A stable continuous detergent process begins with precise gravimetric dosing of all solids and mass-flow-controlled liquid addition. What matters is that the recipe ratio, mixer fill level, residence time, energy input and liquid distribution are kept within a qualified process window. The specific control architecture depends on the recipe, mixer type, throughput, additive quantity, sensitivity of the ingredients and the available measured variables.

The gravimetrically captured mass flow of a main component frequently serves as the reference variable. All other solid and liquid streams are guided proportionally to it via ratio controls. This keeps the recipe proportions constant even with controlled changes to the total throughput. Loss-in-weight feeders are suitable for continuous solid streams, provided bridging, segregation in the storage vessel, refilling operations and dosing fluctuations are kept under control. With critical components, the actual values of all dosing streams should be continuously monitored and checked for plausibility against the recipe ratios. The combination of gravimetric dosing, continuous mixing and real-time measurement is also used in other continuous powder processes to monitor the recipe and product quality.

Liquid addition should preferably be controlled on a mass basis. Coriolis instruments can capture mass flow and density for this purpose; however, their selection and design must suit the viscosity, temperature, pressure range, gas-bubble content and possible solids loading of the additive. The dosing pump is guided in a fast inner control loop to the desired liquid mass flow. A higher-level ratio controller forms the setpoint from the current solids mass flow and the liquid-to-solid ratio defined in the recipe. With highly viscous surfactants, binders or fragrance formulations, temperature-controlled and insulated supply vessels, lines, pumps and nozzles help to limit viscosity fluctuations and the resulting dosing errors.

Cascade control is particularly sensible where both a fast-reacting flow loop and a slower correction via quality variables are required. The fast loop stabilises the actual liquid flow. A slower, higher-level loop can, provided a validated measurement is available, adjust the setpoints. Possible quality variables are residual moisture, bulk density, active-ingredient or marker concentration, and the proportion of fine or oversize particles. Not every variable can be measured inline for every detergent recipe with sufficient accuracy. A pH value, for example, is only a suitable control variable if an actual aqueous neutralisation or reaction step is taking place; for a dry final mixing process it is generally not relevant.

The injection of the liquid is substantially decisive for the stability of the process. Droplet size, spray angle, spray rate, nozzle position and the state of motion of the powder must be coordinated with one another. A liquid should be introduced into a sufficiently agitated product zone so that it distributes quickly. With viscous media, heated single-fluid nozzles, two-fluid nozzles or other suitable atomisation concepts may be required. A two-fluid nozzle, however, additionally needs a stable compressed-air supply and a matched air volume; a higher air pressure alone does not guarantee better distribution. With a wide throughput range, several switchable nozzles or correspondingly turndown-capable dosing and atomisation technology can be sensible.

Mixer fill level, throughput and rotational speed together form the central operating control. The fill level influences the mean residence time and the residence-time distribution. The throughput changes both the dwell time and the ratio of liquid addition to product movement. The tool speed influences mixing intensity, energy input, possible de-agglomeration and heat generation. These variables should not be changed independently of one another. Instead, a permissible operating range is defined within which homogeneity, bulk density, particle structure and moisture are maintained. Residence-time distributions can be characterised with tracer trials; they describe how long different product fractions remain in the mixer, and help in assessing start-up phases, recipe changes and disturbances.

Torque, power consumption or motor current can be used as process signals. A rise can indicate changes in bulk density, an increased fill level, rising viscosity, local over-wetting or caking. However, these signals are not unambiguous. They should therefore not be used as the sole proof of lump formation, but combined with further measured values such as fill level, mass flow, product temperature, moisture or sample analysis. Limit values can serve as an alarm, as a limiting protective function, or as the trigger for a controlled transition to a safe operating state.

For quality monitoring, process analytics and regular atline analyses can be combined. Spectroscopic inline moisture measurement can be suitable for the moisture or composition of powder mixtures, but requires a representative installation situation and a calibration model carefully validated against reference methods. The FDA explicitly points out that inline endpoint or mixing models must be validated for suitability and specificity for the components of interest. In addition, bulk density, moisture, temperature, sieve analysis, colour values, fragrance content or active-ingredient content can be monitored at the discharge or in the laboratory, depending on the recipe and quality requirement.

Start-up, shut-down and recipe-change phases must be defined as their own operating states. All feeders should be ramped up and down synchronously and in the specified recipe ratio. The hold-up in the mixer, the product quantities in the dosing lines and the residence-time distribution must be taken into account here. Only once the mixer is operating in a stable state and the product quality has been demonstrated should product be released for regular filling. With recipe changes, knowledge of the residence-time distribution can help to determine the transition quantity traceably and keep it separate.

Model-based control, for example with feedforward, model predictive control or dead-time compensation, can be sensible where long or variable transport times exist and sufficiently robust process models are available. It is not, however, a standard requirement for every detergent mixing line. A robust basic architecture of gravimetric dosing, ratio control, a stable liquid flow loop, fill-level control, clear operating limits and quality-supported monitoring is often sufficient. Advanced control methods should only be used once they offer a demonstrable benefit over this basis for product quality, throughput or scrap avoidance.