Machine Learning Based Identification of Solvents from Post-Desiccation Patterns

Jesús Israel Morán-Cortés; Felipe Pacheco-Vázquez

cond-mat.softMar 8, 2026

Machine Learning Based Identification of Solvents from Post-Desiccation Patterns

Authors: Jesús Israel Morán-Cortés, Felipe Pacheco-Vázquez

Organizations: Instituto de Física, Benemérita Universidad Autónoma de Puebla, Apartado Postal J-48, Puebla 72570, Mexico.

Abstract

We introduce an optimized protocol of fracture pattern classification using an artificial neural network to identify the solvent involved in the desiccation cracking process of starch-liquid slurries, even after it has been completely evaporated. For this purpose, image analysis techniques were used to characterize patterns obtained from drying suspensions using single solvents (water, ethanol, acetone) and two-component solvents (water-ethanol mixtures at different concentrations). Frequency histograms were generated based on nine morphological features, taking into account their size, shape, geometry and orientational ordering. Subsequently, we used these histograms as input data into artificial neural network variants to determine the set of features that lead to the higher accuracy in solvent identification. We obtained an average accuracy of $96(\pm 1)\%$ considering all solvents in the analysis. The highest accuracy was obtained with sets of features that include the crack area distribution. The proposed protocol can help to determine the combination of features that optimize pattern recognition in other fields of science and engineering.

Figures & tables

Figure 1: (a) Experimental setup components: (1) digital camera, (2) digital thermometer-hygrometer (3) aluminum container with initial mass $m_{0}$ of a specific mixture, and (4) analytical balance. (b) Remanent mass $m$ of the mixtures as a function of time $t$ during the drying process, for three different solvents.

Figure 2: Fracture patterns observed in experimental samples of mixtures using solvents of a) pure liquids and b) water-ethanol mixtures with the indicated percentages of ethanol.

Ethanol content in water	0%	20%	40%	60%	80%	100%
$X$	$0.000$	$0.089$	$0.207$	$0.370$	$0.610$	$1.000$
$\gamma$ (mN/m)	$72.01$	$37.97$	$30.16$	$26.23$	$23.82$	$21.82$

Table 1: Ethanol-water binary solvents. Mole fraction concentration

X

of ethanol

(M=46.07)

in water (

M=18.02

) depending on the indicated percentage of ethanol volume in the mixture, where

M

is the molar mass. The corresponding surface tension

\gamma

25^{\circ}

C is indicated Vazquez et al. (1995) ; Poling et al. (2001) ; Cervantes-Álvarez et al. (2020) .

Figure 3: Image processing. (a) Picture of a fracture pattern. (b) Corresponding centroids (black dots) and (c) Voronoi diagram (black edges) from the real fracture pattern (red edges) after image processing.

Figure 4: Size-shape analysis. (A) Final patterns observed after the desiccation process of layers of cornstarch and the indicated solvent. (B) Corresponding processed images used to identify the cracks (red edges) and their centroids (black dots). In A and B the scale bar is $50\,\textup{mm}$ . (a-e) Histogram distribution of (1) pattern area $A_{\textup{N}}$ , (2) perimeter $P_{\textup{N}}$ , (3) crack area $A_{\textup{crN}}$ , (4) isoperimetric ratio $\lambda$ , and (5) eccentricity $\varepsilon$ , obtained using as solvent: (a) water, (b) ethanol, (c) acetone, and (d) $80/20\%$ and (e) $20/80\%$ water/ethanol. Colored lines represent the average value and the gray shaded area represents the standard deviation.

Figure 5: Orientational ordering analysis. (A-B) Voronoi diagrams (colored cells) and centroids (white dots) obtained from the actual patterns (black lines) depending on the solvent. Colors are used to indicate A) the number of neighbors $N_{B}$ and B) the orientational order (here $\psi_{5}$ is shown for visualization) according to the color bars. (1-4) Frequency histograms of 1) nearest neighbors, and different order parameters 2) $\psi_{4}$ , 3) $\psi_{5}$ and 4) $\psi_{6}$ , depending on the solvent indicated in the same row. Colored lines represent the mean values and the gray shaded areas the corresponding standard deviations.

Figure 6: Framework of the proposed protocol, including: (a) database construction, (b) machine learning implementation, and (c) model evaluation.

Figure 7: Classification accuracy. Performance of (a) the normalized isoperimetric ratio $\lambda$ in three MLP neural network configurations depicted in the inset: increasing, constant, and decreasing architectures as a function of the number of hidden layers (HL). (b) Performance of each feature in the decreasing architecture. Solid lines indicate mean values, and shaded areas represent the corresponding standard deviations.

Figure 8: Classification accuracy for some sets of features (with $N=60$ neurons), comparing (a) three classes corresponding to single solvents (water, ethanol, and acetone) and (b) five classes including the single solvents and binary solvents ( $20\%$ and $80\%$ ethanol content in water). For sets with the same number of elements, the worst, intermediate and best accuracies are indicated by blue, green and red bars, respectively. For comparison, green bars correspond to the same sets in both cases (a) and (b). The vertical black line corresponds to the accuracy using only $A_{\textup{cr}N}$ , which leads to the highest value obtained with an individual feature.

Image Classification

Image classification assigns semantic labels to images, enabling recognition of objects, scenes, conditions, and fine-grained categories.

See all 390 papers

Interpretable liquid crystal phase classification via two-by-two ordinal patterns
Mar 19, 2026Leonardo G. J. M. Voltarelli, Natalia Osiecka-Drewniak, Marcin Piwowarczyk +4Classification Materials Science
Departamento de Física, Universidade Estadual de Maringá, Maringá, PR 87020-900, Brazil · The Henryk Niewodniczański Institute of Nuclear Physics, Polish Academy of Sciences, E. Radzikowskiego 152, 31-342 Kraków, Poland · Departamento de Física, Universidade Tecnológica Federal do Paraná, Apucarana, PR 86812-460, Brazil +4
Knowledge-Driven XRD Phase Identification via Multi-View Retrieval and Explanation
Sep 25, 2026Doaa Mohamed, Markus StrickerMaterials Science
Materials Informatics and Data Science Interdisciplinary Centre for Advanced Materials Simulation (ICAMS) Ruhr University Bochum, Bochum, Germany
Solvating machine-learning interatomic potentials for solution-phase chemistry
Oct 12, 2026Anton Z. Ni, Joonho LeeMachine Learning Interatomic Potentials Molecular Property Prediction
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA

Mar 19, 2026J/K move · Enter open · S save

Interpretable liquid crystal phase classification via two-by-two ordinal patterns

Leonardo G. J. M. Voltarelli, Natalia Osiecka-Drewniak, Marcin Piwowarczyk, Ewa Juszynska-Galazka+3

Departamento de Física, Universidade Estadual de Maringá, Maringá, PR 87020-900, Brazil · The Henryk Niewodniczański Institute of Nuclear Physics, Polish Academy of Sciences, E. Radzikowskiego 152, 31-342 Kraków, Poland · Departamento de Física, Universidade Tecnológica Federal do Paraná, Apucarana, PR 86812-460, Brazil+4

Classification Materials Science Explainable Image Classification

Open paper AskPDF arXiv

Liquid crystal textures encode rich structural information, yet mapping these images to mesophase identity remains challenging because visually similar patterns can arise from distinct structures. Here we present a simple, interpretable representation that maps textures to a 75-dimensional frequency vector of two-by-two ordinal patterns, grouped into eleven symmetry-based types to characterize a large-scale dataset spanning seven mesophases. Combined with a simple machine learning classifier, this lightweight representation yields near-perfect phase recognition, including the difficult distinction between smectic A and smectic B mesophases. Our approach generalizes to unseen compounds and accurately distinguishes between phase identity and material origin. Unlike deep learning methods, each ordinal pattern is readily interpretable, and model explanations augmented with network visualizations of pattern interactions reveal the specific types and pairwise dependencies that drive each mesophase decision, providing compact, physically meaningful summaries of texture determinants. These results establish two-by-two ordinal patterns as an interpretable and scalable tool for liquid crystal image analysis, with potential applications to other complex patterned systems in materials science.

Abstract

Image Classification

Explore similar work

Interpretable liquid crystal phase classification via two-by-two ordinal patterns

Knowledge-Driven XRD Phase Identification via Multi-View Retrieval and Explanation

Solvating machine-learning interatomic potentials for solution-phase chemistry