Milling Process Parameters Statistical Analysis — a Suitable Method to Determine Milling Process Parameters

Editor: Manja Wühr

The reason for unsatisfactory results with milling process optimizations lies in the critical quality parameters. Using a conical sieve mill as an example, the article will show that it is possible to determine the correct process parameters by means of statistical analysis.

Related Companies

Changing the rotor at the new sieve mill ConiWitt (Pictures: Frewitt)
Changing the rotor at the new sieve mill ConiWitt (Pictures: Frewitt)
( Archiv: Vogel Business Media )

Thanks to the introduction of a new generation of conical sieve mills and the upgrading of the technical center with an analytical laboratory; Frewitt has eliminated the trial and error method. Conventional methodology has been replaced with statistical design of experiments (DOE) making it possible to look for a pattern in the available data.

When selecting the specifications, firstly a preliminary analysis is performed to determine the process variables that influence the properties of the process. In collaboration with the user, it is not uncommon that the number of factors can be reduced to a priori, a fact that has been confirmed in this test procedure as well. The head of Frewitt’s laboratory selected the ConiWitt 200 conical sieve mill as a test device for analysis of the following factors: rotor speed, rotor type, sieve type, and the target parameters: flow rate, particle size (distribution) and temperature change (ΔT) of the product.


In the fine chemical and pharmaceutical industries, the specifications are defined very precisely, in order to guarantee consistent quality of the product to be processed as well as the final product, and to assure the efficience of the drug. For this reason, special attention is paid to temperature change, its effect on heat-sensitive products, and particle size distribution. Concerning flow rate is a key factor in the parameters of powder processing.

Now we come to the process analysis procedure itself: Experimental designs are prepared with the help of a program which guides the user through the entire optimization process, from the design phase to the evaluation of the analysis data.

The program runs through three phases: screening (defining the key factors and discarding the unimportant ones), modeling (testing the key factors and minimizing the relevant, influencing variables) and optimization (precise modeling of the key factors with the target variables). A hard, crystalline, inorganic mineral salt was chosen as a test product. The process parameters (factors) were defined as follows: rotor type (round, square), rotor speed (5–14 m/sec), sieve (with 1–2 mm openings).

The selected target parameters were: flow rate (kg/h), particle size (distribution), ΔT of the inorganic mineral salt.

Tests results obtained in past measurements with the “traditional” method showed that the flow rate with sieve X and rotor type round was always considerably lower in comparison to sieve X with rotor type square. Using statistical analysis to evaluate the results; however, led to a surprising finding: the supposed correlation could not be confirmed. After changing several process parameters, i.e. sieve X and rotor (square), a reciprocal relationship could indeed be established, but now the flow rate with sieve X and rotor round was all of a sudden considerably higher than with sieve X with rotor square.

Showing clear relationships

The process model establishes the correlation between the influencing factors identified as important (independent variables) and the target parameters (dependent variables). The process data are used to fit the behavior of the model to that of the actual process itself. The process model has yet another substantial advantage: It provides direct support during the process management, and can be used to simulate the process flow. The reaction to changes in target parameters can be predicted on the basis of the current state of the process. Conversely, by inverting the process model; it is possible to come up with a procedure for intervening in the process flow, in order to achieve a specific, predefined target, value as a process output.

The software models the process parameters in a virtual cube as multi-dimensional grids, in which the measurement points of the individual factors and target parameters are plotted as a function of their values. This graphic presentation makes it possible to discern correlations between the given parameters, as well as interpreting them correctly.

For more efficiency

Analysis of variation is used for analyzing nominal (qualitative) factors, whereas quantitative (metric) factors are analyzed using regression analysis, based on a linear combination of basis functions. The model parameters are defined so that deviations between the data and the model are as small as possible. On the left, it is shown both the results of the analysis and how the factors can be adapted to fit the specifications (target parameters).

  • Specification high flow rate: The customer should use a wide mesh (2 mm) sieve and set the rotor speed to 14 m/sec.
  • Specification very fine powder (<1 mm) with high flow rate: A 1.5 mm mesh sieve and a rotor speed of 5 m/sec should be selected.
  • For a chemically heat-sensitive product that needs to be finely (ca. 0.8 mm) ground, a 2 mm mesh sieve and a rotor speed of 14 m/sec are recommended.

As with many other mechanical procedures, powder milling or sizing involves a product-specific process in which the subprocesses can be controlled and improved once the process parameters (e.g. rotor type, rotor speed, sieve type) and their influence on the target parameters are precisely determined.

The correlations between the factors and the target parameters can be proven, and as the results show; they have a greater influence on the final product, than was generally thought. The results of this analysis have been incorporated in the guideline, aid to Frewitt customers in fulfilling their product and process specifications as early as the specification selection and preliminary analysis of the process variables phase.


With the help of statistical analysis, Frewitt has taken yet another important step on the road to better quality; one that also enables customers with API products to carry out experiments, and fine tune their milling processes for optimum output with minimal use of materials and resources (time, personnel, energy).

Articlefiles and Articlelinks