Second Midds Salvaging Something From Nothing

In my previous article from the Third Quarter issue of Milling Journal, we addressed the need to carefully consider the question(s) to be answered in the conduct of an experiment. The importance of examining system parameters and confirming their stability in all meaningful measures was discussed. The need to ensure those parameters remain stable and do not become noise obscuring your results also was emphasized.

Keep in mind: All of this was identified for the purpose of designing an experiment identifying improvement opportunities leading to measurable benefits to the company. Poor experiment design, planning, execution, and analysis may result in missing an improvement opportunity or encourage costly changes that fail to deliver on intended improvements.

This article attempts to glean insights from data collected from an experiment which considered two factors, namely roll profile and differential in the grinding of 2M stock.

As suggested in the previous article, a simple linear equation becomes more complicated when two factors are considered as shown in Figure 1 below.

Figure 1 describes the relationship between ŷ dependent variable and two independent variables x1 and x2. When the best fit linear regression line is developed, the intercept of the y-axis is β0 , while β1 , β2 , and β3 are the coefficients for the factors x1 and x2 individually and the interaction term x1 x2. Again, the term ε (epsilon) is the error term which accounts for variation not accounted for in the equation by the other terms.

During the experiment, we changed the levels of x1 and x2 making multiple measurement replications to properly estimate value of ŷ and its variability at each combination of level of x1 and x2.

Provided the error term ε remains low, we can determine if changing the independent variables x1 and/or x2 and if the interaction between x1 and x2 is statistically significant, impacting the dependent variable ŷ.

A Better Experimental Plan

A meaningfully designed experiment’s independent variables should consist of two factors at two levels of each factor, specifically roll profile (smooth versus corrugated) and differential (1.23:1 or 1.5:1). For a single replication, four distinct combinations of experiments would have been conducted and measurements taken to build a proper data set for analysis. As indicated in the previous article, all other parameters would have to be tightly controlled as they are not of interest in the experiment as designed.

This experiment reported here was not even a complete block of tests for the factors or variables to be studied. There was also no replication to assess variability of measured results. Therefore, the following analysis is incomplete and certainly not definitive, but it is an exercise in salvaging something from a poorly planned and executed experiment.

Ground Stock Particle Size

The only data reported to the author was particle size analysis using laser diffraction, reporting equivalent spherical diameter of ground 2M stock taken from under the rollermill as indicated in Figure 2.

Figure 2 also shows particle size distribution of commercial cake, pastry, and bread flour for comparison purposes, reflecting differences in soft and hard winter wheat classes.

Particle size in microns is reported at D10, D50, and D90, reflecting the percentage (10%, 50%, and 90%) of particles less than the size reported. Span represents the width of the particle size curve reported and is calculated by subtracting D10 from D90 and dividing the result by D50. While monitoring D50 as measured is a good tool for assessing overall grinding, it doesn’t provide a detailed understanding of the entire product distribution.

General Observations

A D50 of 73.7µ for the corrugated roll at a differential of 1.23:1 is greater than 58.2µ at a high differential (1.5:1) and very similar to the smooth roll with a 1.23:1 differential and a D50 of 60.0µ. At the same differential (1.23:1) the corrugated roll produced a coarser product which makes some sense if the corrugation provides space for particles the smooth roll does not, thereby reducing compression and particle reduction.


Did increasing the differential of the corrugated roll to 1.5:1 increase shear overcoming an assumed reduction in compression resulting at a comparable particle size? It is difficult to resolve, with this limited data, how changing from smooth to corrugated rolls – which increases shear or cutting action while perhaps reducing compression – is impacted by increasing differential which also increases shear.

Were there enough reptations to identify a difference, or have other parameters (i.e., roll gap, stock loading, or spring compression) obfuscated the real outcome? Who really knows?

Sifting Expectations

Ground stock particle size analysis was used to estimate stock distribution from the 2M sifter. Particle size testing suggests no residue or collection stock going to 1QU was collected or perhaps produced. While the percentage of residue or collection stock going to 1QU is expected to be low (0.5-1.5% of 2M), the data suggesting non-existence of the stock is in error or beyond testing machine capability. The quality of this stock is an essential measure of grinding performance, as when sheared it generates ash increase in flour and if over-compressed, increases flour delivered to residue or collection stream.

The balance between these two alternatives is critical to success. See my article from the Second Quarter 2023 issue of Milling Journal for more discussion. Sifter weigh-offs and sample collection would have been useful in this experiment.

Starch Grouping Analysis

Flour physical components include liberated protein matrix, cell wall material damaged, and native starch classified into at least two recognized categories of different particle size ranges, bran and aleurone cell contamination (a significant source of flour ash), and compound particles made up of multiple flour components. Researchers have suggested flour starch particle size as falling into a one of two tri-modal distributions.

Analysis of ground 2M stock produced are shown in Figure 4  and Figure 5 using the two different starch particle size ranges fine and course as shown. Ground 2M stock contains material coarser than wheat flour, resulting in smaller percentages of material in each size distribution shown in this analysis. Figures 4 and 5 present the particle size distribution of commercial cake, pastry, and bread flour for comparison purposes, reflecting differences in soft and hard winter wheat classes.

To express the size distribution based of flour produced and included in the 2M stock analyzed, the particle size ranges can be estimated by mathematical adjustment based on flour yield estimate shown in Figure 3. The adjusted distribution based on 2M flour production estimate is shown in Figure 6. In my opinion, the inability of the initial particle size analysis to account for the actual particles size range in practice and adjusted with the assumed level of flour production using those results, the modified data does not lead to clarity in understanding the impact of roll corrugation or differential in 2M grinding.

Corrugation vs. Smooth Roll

One hypothesis suggests the use of corrugated rolls rather than smooth rolls in the mill reduction passages to provide more shear or cutting action and hence more flour production. Additionally, one might expect the granulation of the flour to be somewhat coarser, given corrugation depth provides space at the grinding line that a smooth roll setup would not provide.

The comparison between smooth and corrugated rolls is confounded, as the corrugated roll may increase shear while reducing compressive forces on compound particles presented to the grinding line. How absorptive properties of flour generated should be considered and perhaps evaluated using Solvent Retention Capacity (SRC) testing.

The data in Figure 2 shows an increase in D50 flour particle size using a corrugated roll compared to a smooth roll at 1.23:1 differential. The percent of particles less than 10 µ and between 10 µ and 45 µ decrease while the percent greater than 45 µ increased. This supports the notion that a corrugated roll may have increased cutting action, however; the corrugations provides space for coarse particles that a smooth roll would have compressed rupturing the particle and creating smaller particles.

Changing to corrugated rolls at the 1.23:1 differential may be a valuable consideration if a coarser flour is required. The lower differential may have also reduced energy, wear, and perhaps capacity, however; these questions were not addressed despite their importance.

Changing Differential

Increasing the differential of the corrugated roll was suspected to increase flour production due to increased shear resulting from the increase in differential. The data in Figure 4 suggests increasing differential of the corrugated roll resulted in comparable results to the smooth roll operating at the lower differential. Clearly, there was no gain in corrugating the roll and increasing differential. Perhaps, however, 2M stock has improved sifting properties which were not assessed. The reduced presence of flour carryover on the various flour cloths might have indicated improved sifting properties if the data were collected.

While it is well known that a 1.23:1 differential works in smooth roll grinding of 2M stock, it would be useful to assess the performance of the smooth roll at the 1.5:1 differential. An opportunity to learn or confirm was missed.

Summary

The time and money spent changing rolls and analyzing samples was wasted. Important considerations and issues were left to chance or unreported. A partial list of issues or questions for consideration are shown below.

  1. Report rate and characteristics (physical and chemical) of stock to each roll. Is it really the same over time or between tests?
  2. Confirm baseline roll gap, pressure, and energy. Are you really grinding with the same roll settings?
  3. How did total power consumption and power consumption per unit of flour generated compare between grinding methods? Only useful if points 1 and 2 are properly confirmed.
  4. Confirm uniform grinding across the grinding line.
  5. Collect samples from under the roll preferable at the point of pick up.
  6. Are the sifter boxes not only flowed but performing the same?
  7. Multiple weigh-offs with samples taken into and out of sifter are essential.
  8. What shifts in ash content were observed by fraction from the sifter?
  9. How was real-time sifting impacted based on sifter efficiency? Yes, collect enough sample to do a carryover analysis on each stream from the sifter.
  10. Was there a shift in production and particle size distribution in the two flour streams produced? Weigh off from the roll and sifter should be reasonably close (within 5% or less).
  11. Were there measurable changes in flour properties such as Moisture Ash and Protein (MAP), SRC for all flour streams weighed off?

The test design and data collection were inadequate to make a meaningful evaluation to recommend or reject proposed roll surface or differential changes. As 2M flour is only 9-15% of wheat to the mill, changes in its properties may not likely result in significant changes to straight grade flour quality. Little if anything was learned. The experiment should have been planned and executed with multiple replications to estimate the mean and variation of collected data to answer important questions of concern. The mill ran with no known significant change to operating or flour quality parameter. Would you really make the decision to spend your company’s funds on making changes based on these results? I hope not. Do it the right way!

Dr. Jeff Gwirtz is CEO of JAG Services, Inc., an international consulting company in Lawrence, KS; 785-341-2371; jeff@jagsi.com. He also is adjunct professor in the Department of Grain Science and Industry at Kansas State University, Manhattan.