A new imaging method for millers

by Jeff Gwirtz
Share This:
This is a brief summary of the paper “Near Infrared Focal Plane Array Quantitative Imaging of Break Stock” presented at the 115th Annual IAOM Conference and Expo, Henry B. Gonzalez Convention Center, May 4, 2011 in San Antonio, Texas, U.S. The presentation was possible through the collaborative work of Mark D. Boatwright, David L. Wetzel, Jeff A. Gwirtz and Tyler R. Nickoley in the Grain Science & Industry Department, at Kansas State University (KSU). Boatwright, a milling science and management graduate, is presently a graduate student supervised by Wetzel in the Microbeam Molecular Spectroscopy Laboratory. Nickoley, an undergraduate milling science and management chemistry option student in the Microbeam Laboratory, contributed in both the analysis and milling aspects of the project. The milling work directed by Gwirtz was performed in the Milling Teaching Laboratory in the Grain Science & Industry Department, KSU.

The purpose of the work was to engage students going into the milling field with exposure to an innovative application of an analytical tool. The information collected is currently used in the classroom to connect students with recently developed technology that has practical application in solving processing challenges and enhances student knowledge. The chemical imaging scheme for the determination of endosperm yield developed in a research project at KSU is already being used as a teaching tool. This project fosters initiative to seek out innovative solutions as an important attribute for upcoming industry leaders. By exposing our undergraduate milling students to the capability of mathematical estimation of the product of successive steps in a segment of the operation, results in endosperm yield and mass balance are revealed. In the article, the recently developed analytical tool is demonstrated for practical use as a method to enable mill optimization.

The optical method applied to flour milling intermediate product on a research project by Wetzel, Posner and Dogan was published “Indium Antimonide Focal Plane Array Chemical Imaging Enables Assessment of Unit Process Efficiency for Milling Operation” in the scientific journal Applied Spectroscopy (December Issue, Appl. Spectroscopy, 2010, 64(10), the official publication of the Society of Applied Spectroscopy. Due to instrument sophistication and cost it is not yet suitable for analysis on a day to day basis in mill operation analysis. However, it is a valuable tool that can be used in the design and optimization of operational settings in the commissioning of new equipment. Table I shows information about method capabilities which have been identified in the Microbeam Laboratory. Near infrared radiation from the quartz tungsten halogen lamp shines on the sample where it is diffusely reflected by the specimen to the lens (Figure 1). The lens focuses the rays onto a tunable filter from there to the focal plane array. Solid state detector elements in the array of the high tech instrument perform approximately 82,000 analyses simultaneously which represents the number of pixels in a field of view on the sample stage. Chemical features distinguish the component of the spectrum from each of these pixels. Image contrast is produced through the choice of wavelength. Endosperm can be highlighted in an image using the 2100 nm wavelength strong absorption band of the spectrum. Similarly, nonendosperm can be highlighted with the 1740 nm wavelength image. A schematic of the process is provided in Figure 1.

The commercial Malvern instrument used at KSU was originally developed by chemists at the National Institute of Health in Bethesda, Maryland, to examine diseased tissue in the health field. In the pharmaceutical industry, the imaging near-infrared spectrometer is used on medication to assure proper dosage of multiple active ingredients in an excipient matrix — i.e. the aspirin, acetaminophen and caffeine in an APC tablet — would appear as different false color coded spots in the tablet image to show the distribution and content of each.

Very pure endosperm, such as in the 1st Midds top flour stream from the KSU Pilot Flour Mill in Shellenberger Hall, is used as the “standard” for comparison. The multiwavelength spectroscopic profile spectra are obtained to profile the endosperm. Bran obtained from the same milling that is relatively free of endosperm is ground, and the multiwavelength spectroscopic profile spectra are obtained to highlight nonendosperm. These reference data enable statistical manipulation of the spectra of each pixel in the image to classify it according to the binary system.

The false color image contrast shows the predominant composition of each pixel as shown in Figure 2. Conceptually, the binary code of 1 or 0 is applied, and then the count of endosperm pixels is divided by the total number of pixels in the image to obtain the endosperm concentration of the optical field of view. In practice, a multivariate statistical process shades each pixel, the summation of which produces the concentration of endosperm and non-endosperm. Validity of the endosperm numerical values are subject to the replication of fields of view. The perception provided by the false color images is numerically evaluated by further treatment of the data by dedicated software furnished with the instrument. The analytical reliability has been established by extensive experiments in the Microbeam Laboratory at KSU.

Figure 2 shows images from a commercial mill purifier before adjustments. The calculated endosperm mass balance is shown in red. The warm color or redness of the image indicates higher amounts of endosperm. Before adjustment, the product was 64% and the rework was 37%. After adjustment, the product yield increased to 77% with 23% rework, which is certainly a dramatic difference.

The diagram in Figure 3 depicts a conventional first and second break laboratory flow. The second cut which is normally directed to a Sizings system for processing was separated further with stocks from both first and second break combined for endosperm content. The fractions imaged are designated A through J in descending particle size. Figure 3 shows typical false color images of intermediate products in which the warm color (red) shows the endosperm present in each image. Note that endosperm content is more prevalent in the finer fractions based on the red false color in the image. Also note that the finer material is more homogeneous.

Endosperm values stated as percent and listed by sieve size are the means of the three replicate specimen images calculated independently. From three different fields of view the image chosen to show in the diagram is the one that best represents the calculated numerical mean. Three fields of view for a heterogeneous material produce numbers that vary. For the coarse heterogeneous material statically a number of fields of views must be averaged. By the third sieve level there is less heterogeneity.

For the top three images tail over of the 1041, 900, 750 micron screen, statistically there are distinct differences of 38%, 52%, 80% endosperm. For the 80% endosperm image, red pixels start to dominate. Material A (tail over of 1041) go to the third break and has the lowest amount of endosperm. Material (-1041, +355; A-G) are sent to sizings in the experimental mill, but in a commercial mill it could be diverted to purifiers 1 and 2 or a fine break. Based on observations and our imaging analysis, one could identify size fraction destination. For example, B could be sent to a 3 BK fine roll. Purifier 1 could process fractions D, E and possibly F since they are similar in endosperm content. Purifier 2 could process slightly smaller, but more pure material could process F and G. Destination of F is dependent on granulation specifications and purity requirements. Of interest is endosperm connected to the non-endosperm (compound particles). Purifiers will aid in the separation of compound particles containing greater proportion of endosperm from those containing predominantly bran.

Quantitative near infrared imaging enables high-tech determination of endosperm content. Endosperm content can then be multiplied by the flow rate to obtain the mass balance of endosperm in the process of grinding and separation. From these data, endosperm yield can be determined as a function of particle size.

Break Setting Study

A study utilizing three first break settings, and three second break settings with fixed setting of a third break, sizings, and 1 Midds (reduction) was conducted in a single replication to determine if near infrared focal plane array quantitative imaging could detect differences in endosperm released from the primary breaks. First break gap settings were 0.020, 0.022 and 0.024 inches (0.508, 0.559, 0.610 mm). Second break gap settings employed were 0.010, 0.012 and 0.014 inches (0.254, 0.305, 0.356 mm). Third break was set at 0.003 inches (0.076 mm), Sizings gap was 0.005 inches (0.127 mm) and 1 Midds (reduction) was 0.001 inches (0.025 mm). Three different roll gap settings for first and second break result in nine different 1BK/2BK combinations.

The first and second break roll gap combination which provided the best recovery of endosperm was 0.022 inch (0.559 mm) for first break and 0.010 inch (0.254 mm) for second break. The first and second break roll gap combination 0.024/ 0.014 inch (0.610/0.356) roll gap for first and second break resulted in the poorest endosperm recovery. The proportion of endosperm per unit of extraction was 0.90 and 0.83 for the optimal and sub-optimal combination, respectively. The endosperm content of the flour released was 100% and 98.8% for the optimal and sub optimal combination, respectively. With the more aggressive combination, the total flour produced was higher. However, as might be expected, the non-endosperm fraction also increased.


Near infrared focal plane array quantitative imaging has potential applications to mill process development and management. It could be further used to enhance understanding of the tempering process along with changes in the grinding process due to roll wear or corrugation selection. Mass balance and control can be obtained by the unit process through screen selection and purifier settings to accomplish the intended goal.

The granulation curve is a valuable tool identifying mass balance based on particle size. Its use in stream separation with respect to quality is dependent on the user’s understanding and knowledge of the milling process and the anticipated quality characteristic of the material.

The cumulative ash curve tool addresses both the measurable but dubious quality characteristic of flour ash and its presence in the flour as the flour mass is known. Again, user understanding and knowledge of the milling process is critical in its use.

The adage of “Grind to Sift and Sift to Grind” is based on the existence of a fixed milling system with a predetermined allocation of grinding, sifting and purification capacity and set up. The saying addresses the need to properly tend to the equipment in place and properly manage machine setting and screen selection. The adage also justifies, in part, alternative mill flow designs and adaptations. However; it does not address “optimization.”

Near infrared focal plane array quantitative imaging has the power to provide both mass and qualitative information about intermediate processing streams. Its focus is on the material millers strive to recover in their process - endosperm. We need to improve the economic return on asset and wheat by taking an innovative approach to the concept of mill balance and optimization.

Jeff Gwirtz is chief executive officer of JAG Services Inc., a consulting company serving the milling and grain industry. He may be reach at jeff@jagsi.com.