Are there best practices for process data acquisition and OEE analysis in powder mixing operations?
Yes. Robust process data capture and OEE analysis in powder mixing operations requires process, batch, quality and downtime data to be recorded and evaluated within a shared context. Unlike continuously producing plants, a batch process consists of individual, often differently long phases such as raw-material provision, dosing, filling, mixing, sampling, discharge, cleaning and product changeover. An OEE metric is therefore only meaningful if the time models, loss categories and quality criteria are clearly defined.
At the outset, the critical process parameters and relevant quality attributes should be defined. In powder mixing technology, these typically include mixer rotational speed, torque, mixing time, batch mass, fill level, bulk density, moisture, temperature, pressure or vacuum, and the dosing accuracy of the individual components. Depending on the product and objective, data on particle size distribution, homogeneity, flowability, bulk density, energy consumption and yield can also be relevant. Suitable metrics for quality assessment include, for example, the relative standard deviation, the coefficient of variation, the proportion of batches releasable on first pass, or the amount of rework and scrap.
What matters is the unambiguous linking of this data. Every relevant data record should be associated with a batch number, product ID, recipe version, plant identifier and timestamp. Where required for root-cause analysis, raw-material lots, operator interventions, shift assignment, ambient conditions and the tool configuration used should also be documented. Only then can it later be traced whether a deviation is attributable, for example, to a particular raw-material lot, a changed fill level, a recipe change, a dosing error or a technical fault.
Data capture should be as automated as possible and as close as possible to the process source. Process values can be taken directly from sensors, drives, scales or the PLC (programmable logic controller) and then stored in a historian, MES (Manufacturing Execution System) or a suitable analytics platform. Manual entries nevertheless remain necessary, for example to justify downtime, classify deviations or document measures. They should be structured through predefined categories, unique user accounts and plausibility checks, so that evaluations are not distorted by inconsistent free-text entries.
The required sampling rate should not be fixed as a blanket rule, but should be oriented to the dynamics of the respective signal and the analysis objective. For OEE evaluation, precise start and end times of the process phases and the correct recording of interruptions are frequently more important than high-frequency time series. For fast dosing operations, for the analysis of torque profiles, for vibration, or for assessing short-term process disturbances, on the other hand, considerably higher temporal resolution may be required. For pure status changes, event-based capture is often sufficient, while process signals for endpoint determination or root-cause analysis should be recorded at a suitable temporal density.
A robust data architecture links the field and control level with batch documentation and the analytics level. The ISA-88 standard provides an important framework for this, because it structures batch processes via recipes, plant modules, operations and phases. ISA-95 supplements this approach with models for integration between control, production and enterprise systems. This makes it possible, for example, to clearly map which recipe version was used on which plant for which batch, in which phase a stoppage occurred, and what quality data was subsequently available for the batch.
Standardised interfaces are particularly valuable for technical communication between machines, controls, historian, MES and the analytics environment. OPC UA (Open Platform Communications Unified Architecture) is an open, platform-independent architecture for the secure, cross-vendor exchange of industrial information. It can help reduce proprietary point solutions and provide data from different plants consistently. Depending on the architecture, event-oriented publish-subscribe approaches can additionally be used. What matters is less the choice of a single protocol than stable, documented and monitorable data transmission with unambiguous data models.
OEE consists of the factors availability, performance and quality. Mathematically:
OEE = Availability × Performance × Quality
In batch operation, however, it must be bindingly defined before the calculation which times count as planned production time and how cleaning, changeover, product changeover, maintenance, quality releases or campaign changes are classified. Without these rules, OEE values are hardly comparable between products, plants or sites. Particularly with powders, cleaning effort can vary considerably depending on the active ingredient, recipe, allergen or cross-contamination risk.
Availability assesses the extent to which the plant was actually ready for use within the defined planned production time. Unplanned downtime caused by technical faults, missing raw materials, dosing problems, sensor faults, cleaning deviations or waiting times for releases should be recorded with clear causes. Depending on the OEE model defined by the operation, planned cleaning can be treated as part of planned non-production time or as a changeover activity. Regardless of this calculational classification, unusually long cleaning is always an important starting point for improvement, for example through optimised campaign planning, improved cleaning sequences or shorter product-changeover times.
The performance metric compares the actual batch time with a realistically defined target time. This target time should be product-specific and should not consider only the actual mixing step. Depending on the objective of the metric system, dosing, filling, discharge and necessary auxiliary times can also be included. Performance losses arise, for example, from poor flow properties, bridging, dosing faults, re-dosing, extended mixing times, restricted discharge or short recurring interruptions. The target times must be reliably derived from process knowledge, validated recipes or long-term reference data. A blanket standard time for very different products leads to distorted results.
The quality component shows what proportion of a produced quantity, or how many batches, meet the defined specification without rework. First pass yield is a particularly meaningful metric for this. Quality losses can arise from inadequate homogeneity, dosing deviations, incorrect moisture, unacceptable particle size distribution, impermissible bulk density, re-mixing, rework or scrap. Adhesion, residual quantities in the mixer, discharge losses and dust losses should also be recorded. Clear logic is important here: if a batch requires more time due to re-mixing, this initially represents a performance or time loss. If product is discarded or cannot be released without rework, this concerns the quality component. Double counting must be avoided.
In-line and at-line analytics can considerably improve the data basis. Near-infrared spectroscopy, for example, can be used to monitor concentrations, moisture or homogeneity, provided the measuring point, method, calibration and data model are suitable for the specific application. Torque, power or current-draw profiles can likewise provide additional indications of product behaviour, mixing progress or process deviations. However, such signals should not be interpreted in isolation. Their informative value only emerges from comparison with product and quality data and from a traceable assessment across multiple batches.
An adaptive mixing time, in which the process is ended not after a fixed time but on reaching a defined mixing endpoint, can increase efficiency while also avoiding over-mixing. This requires a sufficiently robust measurement method, a secured decision logic and clearly regulated handling of measurement errors or implausible signals. In regulated environments, the measurement and evaluation methods must be fit for the intended purpose and, where required, qualified or validated.
Statistical Process Control, trend analyses, Pareto evaluations and structured root-cause analyses have proven effective for evaluation. SPC helps to detect changes, drift or increasing scatter in critical process parameters at an early stage. Pareto analyses show which causes of downtime, fault types or scrap causes have the greatest impact. Methods such as 5 Why or Ishikawa then help to avoid stopping at symptoms and instead identify technical, organisational, material-related or procedural causes. For complex relationships between many process and quality data, multivariate data analysis can be worthwhile.
In pharmaceutical or comparably regulated applications, the continuous capture and statistical evaluation of process and quality data is also a core component of Continued Process Verification. The FDA describes process validation as the collection and evaluation of data from process design through to commercial production. In the third stage, Continued Process Verification, routine production is meant to provide ongoing evidence that the process remains in a state of control. Relevant trends in process data, input materials, intermediates and finished products should be statistically evaluated and reviewed by trained personnel for this purpose.
A data integrity concept should exist for all process, batch and OEE data. Data must be unambiguously attributable to a person or system, legible, captured contemporaneously, original or traceably derived, accurate, complete, consistent, durably available and protected. In practice, this means, among other things, role-based user rights, controlled recipe and parameter changes, audit trails for relevant corrections, traceable data transfers, regular backups and tested recovery procedures. Calculation logic for OEE, target times and loss categories should also be version-controlled and documented, so that metrics remain correctly comparable over longer periods.
The benefit of data capture ultimately does not arise from dashboards alone, but from a lived improvement process. Regular shop-floor reviews should answer concrete questions: which causes of downtime account for the most lost minutes? For which products does re-mixing occur particularly frequently? Which raw materials or recipe variants increase dosing time? Which cleaning operations take longer than planned, and what technical or organisational measure can improve this? Blanket benchmarks such as 85 percent OEE are rarely directly transferable in this context. Plant- and product-specific target values that take into account the actual batch sequence, cleaning effort, product changeover, quality requirements and operational planning are more meaningful.
How amixon® can support the OEE of powder mixing plants
The OEE of powder mixing plants is essentially determined by availability, performance and quality. A reliable assessment requires product- and plant-specific times, downtime causes, batch yields, quality data and cleaning effort to be systematically recorded. amixon® can support operators in this through apparatus design geared towards cleanability, maintenance-friendliness, process stability and dischargeability, as well as through pilot-plant trials and qualification-relevant documentation. The actual OEE, however, always also depends on the product, recipe, production planning, raw-material quality, operation, maintenance strategy and the operational environment.
For availability, design features that can reduce unplanned downtime and simplify maintenance work are important. In many amixon® mixers, the mixing tool is supported exclusively above the mixing chamber. This eliminates a lower shaft passage as an additional product-contact sealing system. A comparatively low rotational speed and a reduced number of wear-relevant components can reduce maintenance requirements. The actual service life of individual components, however, depends on the product, mode of operation, cleaning method, load and maintenance, and should therefore be assessed for the specific application.
Regular inspections and preventive maintenance help to make maintenance measures plannable and reduce unplanned failures. Condition-based or predictive maintenance can be worthwhile if suitable operating, condition and fault data are available and the measured values are linked to clear actions. amixon® supplies selected wear parts already with the initial delivery of a mixing plant. In addition, according to the company, it holds the majority of spare parts in stock at the Paderborn site and offers a lifetime spare-parts service; additional service bases are located, among other places, in China, Japan, India, Korea, Thailand and the United States.
For discontinuous powder mixing processes, the performance metric of the OEE should be calculated on the basis of a clearly defined target batch time. Depending on the OEE model, this comprises not only the mixing time but also dosing, filling, discharging, product changeover and, where applicable, production-relevant auxiliary times. amixon® can influence the duration of these process steps through the choice of mixing system, the design of the discharge, the cleaning concepts and the accessibility of the mixing chamber.
For the mixing time, what matters is that the required homogeneity is reproducibly achieved without unnecessarily subjecting the product to mechanical or thermal stress. The actually required mixing duration cannot be reliably generalised, because it depends on the powder and particle properties, the recipe, the fill level, the dosing sequence, the tool design and the quality requirements. It should therefore be determined with the original product and the intended operating conditions. For certain applications, amixon® mixing systems, such as the KoneSlid® mixer type KS, can enable short mixing times and fast discharge. Statements such as an ideal mixing quality after a fixed number of revolutions should always be understood as product- and application-specific trial results, not as a general performance commitment.
Discharge, too, influences the effective batch time, yield and cleaning effort. According to amixon®, the KoneSlid® mixer type KS is designed for fast discharge; the product can be discharged within a few seconds, depending on the product and operating conditions. ComDisc® technology is intended to convey product residues gently and without segregation to the outlet at the end of discharge. According to amixon®, discharge rates of up to 99.997 percent or more are achievable depending on the product and application. Such figures are application-dependent and should be verified through trials for the specific product. A high level of residual discharge can support batch yield and product changeover as well as reduce cleaning effort.
Fast product changeovers require a mixing chamber that is readily accessible and designed for easy cleaning. Constructional measures such as smooth product-contact surfaces, reduced dead spaces, a top-mounted mixing tool and suitable access points can make manual inspection and cleaning easier. Spray lances, WIP or CIP concepts can be provided on a project-specific basis depending on the product, cleaning strategy and hygienic requirements. The actual cleaning time and cleaning effectiveness, however, must be defined by the operator for the product, soiling, cleaning medium and method, and validated where necessary.
When using a container mixer, the spatial separation of mixing, container transport and cleaning can improve the availability of the central mixing station. Whether this actually means cleaning time no longer limits mixing capacity depends on the number of containers, container logistics, the available cleaning station, the product sequence and the release processes. The economic benefit should therefore be assessed using a specific material-flow and capacity model.
The quality component of the OEE is largely determined by the reproducibility of the mixing quality, adherence to the recipe, yield and the number of instances of rework. A suitable mixer design can help to reproducibly achieve homogeneity within a defined operating window. Nevertheless, the mixing quality must be demonstrated for each recipe on the basis of the raw-material properties, the fill level, the batch size, the dosing sequence and the selected process parameters. Mixing quality that is fully independent of fill level across a range of 10 to 100 percent cannot be committed to as a blanket rule, because flow conditions, mixing kinetics and the ratio of product quantity to tool geometry can change depending on the application.
Recipe management in the PLC or in a higher-level system can help to apply approved process parameters in a controlled and repeatable manner. Barcode-based material identification, batch documentation and integration with ERP or MES systems can improve traceability and make deviations visible earlier. This requires a suitable data model, clear roles and responsibilities, controlled changes and reliable transfer of the relevant data. In GMP-regulated applications, data integrity, user rights, audit trails and the validation of the systems used must additionally be governed according to the respective risk profile.
For the continuous mixer type AMK, amixon® describes a controlled start-up procedure: the gravimetric feeders initially start with a low mass flow, the mixer is put into operation at approximately half the fill quantity, and the discharge is opened step by step after reaching approximately 80 percent of the usable volume. At the end of production, the dosing flows are reduced in a controlled manner and the mixer is fully discharged. This concept can reduce start-up and run-down losses. Whether and to what extent off-spec quantities are avoided, however, depends on the design, the stability of the dosing, the properties of the raw materials, the control technology, process monitoring and the specific quality criteria.
To quantify OEE potential, target and actual times should not be estimated but determined, as far as possible, under realistic conditions. amixon® can carry out mixing, discharge and, where applicable, cleaning trials with the original product at the pilot plant and document the results. The transferability of these results to a target size must be checked on the basis of the product, geometry, fill level, tool design and process requirements. Pilot-plant data can form a reliable basis for the technical design, the definition of provisional target times and comparison with an existing plant. They do not, however, replace acceptance testing and performance assessment in the later production operation.
In regulated environments, amixon® can provide qualification-relevant technical documentation and support the operator with the DQ, IQ and OQ. The starting point is the operator's User Requirement Specification. Depending on the project, material certificates, surface specifications, welding documentation, test protocols and documentation on product-contact components, for example, can be provided. The plant can be designed to project-specific requirements such as GMP-compliant hygienic design, ATEX, ASME, EHEDG guidelines, FDA requirements or 3-A Sanitary Standards. The final regulatory assessment, process validation and system validation, including the requirements for electronic records and signatures, remain with the operator.
According to the company, amixon® develops and manufactures at the Paderborn site. The high level of in-house manufacturing and central quality control can support project-related documentation, the traceability of technical changes and long-term spare-parts supply. For long-term maintenance, it is above all important that technical drawings, material information, specifications and component identification are maintained in a controlled manner. Remanufacturing components after decades depends on the original specifications, the availability of suitable materials and components, and any applicable technical or regulatory requirements.