Which metrics (RSD = relative standard deviation, CoV = coefficient of variation) are typically used to assess mixing quality in the food industry?
To assess mixing quality in the food industry, standard deviation and the coefficient of variation are used above all. CoV and RSD are mathematically identical: both relate the scatter of the measured concentrations to the mean value. Whether the term CoV or RSD is used is largely a matter of the technical field: CoV is common in process engineering and mixing technology, RSD more often in laboratory analytics and quality control.
Basic characteristic values
The standard deviation s describes the absolute scatter of the measured values around their mean xˉ. However, it depends on the unit and the order of magnitude of the measured values. For comparing different batches, formulations or marker components, the dimensionless coefficient of variation is therefore better suited:
CoV=sxˉ×100%
The relative standard deviation is calculated in the same way:
RSD=sxˉ×100%
A low CoV or RSD means that the concentrations of the target component under investigation vary little between the individual samples. This is an indication of an even distribution. CoV and RSD can, however, only be meaningfully interpreted if the selected marker component can be reliably captured analytically and the samples were taken representatively from the batch.
Interpreting the values
For foods, there is no uniform CoV or RSD limit value that applies to all product groups. The appropriate target value depends, among other things, on the significance of the component, its dosing fraction, the number of particles per sample, the analytical uncertainty, the intended consumption quantity, and the legal or customer-specific requirements. For spice blends, baking mixes, instant products, nutritional supplements, baby food or vitamin-fortified formulations, the homogeneity requirements are often higher than for less critical major components.
As a practical guide, values up to about 5 per cent can indicate very high homogeneity. Values up to 10 per cent are classed as good or acceptable in many technical applications. At values above 10 per cent, it should be checked whether the scatter is caused by inadequate mixing, segregation during discharge, raw material fluctuations, unsuitable sampling, or the analytics. In the animal feed sector, which is methodologically well comparable for dry food mixtures, CoV limits of a maximum of 5 per cent for diluted medicated premixes, 10 per cent for micro- and macro-premixes, and 15 per cent for complete feed are used, for example. These values are not direct legal limits for food, but they provide a useful technical benchmark for risk assessment.
For critical food components, individual value limits should be defined alongside the CoV. A low overall scatter can mask individual local over- or under-concentrations. Minimum and maximum values, the ratio of minimum to maximum, confidence intervals and adherence to the target concentration are therefore important supplements. For ingredients with an allergen, vitamin, mineral, flavour, colour or functional role, the assessment is generally stricter than for constituents that have only a minor effect on safety, labelling or sensory quality.
Sampling and analysis
Sampling is often the largest source of uncertainty in the assessment of mixing quality. A large number of analytically precise measured values does not help if the samples come from only one product zone or do not represent the whole course of the batch during discharge. For an initial robust assessment, many homogeneity tests use at least ten individual samples from a batch. These should be taken distributed spatially or over time, preferably throughout the entire discharge at equal time intervals. An official Canadian test protocol for mixer testing requires at least nine representative individual samples across the whole batch; other guidelines for the animal feed sector specify at least ten samples.
The sample size must match the particle size, the dosing concentration and the later relevant portion. If a sample is too small, the statistical scatter can appear high due to the low number of marker particles it contains, even though the overall batch is sufficiently homogeneous. With very low-dose components, the analytical measurement uncertainty can also noticeably affect the CoV. Sample quantity, extraction method, analytical method, detection limit and repeatability precision should therefore already be considered when planning the homogeneity study.
Suitable markers are components that can be detected unambiguously, selectively and with sufficient precision. Depending on the product, these can be salt or chloride, an added acid, a dye, a mineral, a vitamin, a flavour indicator or a deliberately dosed tracer. The marker component should represent the mixability of the most critical formulation component as closely as possible. A coarse-grained salt, for example, is not necessarily a suitable marker for a very fine, cohesive spice or a low-dose dye.
Supplementary evaluation
Besides CoV and RSD, the mean relative to the target value, minimum and maximum, range, quantiles, confidence intervals and the distribution of the individual values can be evaluated. In continuous mixing processes, time-related characteristic values are added as well: the concentration profile over time, the mean residence time and the residence time distribution help to detect dosing fluctuations, dead zones or delays between addition and discharge.
A mixing kinetics study is also helpful for assessing a batch mixer. Samples are taken at several mixing times and the CoV profile is plotted. Typically, the CoV initially decreases with increasing mixing time. Once a stable minimum has been reached, further mechanical stress can worsen the CoV again with formulations prone to segregation. The goal is therefore not the longest possible mixing time, but the shortest robust mixing time within a validated process window.
With suitable formulations, inline or online methods such as NIR or Raman spectroscopy can supplement classic sampling. They capture concentration or moisture profiles in real time at defined measurement points. For release purposes or for defining the mixing endpoint, however, such systems must be calibrated against representative reference sampling and validated laboratory analytics.
How amixon® makes mixing quality measurable, monitorable and documentable
To assess mixing quality in food applications, the coefficient of variation (CoV) and the relative standard deviation (RSD) are used above all. Both characteristic values describe the scatter of the measured concentrations of a target or marker component relative to the mean. Mathematically, CoV and RSD are identical: CoV=RSD=(s/xˉ)×100%, where ss denotes the standard deviation and xˉxˉ the mean of the individual measurements. Low values indicate low scatter and therefore an even distribution of the component under investigation. CoV is used more often in mixing and process engineering, RSD more in laboratory analytics and quality control.
For food mixtures, there is no universally binding CoV or RSD limit value. The acceptable scatter depends on the formulation, dosing fraction, analytical method, particle size, intended consumption quantity, and the significance of the respective ingredient for safety, labelling, sensory quality and overall quality. As a technical guide, values up to about 5 per cent are often regarded as very good, while values up to approximately 10 per cent can still be acceptable in many applications. For critical ingredients such as vitamins, minerals, dyes, flavours, allergens or low-dose additives, stricter product-specific criteria are frequently set. In the animal feed sector, whose test methodology provides a useful benchmark for dry food mixtures, CoV limits of a maximum of 5 per cent for diluted medicated premixes, 10 per cent for micro- and macro-premixes, and 15 per cent for complete feed are used.
The informative value of any mixing quality metric stands or falls with the sampling. A large number of precise laboratory values is not meaningful if the samples do not represent the spatial and temporal course of the entire batch. Sampling should therefore be stratified: from different positions and especially throughout the entire discharge. Many homogeneity tests use at least ten individual samples per batch. In doing so, sample quantity, particle size, the dosing fraction of the target component, the analytical detection limit and the method's repeatability precision must be matched to one another. Official guidelines for mixer testing require at least nine representative samples across the batch; other industry-specific recommendations specify at least ten samples.
amixon® supports the assessment of mixing quality with a sampling strategy tailored to the specific formulation. On batch mixers, low-dead-space discharge elements and the large CleverCut® accessibility support defined sampling points within the mixing chamber and across the time course of discharge. In the amixon® pilot plant, a trial and sampling plan can be developed with the original product. This defines sampling locations, number of samples, sample quantity, mixing times, analytics and acceptance criteria. Separating method variance from process variance is essential here: imprecise sampling or laboratory analysis must not be wrongly assessed as inadequate mixing quality.
Besides CoV and RSD, adherence to the target value, the minimum and maximum of the individual values, range, confidence intervals and the trend of the values across discharge should be taken into account. A low overall CoV can mask individual local over- or under-concentrations. In continuous processes, time-related concentration profiles, mean residence time and residence time distribution are added as supplementary metrics. For batch processes, a mixing kinetics study is useful: the formulation is examined at several mixing times, so that a robust mixing time window can be defined. The goal is not the longest possible mixing time, but the shortest reproducible time at which the specification is reliably met. With formulations prone to segregation, an excessively long mixing time or unfavourable emptying can even worsen homogeneity again.
A stable process basis is created when the validated mixing parameters are consistently adhered to. Mixing time, circumferential speed, fill level, raw material sequence, dosing quantities, addition times and, where required, temperature or liquid addition profiles can be stored as a formulation program in the PLC. Deviations from the target time, target speed or other critical parameters are thereby detected and documented immediately. Monitoring these parameters is an important prerequisite for reproducible mixing batches, but it does not replace periodic analytical proof of the actual homogeneity. In particular with fluctuating raw material batches, low-dose components or high requirements for allergen control, a product-related check remains necessary.
An ERP connection as well as barcode or RFID identification can supplement batch documentation. This allows formulation, raw material batches, mixing program, fill level, process values, operator interventions, cleaning status and product release to be linked seamlessly. This supports batch traceability, root-cause analysis for deviations, audit readiness and the validation of critical process steps. For GMP-oriented projects, amixon® can support DQ, IQ and OQ; documentation and automation can be aligned project-specifically with EU-GMP and FDA 21 CFR Part 11.
Warning signals such as fluctuating analysis values, visible segregation during discharge, differing colouration, uneven moisture, batch-dependent bulk density or unstable dosing behaviour should be consistently investigated. Causes can include unsuitable mixing parameters, inadequate premixing, dosing errors, raw material fluctuations, segregation after the mixer, or non-representative sampling. In the amixon® pilot plant, such effects can be distinguished from one another using the original product through mixing, discharge and, where applicable, conveying trials, before they lead to scrap or rework in production.
For food applications, amixon® can provide a product-gentle, hygienic design. SinConvex® forced restratification conveys the product upward at the periphery and guides it back down under gravity at the centre. Tool circumferential speeds can be set at approximately 0.8 to 3.5 m/s, depending on the task. The mixing effect arises through controlled product circulation rather than pronounced throwing, impact or crushing zones. This allows sensitive structures, coatings and agglomerates to be preserved with limited heat input. Where targeted de-agglomeration is required, cutting rotors can be switched on locally and for a limited time. The actual degree of product protection must be assessed for the specific formulation through trials.
Mixing chambers welded free of crevices and ground smooth, mixing tools supported only at the top without a lower shaft passage in the product area, CleverCut® inspection doors with OmgaSeal® seals, low-dead-space discharge elements and integrated washing lances support hygiene, accessibility and cleaning. Depending on the project, dry or wet cleaning as well as a validatable cleaning strategy can be provided. A design to EHEDG as well as taking FDA hygiene guidelines and 3-A Sanitary Standards into account is possible project-specifically. Extensive residual discharge reduces product losses and eases rapid product changes; however, it does not replace the necessary cleaning validation for allergen-critical formulations.