Application of automatic image analysis to evaluate the anisotropy of autoclaved aerated concrete for moisture transport

Autoclaved aerated concrete (AAC) is an artificial building material commonly applied in energy efficient buildings. Blocks of AAC are characterized by relatively homogenous distribution of pores and their dimensions. However, during manufacturing there occurs irregular pores positioning which may influence water transport processes. This phenomenon is called anisotropy which could be evaluated using standard gravimetric or electric methods of moisture detection. In this article we propose a method of level of anisotropy evaluation using image analysis. During the research a set of AAC samples was prepared and examined using time domain reflectometry (TDR) method for capillary uptake phenomenon and in parallel the cross-sections visual analyses were conducted. Both techniques confirmed the anisotropic features of the tested material.


Introduction
Ensuring the efficient use of energy in buildings is an important issue in today's construction industry.Materials with good thermal properties include, for instance, clay brick or aerated concrete.Currently, autoclaved aerated concrete (AAC) constitutes a popular and widely used construction material.In addition to thermal properties, a factor considered in terms of building operation is the moisture content of the partitions.Hence, it is important to study the absorption capacity or saturation of commonly used materials.Aerated concrete is characterized by relatively homogenous general distribution of pores and their dimensions however occurs irregular pores positioning, which is due to its manufacturing process [1].This can affect the rate of water transport.Because of the porosity configuration of aerated concrete, its mechanical and thermal anisotropy is studied [2][3][4][5].
There are different methods for testing moisture content.The standard method is the gravimetric method, which involves weighing a sample before and after drying.However, this method requires taking a sample, which is invasive.Another group of methods are the indirect techniques of moisture detection, that enable to measure physical values that are dependent on moisture.Among them the electric methods can be mentioned, mainly resistance and dielectric ones.The first method performance is based on measuring the resistance of flowing current while the second is based on measuring the apparent permittivity value.
A property that can describe the anisotropy of water transport is the distribution and shape of pores.It is likely that the rate of water rise varies according to the direction of growth [3].The issue of porosity is taken up to evaluate porous surfaces [11], soil [12], thermally sprayed coatings [13], sintering [14], and clastic rocks [15].
Furthermore, in this type of research, it is possible to link the characteristics of pores to the physical properties of materials, such as frost resistance tests [16,17] or compressive strength of cement mortars [18].
In this paper, the use of image analysis to evaluate water rise was proposed.Images of aerated concrete specimens are analysed to extract pore features that affect the rate of rise.These may include circumference, length, maximum and minimum Feret's diameter, etc. [19].

Materials and Methods
The experiment was conducted on cuboid blocks of aerated concrete with a density of 400 kg/m 3 .It was assumed that due material curing processes AAC is anisotropic, and the shape of pores is different for parallel and perpendicular direction and thus moisture migration processes may be different.For that aim the samples were cut in two different directions.One type of sample cutting direction was according to the autoclaved concrete curing direction and the second type of samples slicing direction was perpendicular to the curing process direction.Pictures of the samples were taken from a fixed distance, with the same camera settings.
For the laboratory experiment the samples were placed in water at a depth of 1 cm.Moisture migration process was monitored using the TDR equipment (LOM multimeter, FP/mts probes; ETest, Lublin, Poland).FP/mts probes were placed at two heights: 5 and 10 cm above the water surface.The dimensions of the samples were the following: 24 cm × 24 cm × 24 cm.The experiment was conducted in laboratory conditions in constant temperature 20ºC ±1ºC and relative air humidity 50% ± 10%.Duration of the TDR experiment depended on water uptake process and was established for 7 days.
Fiji software was used for image processing and analysis.It is a distribution of the open-source ImageJ software with many plug-ins included.The author of ImageJ is Wayne Rasband while the Fiji project is created by many people including Johannes Schindelin, Ignacio Arganda-Carreras, Albert Cardona, Mark Longair, Benjamin Schmid.It is an image processing program written in Java [20].
Digital photographs were taken of the sample positioned parallel and perpendicular to the direction of growth.To ensure comparability of results in both directions, squares of 2000 x 2000 pixels were cut out seen on figure 1.The initial operations performed on the image involved conversion to an 8-bit grayscale.Finally, both images were thresholded using Huang's method [21].The result of this transformation was images on figure 2.
• AR -ratio of the length of the axis of the ellipse fitted to the shape of the selection.
where  $-(.denotes the major axis of the ellipse fitted to the selection.• Feret's diameter -the largest distance between any two points along the border of the selection.• Feret's Angle -(0-180 degrees) is the angle between the Feret's diameter and the line parallel to the x-axis of the image.• MinFeret -the smallest distance between any two points along the selection boundary.
• Solidity -the ratio of the area of the selection to the convex hull.

Results
A comparative analysis was carried out on a sample of 2862 pores for the sample aligned with the growth direction and 3334 for the perpendicular direction.Initially, the distribution of individual variables was examined.In terms of shape, the distributions are similar.Slight differences can be observed in the intensity of the observations depending on the direction of the block.
To test the significance of the differences, comparative tests were conducted.Since most distributions are not normal, the Wilcoxon-Mann-Whitney test [22] will be used for comparison.With similar shapes of the distributions, it verifies the hypothesis that the medians of the two groups are equal.The choice of such statistic was dictated by the asymmetry of the distributions.As a result, we obtained that only Circ, AR and Round are not significantly different from all measures described in materials and methods chapter.The others differ significantly as can be seen in figure 3 and figure 4.  A: cross-section along curing direction, B: cross-section perpendicular to curing direction.
In figure 5 here are visible moisture changes measured using the TDR equipment.a) b) Figure 5. Uptake process in two types of the samples a) parallel to curing process, b) perpendicular to curing process It can be noticed from the diagrams that in case of the sensors close (5 cm) to water surface (blue line -probe 0 and probe 2) the differences are not significant for both types of samples.Moisture readouts are similar, water appears at the same time and reaches maximal values (0.3 cm 3 /cm 3 ) within the first day of experiment.On the other hand it is clearly visible that the process of capillary uptake is different for both types of samples when observing the readouts of probes located 10 cm above water surface (orange line).In case of the samples (a) on probe 1 located at the level of 10 cm above the water surface, moisture starts to appear during the first day since beginning of experiment and the readouts are stable since the third day.In case of samples (b) moisture readouts at probe 3 begin visible during the third day of experiment and stable state doesn't appear even after 7 days.
Sample (a) is characterized by a larger surface area and perimeter of pores arranged in the uptake direction, longer major and minor axes of the fitted ellipse as well as longer Feret I MinFeret diameters.The samples also differ in the angle between the major axis and the x-axis, as well as the angle between the Feret diameter and the x-axis.Significant differences in FeretAngle and Angle metrics may indicate differences in pore alignment.Comparing this with the results from the TDR, it can be seen that sample (b) having smaller pore sizes although it quickly saturated to the height of the lower probe the moisture in 5 cm height is lower than in sample (a) after the same time.

Summary and conclusions
The study focused on evaluating the anisotropic characteristics of autoclaved aerated concrete (AAC) and its impact on water transport processes.The authors applied image analysis and time domain reflectometry (TDR) to analyse the pore distribution and moisture uptake of AAC samples.By conducting laboratory experiments on AAC cuboid blocks, the researchers compared samples cut in two different directions: one aligned with the curing process and the other perpendicular to it.Digital images of the samples were captured and processed using Fiji software for pore feature extraction.The results showed that the shape distributions of the pores were similar between the two sample types, but slight variations in intensity were observed depending on the sample direction.Comparative tests using the Wilcoxon-Mann-Whitney test revealed significant differences in various pore parameters, indicating differences in pore alignment between the samples.In addition, the moisture uptake process was monitored using TDR equipment.The measurements demonstrated that the sample aligned with the growth direction exhibited faster and more stable moisture absorption compared to the sample perpendicular to the growth direction.The findings of the study emphasized the importance of considering anisotropy in AAC's water transport properties.The shape and distribution of pores are related to the rate and uniformity of moisture absorption.The research highlighted the potential of image analysis for evaluating anisotropy and provided insights into the material's performance in terms of moisture management.Understanding the anisotropic behavior of AAC can contribute to improving the design and performance of energy-efficient buildings.It enables better moisture management, enhances building efficiency, and ensures the effective use of energy resources.

Figure 1 .
Figure 1.Images cropped to 2000 x 2000 px.On the left, cross-section along curing direction, on the right cross-section perpendicular to curing direction.

Figure 3 .
Figure 3.The boxplot of Area variable with result of the test of equality of medians.A: cross-section along curing direction, B: perpendicular to curing direction.

Figure 4 .
Figure 4.The boxplot of FeretAngle variable in pixels with result of the test of equality of medians.A: cross-section along curing direction, B: cross-section perpendicular to curing direction.