Accelerating and automating sorptivity measurements in cementitious systems via computer vision

Cementitious materials form the backbone of modern infrastructure, making durability assessment a critical priority. Sorptivity, a key parameter governing concrete service life, influences deterioration mechanisms such as freeze-thaw damage, sulfate attack, and chloride-induced corrosion. However, traditional methods like ASTM C1585 are time-consuming and labor-intensive, highlighting the need for faster and automated alternatives. This PhD thesis introduces two new automated approaches—the Droplet Method and the Waterfront Method—leveraging computer vision and machine learning to improve sorptivity prediction across cement pastes, mortars, and concretes. Firstly, the Droplet Method was developed to estimate the 6-hr initial sorptivity rapidly by analyzing the wetting behavior of droplet dynamics on the scale of minutes to seconds. Applied to 63 paste systems with water-to-cement (w/c) ratios ranging from 0.4 to 0.8, this approach yielded strong correlations (adjusted R² ≥ 0.9) between the dynamics of droplets and initial sorptivity. In addition, to streamline contact angle measurements, we introduced a low-cost contact angle goniometer (~$200) integrated with a convolutional neural network trained on ~3,000 images that enhances measurement precision and reduces the standard deviation from 14.6° to 6.7°. Secondly, to predict initial and secondary sorptivity in pastes, mortars, and concretes, the Waterfront Method was developed using an EfficientNet-based vision model trained on ~6,000 images to segment wetted regions in real-time. This novel approach enabled continuous and automated absorption tracking across 1,440 measurements, achieving R² > 0.9 for sorptivity predictions. Finally, these two novel methods were applied to a series of concrete mixtures, revealing strong correlations (R² > 0.9) between initial sorptivity and electrical resistivity, secondary sorptivity and freeze-thaw performance. By accelerating and automating sorptivity measurements, we get one step closer to efficiently predicting long-term durability.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civil Engineering

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Kabir, Hossein

Contributors dc:contributor

• Garg, Nishant
• Popovics, John S
• Roesler, Jeffery R
• Olek, Jan

• Computer Vision
• Machine Learning
• Durability
• Sorptivity
• Droplet Method
• Waterfront Method

Statement dc:rights

• Copyright 2025 Hossein Kabir

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/129763

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F129763