns over evenly or develops a dead zone at one end.
It is worth asking any supplier what clearance they hold and how they check it, because that number is what the flat region of the mixing curve is resting on.
