Estimation of the degree of hydration of concrete through automated machine learning based microstructure analysis – A study on effect of image magnification

doi:10.1016/j.aei.2019.100975

Advanced Engineering Informatics

Volume 42, October 2019, 100975

https://doi.org/10.1016/j.aei.2019.100975 Get rights and content

Abstract

The scanning electron microscopy (SEM) images are commonly used to understand the microstructure of the concrete. With the advancements in the field of computer vision, many researchers have adopted the image processing technique for the microstructure analysis. Most of the previous methods are not adaptable, non-reproducible, semi-automated, and most importantly all these methods are highly influenced by image magnification. Therefore, to overcome these challenges, this paper presents a machine learning based image segmentation method for microstructure analysis and degree of hydration measurement using SEM images. In addition, the authors looked into the impact of magnification of SEM images on the model accuracy and classifier training for the degree of hydration measurement considering two scenarios. First, the image segmentation was performed using a classifier of specific magnification, and then a common classifier is trained using the image of different magnification. The results show that the Random Forest classifier algorithm is suitable for microstructure analysis using SEM images. Through the statistical analysis, it has been proved that there is no significant effect of magnification on model training and accuracy for the degree of hydration measurement. So, a single classifier can be used to process the images of different magnification of a specimen which reduces the effort of training and computational time. The proposed method can generate highly accurate and reliable results in a shorter time and lower cost. Moreover, the findings in this research can be useful for researchers to determine the optimum magnification required for the microstructure analysis.

Introduction

It is estimated that the concrete is consumed at a rate of 11 billion metric tons per year, and it is the most widely used construction material for buildings, bridges, roads, and dams because of the exceptional properties such as workability, durability, and strength [1]. The physical and mechanical properties of the concrete are governed by the microstructure of the concrete [2], and they can be modified by making suitable changes to the microstructure of the concrete. The microstructure of concrete is a complex system with a high degree of heterogeneity, composed of solid phases, pores, and water. At the simplest heterogeneity level, concrete consists of aggregates, hydrated cement paste, anhydrous cement and pores [3]. The quantitative microstructure characterization of concrete is useful to determine the relationship to macrostructure properties [4]. Understanding the microstructure and hydration of the cement will ensure the strength and durability of concrete, helps in recognizing and mitigating the stresses to prevent cracking, and assists in determining the curing conditions and construction practices. Therefore, it is essential to understand the relationship between microstructure and the properties of concrete because the inhomogeneity in the microstructure of the concrete can lead to severe effects on strength and other related physical and mechanical properties. Among these properties, the strength and durability of the concrete can be highly influenced by the cement hydration rate. The analysis and understanding of the microstructure of concrete, more specifically the degree of hydration, can help researchers and practitioners to improve the performance of the concrete, modify the bond strength and make durable concrete repairs. In addition, it can also play a vital role in determining the cause of failure or damage to infrastructure during concrete forensics studies.

The microstructure analysis of concrete has a wide range of applications; some of them include concrete forensics, bridge condition monitoring, etc. There exist various methods for concrete microstructure investigation such as light optical microscopy with associated digital image analysis, scanning electron microscopy (SEM), X-ray computed tomography with image processing [5]. With the advancement in the image processing techniques, the scanning electron microscopy (SEM) image analysis has been widely adopted for the microstructure characterization. This SEM image analysis technique usually involves specimen preparation, SEM image collection, and image segmentation. For an accurate quantitative estimation of properties such as the degree of hydration, porosity, or water-cement ratio, the selected image segmentation algorithm should be precise and reproducible [6]. Various researchers [7], [8], [9] presented different SEM image segmentation methods using edge detection, particle swam optimization, histogram intensity method, and etc. Some of the limitations in using these techniques are that these algorithms are not adaptable, non-reproducible, semi-automated, and most importantly rely on the specific magnification image. The effect of magnification on microstructure analysis is one of the significant challenges because researchers or practitioners capture SEM images at different magnification to better understand the sample, but the existing methods are limited on processing images with certain preset magnification.

Therefore, this study presents a machine learning based SEM image segmentation methodology for the concrete microstructure analysis, more specifically it focuses on the estimation of the degree of cement hydration through the proposed approach. The proposed machine learning based SEM image segmentation method adopts pixel-based image segmentation which generates highly accurate results for microscopy images. This method can produce consistent results provided if the model is trained on same features and precise training samples. Some most commonly used machine learning classifier algorithms were tested and evaluated in the experiment to identify the best fit for the proposed machine learning based concrete SEM image segmentation. This study also further investigates the effect of magnification on classifier training, and model accuracy for microstructure characterization and degree of hydration measurement considering two scenarios. In the first scenario, a classifier is trained using images of specific magnification. Whereas in the second scenario, a common classifier is trained using images of all the magnification. Understanding the effect of magnification on classifier performance helps researchers and practitioners to process the images of different magnification using single classifier which reduces the intensive manual effort and long computation time.

The rest of the paper is structured as follows. First, the literature review section reviews the literature on the significance of microstructure analysis and degree of hydration measurement, existing methods for microstructure analysis and machine learning based segmentation. Then, the methodology section illustrates the proposed machine learning based microstructure analysis using SEM images. Followed by the experimental validation section, in which it evaluates the performance of the proposed methodology. In the end, this study is concluded with a summary of all results and findings, limitations to this research, and recommendations for future research.

Section snippets

Current methods for concrete SEM image analysis

SEM image analysis and X-ray Diffraction are the most commonly used techniques to quantify microstructure phases of the concrete. Due to the presence of amorphous phases in the concrete, SEM image analysis technique is more accessible to apply compared to the X-ray diffraction method [10]. SEM images of concrete are segmented into various components such as aggregates, hydrated cement, anhydrous cement, and pores using image-processing techniques. The segmented images allow measurement of areas

Automated machine learning based microstructure analysis

The framework of the proposed machine learning based microstructure analysis is shown in Fig. 1. The first step of the methodology is the preparation of samples for SEM imaging. Multiple SEM images are then collected from each specimen. The collected SEM images are used as a training and testing data for machine learning classifier algorithms. The trained classifier is used for the segmentation of images. The segmented images are used for microstructure characterization and degree of hydration

Experiment setup

The SEM images of four specimens were used in this study to achieve the proposed objective. The SEM images of the specimens were obtained from a study of determining the crack-repair efficiency of self-healing concrete through microencapsulated calcium nitrate (microcapsule) [34] Table 2 provides the details of cement concentration, agitation rate, and average microcapsule size used to prepare the specimens. For the specimen preparation, Type I cement with no supplementary cementitious

Conclusion

Researchers or practitioners capture multiple SEM images of concrete at different magnifications to better understand the sample. The SEM imaging at different magnification involves high cost and time. The existing methods for determining the degree of hydration requires SEM images of specific magnification, where the other images are not utilized. Moreover, the existing methods for microstructure characterization and degree of hydration measurement are highly influenced by image magnification,

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (34)

K.L. Scrivener
Backscattered electron imaging of cementitious microstructures: understanding and quantification
Cement Concrete Compos.
(2004)
H. Wong et al.
Pore segmentation of cement-based materials from backscattered electron images
Cement Concrete Compos.
(2006)
H. Wong et al.
Determining the water–cement ratio, cement content, water content and degree of hydration of hardened cement paste: Method development and validation on paste samples
Cement Concrete Res.
(2009)
R. Yang et al.
Binary segmentation of aggregate in SEM image analysis of concrete
Cement Concrete Res.
(2001)
B. Gaël et al.
Determination of the proportion of anhydrous cement using SEM image analysis
Constr. Build. Mater.
(2016)
P.K. Mehta, P.J. Monteiro, Concrete Microstructure, Properties and Materials,...
D.M. Roy, G. Idorn, Concrete microstructure,...
K. Scrivener et al.
A Practical Guide to Microstructural Analysis of Cementitious Materials
(2016)
A.A. Hilal
Microstructure of Concrete, High Performance Concrete Technology and Applications
(2016)
J.H. Lee et al.
An effective image segmentation technique for the SEM image

H. Feng et al.

Segmentation-assisted edge extraction algorithms for SEM images

S. Feng et al.

SEM-backscattered electron imaging and image processing for evaluation of unhydrated cement volume fraction in slag blended Portland cement pastes

J. Wuhan Univ. Technol.-Mater. Sci. Ed.

(2013)

P. Jaglan et al.

A comparative analysis of various image segmentation techniques

H. Son et al.

Automated color model–based concrete detection in construction-site images by using machine learning

ASCE J. Comput. Civ. Eng.

(2011)

P. Prasanna et al.

Automated crack detection on concrete bridges

IEEE Trans. Automat. Sci. Eng.

(2016)

N.D. Hoang

Image processing-based recognition of wall defects using machine learning approaches and steerable filters

Comput. Intel. Neurosci.

(2018)

W.R.L.d. Silva, D.S.d. Lucena, Concrete cracks detection based on deep learning image classification, Multidisciplinary...

Cited by (36)

Image enhancement and microstructure characterization of energy dispersive X-ray spectroscopy images of blended cement pastes
2024, Expert Systems with Applications
Energy dispersive X-ray spectroscopy (EDS) image analysis has been widely used to study the microstructure of cementitious materials, whereas in some cases the efficiency of the characterization suffered from the extremely long time for image acquisition. In this study, we proposed an EDS image analysis methodology including image enhancement, phase extraction, and characterization of reaction products. The guided filter and Logarithm correction were combined and employed to improve the resolvability of low-quality EDS images. Gaussian Mixture Model (GMM) algorithm was applied to segment the unreacted phases and two different groups of reaction products from the enhanced EDS images. Moreover, the spatial distribution and chemical composition of reaction products were characterized using the lineal-path function and simple lineal iterative clustering (SLIC) algorithm. Results showed that low-quality EDS images after enhancing could achieve the comparable performance with high-quality EDS images in terms of the phase proportions. This study offers a practical and effective methodology for enhancing and analyzing the low-quality EDS images, giving insights into the phase assemblage and microstructure evolution of cementitious materials.
Benchmarking AutoML solutions for concrete strength prediction: Reliability, uncertainty, and dilemma
2024, Construction and Building Materials
Building precise machine learning and deep learning models has traditionally required a combination of mathematical skills and hands-on experience to meticulously adjust hyperparameters that significantly impact the learning process. As datasets continue to expand across various engineering domains, researchers increasingly turn to machine learning methods to uncover hidden insights that may elude classic regression techniques. This surge in adoption raises concerns about the adequacy of resultant meta-models and the interpretation of findings. In response to these challenges, automated machine learning (AutoML) emerges as a promising solution, aiming to construct machine learning models with minimal intervention or guidance from human experts. This paper benchmarks AutoML solutions by providing an overview of their principles and applying them to predict the most important mechanical properties of different concrete datasets, i.e., compressive strength. Nine datasets from various concrete types, sample sizes, and features are utilized, with a detailed discussion on the benchmark dataset from high-performance concrete, applying best practices to the other eight datasets. For each case, the importance of hyperparameter tuning is discussed, alongside the ensemble and stacking models. Tree-based models are employed for each dataset to develop SHAP plots, interpret results, and understand the contribution of each component in the mix design to the overall strength of the concrete. This paper further explores three unique aspects of benchmarking AutoML in material science: (1) “reliability” by contrasting the benchmark dataset’s error metric with literature collected over the past 20 years, (2) “uncertainty” by quantifying the variability in the mean and standard deviation of the error metric from different datasets and its correlation with the sample-to-feature ratio, and (3) “dilemma” by discussing the shortcomings of AutoML in specific concrete datasets.
Deep learning based automated fracture identification in material characterization experiments
2024, Advanced Engineering Informatics
In the current work, the automated fracture identification in material testing experiments is investigated through deep learning convolutional neural network (CNN) techniques. Three widely employed material characterization experiments are considered, namely uniaxial tensile, punch, and shear experiments. Initially, the surface crack identification performance of well-known, minorly modified CNN architectures is investigated in the context of transfer learning. In particular, the transfer learning performance of VGG, ResNet, and Inception CNN architectures is analyzed. It is found that moderate-size filters and deeper rather than wider network architectures yield significantly higher validation accuracies in all cases, with validation scores that approach and exceed 95% accuracies. Moreover, low computational cost models that can run on minimal storage and computing power devices and deliver superior performance for all experimental testing setups are identified through greedy network searches. Thereupon, CNN deep learning architectures that allow for validation accuracies above 99% are reported, delimiting low-performing model counterparts. Moreover, CNN-based deep learning «all-in-one» material fracture models are elaborated, furnishing automated, computer vision-based fracture characterization techniques. Furthermore, feature map insights are provided, highlighting the salient-feature identification potential of the networks, while the automated crack localization and profile characterization are demonstrated through Local Interpretable Model-agnostic Explanations (LIME) machine learning (ML) methods and image subset creation techniques. The results are expected to serve as a basis for material testing automation with the support of deep learning techniques, providing reliable analysis methods in labor-intensive engineering tasks.
Automated pixel-level crack detection and quantification using deep convolutional neural networks for structural condition assessment
2024, Structures
Crack detection is a crucial task in assessing the condition of concrete structures. Herein, a novel deep learning method based on convolutional neural networks, referred to as R-FPANet, is proposed for crack detection. The R-FPANet performs automatic segmentation and quantification of crack morphology at the pixel level. In this methodology, the modularization concept based on the following three modules is adopted: ResNet-50 is chosen as the backbone to extract features from images, the Feature Pyramid Network with Dense Block is integrated to promote the fusion of both shallow and deep features as well as enhance feature reuse, and self-attention mechanisms such as Channel Attention Module and Position Attention Module are introduced to strengthen the dependency between features. Based on the crack segmentation results, a suitably established framework is developed for quantitative analysis of the major geometric parameters, including crack area, crack length, crack mean width and crack max-width at the pixel level. To verify the effectiveness of the proposed method, a large-scale concrete crack image dataset was produced and carefully labeled at the pixel level and then utilized to train the model. Finally, our experiments reveal that the proposed approach achieves an Intersection over Union of 83.07%, further indicating that the segmentation performance of the proposed method is better than the state-of-the-art models and also confirming that the crack quantification results are close to reality. Overall, the proposed method performs well, contributing to crack detection and quantification with great potential for practical use.
Identification of experienced temperature in mortar and concrete using microstructural image and deep learning
2023, Construction and Building Materials
This study presents a framework for identifying the temperature experienced by fire-damaged mortar and concrete using scanning electron microscopy (SEM) images and deep learning. A dataset of 16,484 SEM images of cement paste mixes with varying water-to-binder ratios and pozzolanic materials, exposed to temperatures ranging from 200 to 800 °C was established. Then a deep-learning model based on a convolutional neural network (CNN) for SEM image classification was trained, achieving a high accuracy above 98 %. To test the method's generalizability, cement paste mixture with a different water-to-binder ratio, mortar mixture with sand inclusion, and concrete mixture with coarse aggregates were prepared and exposed to different temperatures. The predicted temperatures deviated from the target temperatures within 8.6 %. Finally, visualization of the deep learning model was used to identify the critical features that influenced the prediction. The outer hydration products with smaller pores had a higher influence on samples before heating, whereas porous dehydrated products were more influential in samples exposed to high temperatures.
Advances in imaging, scattering, spectroscopy, and machine learning-aided approaches for multiscale characterization of cementitious systems
2023, Cement and Concrete Research
Recent progress in methods used in the multiscale characterization of cementitious systems is reviewed, focusing on advances in imaging, scattering, and spectroscopy. The review includes relevant applications and developments in machine learning and other data analytics approaches to enhance characterization. Developments in imaging using light and electron microscopy as well as x-ray (i.e., from synchrotron) methods are summarized. Updates include scanning electron microscopy (SEM), transmission electron microscopy (TEM), tomography, and holography. A critical overview of spectroscopy (e.g., MAS NMR, Raman) and scattering (e.g., neutron, x-ray, synchrotron x-ray) methods is provided, and the intersection of these with imaging is developed (e.g., Raman imaging). Additionally, the paper summarizes recent developments in and implementations of state-of-the-art machine-learning algorithms and data analytics methods for automated, systematic, and/or quantitative analyses of image data sets.
The review considers but is not limited to the application of these methods for investigating the hydration and microstructure development of cement phases, low-carbon-footprint cements (e.g., limestone calcined clay cements, LC3), environmental interactions (e.g., ASR) and model systems. Thus, the present work provides a critical presentation of advances in characterization methods that link together the composition and multiscale structure of cementitious materials.

View all citing articles on Scopus

View full text

Full length articleEstimation of the degree of hydration of concrete through automated machine learning based microstructure analysis – A study on effect of image magnification