Components of bean geometry

A coffee bean's geometry isn't just length or diameter. A bean can be described by size, flatness, major and minor axes, overall shape, and asymmetries arising from its natural structure. When fed into a grinding mechanism, these characteristics determine how the bean contacts the grinding surface and where stress concentrates. However, geometry shouldn't be turned into the sole explanation for post-grind PSD, since roasting has already significantly altered the bean's internal structure.

Geometry and the breaking mechanism

The breaking mechanism depends on how force is transmitted into the bean. A bean with an inhomogeneous structure can crack in different directions under compression, shear, or impact. After roasting, porosity and brittleness change, so the same outward shape can have different mechanical behavior. Research on texture changes caused by roasting shows beans lose strength and become more brittle as water decreases and tissue structure changes. So geometry visible to the eye can't fully represent breaking characteristics.

Evidence from temperature and structure

Evidence on bean temperature before grinding also reinforces this view. Uman and colleagues studied the effect of bean origin and temperature on grinding, showing the bean's state before entering the grinding mechanism can affect the outcome. This helps distinguish two concepts often lumped together: original geometry and the bean's physical state. If you only measure green-bean size and try to predict roasted-coffee PSD, the model will miss roasting-induced changes in density, moisture, structure, and brittleness.

Designing a geometry-PSD test

To test the geometry-PSD relationship, an experiment should measure geometry at the same level as the output it aims to explain. Samples can be classified by size or shape, then roasted under controlled conditions and ground on the same machine with the same settings. Output PSD is measured using a consistent method. If groups with different geometry show different distributions, you still need to check whether the difference genuinely comes from geometry or from accompanying density, moisture, roast level, or mechanical structure.

Limits of the inference

An important point is distinguishing green-bean geometry data from post-roast bean data. Roasting changes volume, density, and cell structure; so a relationship observed in green beans can't automatically apply to roasted beans. Models predicting physical properties during roasting show these variables change with thermal conditions. That's why, when building a PSD model, you should clearly define what state the input data was collected in.

So bean geometry has value as an explanatory variable, but not as a simple formula like “big beans make big grounds” or “round beans make narrow PSD.” Grinding behavior is the result of many layers of variables: geometry, structure, moisture, roast level, temperature, grinding mechanism, and operating conditions. A good analysis doesn't look for one single variable to explain PSD, but identifies which variables contribute meaningfully under specific conditions, then verifies with output data.

When analyzing geometry, you should also avoid reducing “size” to a single variable. Two beans of the same length can differ significantly in flatness or thickness, and this difference can change how they enter the grinding gap. If possible, describe geometry with multiple parameters and see which parameter actually relates to PSD. This approach better fits the multi-dimensional nature of the breaking mechanism than using one representative diameter.

A good experiment also needs a large enough sample to avoid a few unusual beans deciding the result. Defective beans, pre-broken beans, or beans with a distinctly different structure should be handled according to a predefined rule. Then the PSD of each group must be measured with the same method. If results differ, you can further check density or moisture to see whether geometry is a direct cause or just accompanies another physical characteristic.

If possible, geometry should also be measured after roasting as part of the experiment, to check whether roasting-induced deformation changes the original relationship. Comparing before and after roasting helps identify which variables stay stable and which change. This is also a way to avoid using green-bean data as a stand-in for the material the grinder actually receives.

In multi-variable models, geometry may be just a supporting variable. If, after adding density, moisture, and roast level to the model, geometry's contribution decreases, that doesn't make the geometry data useless; it shows geometry is traveling alongside other variables. A good conclusion should reflect that rather than insisting on treating one variable as the main cause just because it's easy to observe.

If geometry is used to classify raw material into groups, the classification rule needs to be defined in advance. You can use a defined size range or shape index rather than choosing groups after seeing PSD. Doing so beforehand reduces the risk of choosing a classification method that favors the result. Each group is then roasted and ground under the same conditions so the comparison is meaningful.

References

• Uman et al. (2016), Scientific Reports 6, 24483, DOI 10.1038/srep24483.
• Bustos-Vanegas et al. (2018), Developing predictive models for determining physical properties of coffee beans during the roasting process, Industrial Crops and Products 112, 839–845, DOI 10.1016/j.indcrop.2017.12.015.
• Pittia, Dalla Rosa & Lerici (2001), Textural Changes of Coffee Beans as Affected by Roasting Conditions, LWT 34, 168–175, DOI 10.1006/fstl.2000.0749.
• Crushing of roasted Arabica coffee beans (2014), Acta Universitatis Agriculturae et Silviculturae Mendelianae Brunensis.