Gene Expression Programming Model Research Articles

Utilization of GEP and ANN for predicting the net-zero compressive strength of fly ash concrete toward carbon neutrality infrastructure regime

Abstract The present infrastructure regime being promoted by the United Nations Sustainable Development Goals is such that by the year 2050, the use of cement in the production of concrete and its use in the general construction activities as to reduce carbon emissions to zero must be replaced with net-zero construction materials. These cement replacement materials should be pozzolanic enough to either partially or totally replace the conventional cement and reduce its carbon footprint. The current study adopts two machine learning techniques: gene expression programming (GEP) and artificial neural network (ANN) to determine the 56 days and 180 days of net-zero compressive strength of fly ash concrete. The study effectively depicts how machine learning techniques can be used for the prediction of long- and short-term compressive strength of fly ash concrete toward a carbon neutrality infrastructure regime. The dataset has been compiled by various researchers, and the input parameters include cement, fine aggregate, coarse aggregate, fly ash, water, and water/binder ratio. And the 56 days and 180 days compressive strength (fck) values are the targeted output values. In order to determine a better model, both GEP and ANN were assessed based on the values of the correlation coefficient and crosschecked by other statistical parameters. Both models performed well; however, GEP outweighs the ANN model in estimating the fck at 56 days and 180 days. Moreover, the GEP model generated a simplified equation for foreseeing the value of fck for different ages of net-zero fly ash concrete.

Experimental-Correlative Framework to Evaluate Critical State Deformability Parameters and Free Swelling Index of Cohesive Soils Using Gene Expression Programming

Expansive soils stimulate serious problems for infrastructures and the near surface facilities due to volume change patterns it experiences during the wet and dry seasons. This research study is intended to implement Gene Expression Programming (GEP) scheme and the Multiple Linear Regression (MLR) method in order to assess the swelling aspects and deformability parameters of moderately expansive soils based on a comprehensive exploration and testing program for the Northeastern areas of Zarqa Governate in Jordan. Representative soil samples have been collected from twenty different locations in the area. The soil has been tested for free swelling, matric suction, permeability, consolidation, water content, hydrometer, and soil index properties. The free swelling index and critical state deformability parameters, obtained from the consolidation test, are the primary focus of this research. Several models that account for different situations based on opportunities of the data availability using an extensive GEP modeling scheme have been successfully developed. MLP method has been managed to develop one reasonable formula with less efficiency than the ones schemed by GEP procedure. The successful models have been validated using test results and error analysis procedures. The models have been ordered according to their according to a proposed ranking procedure. GEP method has showed higher performance for the given conditions.

Comparing machine-learning-based black box techniques and white box models to predict rainfall-runoff in a northern area of Iraq, the Little Khabur River.

The rainfall-runoff process is one of the most complex hydrological phenomena. Estimating runoff in the basin is one of the main conditions for planning and optimal use of rainfall. Using machine learning models in various sciences to investigate phenomena for which statistical information is available is a helpful tool. This study investigates and compares the abilities of HEC-HMS and TOPMODEL as white box models and adaptive neural fuzzy inference system (ANFIS) and gene expression programming (GEP) as black box models in rainfall-runoff simulation using 5-year statistical data. Using the inputs of rainfall and temperature of the previous day and discharge in the steps of the previous 2 days reduced the prediction error of both models. Examining the role of different parameters in improving the accuracy of simulations showed that the temperature as an effective parameter in cold months reduces the amount of prediction error. A comparison of R2, RMSE, and MBE showed that black box models are more effective forecasting tools. Among the black box models, the ANFIS model with R2 = 0.82 has performed better than the GEP model with R2 = 0.76. For white box models, the HEC-HMS and TOPMODEL had R2 equal to 0.3 and 0.25, respectively.

Indirect Estimation of Swelling Pressure of Expansive Soil: GEP versus MEP Modelling

In this article, detailed trials were undertaken to study the variation in genetic parameters in order to formulate more robust predictive models using gene expression programming (GEP) and multigene expression programming (MEP) for computing the swelling pressure of expansive soils (Ps-ES). A total of 200 datasets with ten input parameters (i.e., clay fraction CF, liquid limitwL, plastic limitwP, plasticity index IP, specific gravity Gs, swell percent Sp, sand content, silt content, maximum dry density ρdmax, and optimum water contentwopt) and one output variable, i.e., Ps-ES are collected from the literature, which comprises 120 internationally publications. The effect of input parameters in contributing to Ps-ES has been validated using Pearson correlation (r), sensitivity analysis (SA), as well as a parametric study. The results reveal that the GP-based techniques correctly characterize the swelling characteristics of the ES, thus leading to reasonable prediction performance; however, the MEP model yielded relatively better performance. Also, the proposed predictive models were compared with widely used AI models (ANN, ANFIS, RF, GB-T, DT, and SVM). The ANN performed relatively better; however, it is recommended to use the GEP and MEP due to the blackbox nature of the ANN. Other models exhibited inferior performance. The SA revealed different importance by the GEP and MEP models, however, its confirmed that the maximum dry density and optimum moisture content significantly affect the Ps-ES. The variation in Ps-ES with changes in input attributes is further corroborated from literature. Hence, it is recommended that the proposed GEP and MEP models can be deployed for computing the Ps-ES which efficiently lessens the laborious and time-consuming testing.

Application of ordinal logistic and gene expression programming methods to predict the collapse sensitivity classes of loess soils, a case study: Golestan Province, northeastern Iran

The current study assesses the collapse sensitivity classes of loess soils using gene expression programming (GEP) and ordinal logistic regression (OLR). The crucial variable to forecast the possible development of loess caves in the Golestan Province (northeast of Iran) is the collapse sensitivity factor (Is). A database of 62 records, including the mechanical and physical characteristics of soils, was used. Oedometer tests were used to estimate the parameters of the collapse coefficient, the time needed for 90% settlement (T90%), and collapse sensitivity. The database includes 10 inputs (grain size, porosity, initial water content, precipitation, climatic data, liquid limit, calcium carbonate content, vegetation, and degree of soil saturation) and one output (collapse sensitivity classes). This is a complicated approach due to the complexity of setting up and performing such kinds of tests in the laboratory. The likelihood of soil classification ranks as severe, moderately severe, moderate, and small sensitivity was inspected using OLR and GEP. This study demonstrated that the OLR approach could effectively differentiate among more than 70% of distinct groups. Furthermore, experimental data reported from Semnan, Sarakhs, and Mashhad also attests to the accuracy of the OLR model. The sensitivity analysis indicated that silt fraction imparts the maximum effect on the collapse sensitivity classes. The trial-and-error method was used to determine the configurations of the GEP model prior to developing an ideal model. The performance of the GEP model to estimate the collapse sensitivity categories in a trustworthy, strong, and useful way is well documented by comparison between the results of the GEP and the experimental findings, which are affordable.

An Application of Machine Learning to Estimate and Evaluate the Energy Consumption in an Office Room

There are no exact criteria for the architecture of openings and windows in office buildings in order to optimize energy consumption. Due to the physical limitations of this renewable energy source and the lack of conscious control over its capabilities, the amount of light entering offices and the role of daylight as a source of energy are determined by how they are constructed. In this study, the standard room dimensions, which are suitable for three to five employees, are compared to computer simulations. DesignBuilder and EnergyPlus are utilized to simulate the office’s lighting and energy consumption. This study presents a new method for estimating conventional energy consumption based on gene expression programming (GEP). A gravitational search algorithm (GSA) is implemented in order to optimize the model results. Using input and output data collected from a simulation of conventional energy use, the physical law underlying the problem and the relationship between inputs and outputs are identified. This method has the advantages of being quick and accurate, with no simulation required. Based on effective input parameters and sensitivity analysis, four models are evaluated. These models are used to evaluate the performance of the trained network based on statistical indicators. Among all the GEP models tested in this study, the one with the lowest MAE (0.1812) and RMSE (0.09146) and the highest correlation coefficient (0.90825) is found to be the most accurate.

Plastic concrete mechanical properties prediction based on experimental data

The industrial revolution brought environmental degradation to light. Concrete and plastic degrade the ecosystem and cause unsustainable development. Academic and industrial sectors are interested in lowering carbon emissions associated with concrete. Meanwhile, global sand scarcity worries environmentalists. To reach sustainable development goals, cement and fine aggregate must be substituted with other abundant waste/natural materials. This study aimed to develop a green concrete by utilizing plastic waste and creating modelling tool for predicting the mechanical properties of plastic concrete. Different composition of silica fume and superplasticizers substituted fine aggregate and cement in both irradiated (treated) and regular (untreated) plastic concrete. Compressive strength (fc’) and split tensile strength (fst) of the resulting concrete were studied. Moreover, from literature data, 320 data points each for fc' and fst were used to train gene expression programming (GEP) models. Models’ accuracy was evaluated employing various statistical measures. Regular plastic waste concrete has demonstrated a lower fc’ and exhibited anomalous behavior for fst. While irradiated plastic waste concrete has demonstrated improved mechanical characteristics, comparatively. Correlation coefficients using GEP models for fc’ and fst were found to be 0.92 and 0.88, respectively. Furthermore, sensitivity analysis revealed that plastic was the most significant in the GEP model’s development. K fold validation was employed to prevent over-fitting of the models. GEP provides an empirical expression for each outcome to predict future database features. This research improves green concrete's long-term sustainability by reducing carbon emissions and alleviating fine aggregate scarcity.

Turbulent forced convective flow in a conical diffuser: Hybrid and single nanofluids

Turbulent forced convective flow of hybrid and single nanofluids in a conical diffuser is investigated numerically. Simulations are conducted for various Reynolds (Re=10000-70000) and different concentrations (ϕ=0-1.5 vol%) at equal ratio of TiO2:SiO2. The impact of using theoretical and experimental correlations for dynamic viscosity and thermal conductivity on turbulent forced convection of TiO2 showed that the mean Nusselt (Nu) number is considerably reduced with the use of the experimental model. However, when the theoretical model is used, Nu varies insignificantly. Addition of TiO2 nanoparticles decreases the heat transfer inside the diffuser, whereas addition of TiO2–SiO2 nanoparticles either enhances or decreases the heat transfer rate. Compared to the pure fluid, hybrid nanofluids show a maximum enhancement of 5% and a maximum decrease of 9.7% at ϕ= 0.5 vol% and ϕ= 0.5 vol% at Re=10000, respectively. However, TiO2 nanofluids show a maximum decrease of 19% at ϕ=1.5 vol% and Re= 30000. As Re increases, the deviations between TiO2 and SiO2-TiO2 nanofluids diminish. Moreover, the Gene Expression Programming model can accurately evaluate Nu versus Re number and nanoparticle concentration.

Prediction of blast-induced air overpressure using a hybrid machine learning model and gene expression programming (GEP): A case study from an iron ore mine

<abstract> <p>Mine blasting can have a destructive effect on the environment. Among these effects, air overpressure (AOp) is a major concern. Therefore, a careful assessment of the AOp intensity should be conducted before any blasting operation in order to minimize the associated environmental detriment. Several empirical models have been established to predict and control AOp. However, the current empirical methods have many limitations, including low accuracy, poor generalizability, consideration only of linear relationships among influencing parameters, and investigation of only a few influencing parameters. Thus, the current research presents a hybrid model which combines an extreme gradient boosting algorithm (XGB) with grey wolf optimization (GWO) for accurately predicting AOp. Furthermore, an empirical model and gene expression programming (GEP) were used to assess the validity of the hybrid model (XGB-GWO). An analysis of 66 blastings with their corresponding AOp values and influential parameters was conducted to achieve the goals of this research. The efficiency of AOp prediction methods was evaluated in terms of mean absolute error (MAE), coefficient of determination (R<sup>2</sup>), and root mean square error (RMSE). Based on the calculations, the XGB-GWO model has performed as well as the empirical and GEP models. Next, the most significant parameters for predicting AOp were determined using a sensitivity analysis. Based on the analysis results, stemming length and rock quality designation (RQD) were identified as two variables with the greatest influence. This study showed that the proposed XGB-GWO method was robust and applicable for predicting AOp driven by blasting operations.</p> </abstract>

Statistical Analysis Approaches in Scour Depth of Bridge Piers

A local scour is the removal of bed material from around the pier of the bridge. This bed removal is considered a big problem and is of great concern for hydraulic engineers. They should find economic solutions for this problem. The exaggerated local scour around bridge piers leads to many problems for the whole bridge structure, such as stability problems that may lead to the bridge's destruction. This paper aims to verify the scour depth around different shapes of uniform bridge piers for different flow conditions than those done by previous researchers using different prediction models. Where the consistency of previous experimental investigations is verified by multiple nonlinear regression analysis (MNLR), Gene Expression Programming (GEP) and Artificial Neural Network (ANN) models. In the comparison of values that were measured and predicted by the four models (CFD, MNLR, ANN, and Gene), it is seen that the ANN model has the ability to predict the Ys/b values higher than other models used in relation to the measured values. This makes the ANN model superior in predicting the Ys/b value over the other used models, followed by the Gene model. In comparison, the values of the R2and RMSE for the four models that were used in this study, for the Ys/b model using the ANN had a value of 0.9978 and 0.0147, respectively, while those for the Ys/b model using the Gene model were 0.9800 and 0.0375, respectively. Doi: 10.28991/CEJ-2023-09-01-011 Full Text: PDF

Optimal Location to Use Solar Energy in an Urban Situation

This study conducted in Lima, Peru, a combination of spatial decision making system and machine learning was utilized to identify potential solar power plant construction sites within the city. Sundial measurements of solar radiation, precipitation, temperature, and altitude were collected for the study. Gene Expression Programming (GEP), which is based on the evolution of intelligent models, and Artificial Neural Networks (ANN) were both utilized in this investigation, and the results obtained from each were compared. Eighty percent of the data was utilized during the training phase, while the remaining twenty percent was utilized during the testing phase. On the basis of the findings, it was determined that the GEP is the most suitable network for predicting the location. The test state’s Nash-Sutcliffe efficiency (NSE) was 0.90, and its root-mean-square error (RMSE) was 0.04. Following the generation of the final map based on the results of the GEP model, it was determined that 9.2% of the province’s study area is suitable for the construction of photovoltaic solar power plants, while 53.5% is acceptable and 37.3% is unsuitable. The ANN model reveals that only 1.7% of the study area is suitable for the construction of photovoltaic solar power plants, while 66.8% is acceptable and 31.5% is unsuitable.

Dalgacık-Gen İfade Programlama ile Meteorolojik Kuraklık Tahmini: Çanakkale Örneği

The differences in duration and amount of precipitation significantly affect the drought due to global climate change in the last decades. Therefore, drought is one of the parameters to be considered for sustainable water resources studies. In this study, firstly, by using historical precipitation records between the years 1975-2010, the drought indices of 3-, 6-, 9- and 12- months of Çanakkale, Bozcaada, and Gökçeada stations were determined with the standardized precipitation index (SPI). Then, gene expression programming (GEP) models were developed in which the drought values of Bozcaada and Gökçeada stations were selected as input parameters for the drought prediction of Çanakkale province. In addition, W-GEP models were developed using a sub-series of the same inputs produced with wavelet transform (W). Examining the developed models, the determination coefficients (R2) for the 6-, 9- and 12-months periods were generally higher than 0.80 for GEP and W-GEP models. In contrast, the R2 value for the 3- month period was approximately 0.657 and 0.704, respectively. The highest R2 value was determined as 0.868 for the W-GEP model during the 6- month period. As a result, the GEP and W-GEP approaches were found to be successful in the estimation of drought.

Development of the New Prediction Models for the Compressive Strength of Nanomodified Concrete Using Novel Machine Learning Techniques

Concrete is a heterogeneous material that is extensively used as a construction material. However, to improve the toughness and mechanical properties of concrete, various ingredients (fillers) have been added in the past. The addition of nanomaterials for the improvement of the aforementioned properties has attracted many researchers worldwide. The high surface area, high reactivity, and finer size of various nanomaterials have made them preferable for the enhancement of durability, as well as compressive and flexural strength. The aim of the current research is focused on the estimation of compressive strength for the concrete modified with various nanomaterials using two machine learning techniques, namely decision tree technique (DTT) and random forest technique (RFT), and comparison with existing models. The database is collected for different percentages of four major widely used nanomaterials in concrete, i.e., carbon nanotubes, nano silica, nano clay, and nano alumina. The other four input variables used for the calibration of the models are: cement content (CC); water–cement ratio (W/C); fine aggregate, i.e., sand (FA); and coarse aggregate (CA). Both DTT and RFT models were developed for 94 collected experimental datasets from the published literature. The predicted results are further validated through K-fold cross-validation using correlation coefficient (R2), mean absolute error (MAE), root mean square error (RMSE), relative root mean square error, relative square error (RRMSE), and performance index factor (PiF). The RFT model was found to have the lowermost MAE 3.253, RMSE 4.387, RRMSE 0.0803, and performance index factor (PiF) 0.0061. In comparison, predicted results overall revealed better performance and accuracy for the RFT-developed models than for DTT and gene expression programming (GEP) models, as illustrated by their high R2 value, equal to 0.96, while the R2 value for DTT and GEP was found 0.94 and 0.86, respectively.

Thermal Conductivity of Coconut Shell-Incorporated Concrete: A Systematic Assessment via Theory and Experiment

To minimize the energy consumption and adverse impact of excessive waste accumulation on the environment, coconut shell (CA) became a potential (partial) replacement agent for fine aggregates in structural concrete production. Thus, systematic experimental and theoretical studies are essential to determine the thermal and structural properties of such concrete containing optimum level of CA. In this view, an artificial neural network (ANN) model, gene expression programming (GEP) model, and response surface method (RS) were used to predict and optimize the desired engineering characteristics of some concrete mixes designed with various levels of CA inclusion. Furthermore, the proposed model’s performance was assessed in terms of different statistical parameters calculated using ANOVA. The results revealed that the proposed concrete mix made using 53% of CA as a partial replacement of fine aggregate achieved an optimum density of 2246 kg/m3 and thermal conductivity of 0.5952 W/mK, which was lower than the control specimen (0.79 W/mK). The p-value of the optimum concrete mix was less than 0.0001 and the F-value was over 147.47, indicating the significance of all models. It is asserted that ANN, GEP, and RSM are accurate and reliable, and can further be used to predict a strong structural–thermal correlation with minimal error. In brief, the specimen composed with 53% of CA as a replacement for fine aggregate may be beneficial to develop environmentally amiable green structural concrete.

Boundary Lubricity of Vegetable-Oil-Derived Trimethylolpropane (TMP) Ester

Vegetable-oil-based biolubricants are an excellent alternative to conventional lubricants. Instead of focusing on novel feedstocks, these biolubricants should be further elucidated based on their fatty acid composition, which influences their tribological properties. Therefore, the study utilises gene expression programming (GEP) to derive a boundary lubricity model for vegetable-oil-derived trimethylolpropane (TMP) esters, considering the fatty acid composition (saturation and monounsaturation levels), load and speed. Neat vegetable oil and blends from seven feedstocks are selected following a wide range of fatty acid profiles to synthesise TMP esters using a two-stage transesterification process. The TMP esters are spin-coated on wear discs that are subsequently rotated against a ball using a purpose-built tribometer. The frictional performance of the TMP esters with balanced saturation and monounsaturation levels of fatty acid are measured to improve it at higher speeds. The GEP model is statistically evaluated by adopting the friction data, a showing good generalisation and predictability capability. The model demonstrates that friction decreases with increasing saturation levels of the TMP ester. The GEP model for vegetable oil TMP esters allows for the tribological performance prediction of TMP esters following the fatty acid profile, providing a platform to optimise such biolubricant for desired applications.

Gene Expression Programming Model for Tribological Behavior of Novel SiC-ZrO2-Al Hybrid Composites.

In order to improve product format quality and material flexibility, variety of application, and cost-effectiveness, SiC, ZrO2, and Al hybrid composites were manufactured in the research utilizing the powder metallurgy (PM) technique. A model was created to predict the tribological behavior of SiC-ZrO2-Al hybrid composites using statistical data analysis and gene expression programming (GEP) based on artificial intelligence. For the purpose of examining the impact of zirconia concentration, sliding distance, and applied stress on the wear behavior of hybrid composites, a comprehensive factor design of experiments was used. The developed GEP model was sufficiently robust to achieve extremely high accuracy in the prediction of the determine coefficient (R2), the root mean square error (RMSE), and the root relative square error (RRSE). The maximum state of the RMSE was 0.4357 for the GEP-1 (w1) model and the lowest state was 0.7591 for the GEP-4 (w1) model, while the maximum state of the RRSE was 0.4357 for the GEP-1 (w1) model and the minimum state was 0.3115 for the GEP-3 model (w1).

Synthesis of heated aluminum oxide particles impregnated with Prussian blue for cesium and natural organic matter adsorption: Experimental and machine learning modeling

Heated aluminum oxide particles impregnated with Prussian blue (HAOPs-PB) are synthesized for the first time using different molar ratios of aluminum sulfate and PB to improve the adsorption of cesium (133Cs+) and natural organic matter (NOM) from an aqueous solution. The Cs+ adsorption from various aqueous solutions, including surface, tap and deionized water by synthesized HAOPs-PB, is investigated. The influencing factors such as HAOPs-PB mixing ratio, pH and dosage are studied. In addition, pseudo 1st and 2nd order is tested for adsorption kinetics study. A machine learning model is developed using gene expression programming (GEP) to evaluate and optimize the adsorption process for Cs+ and NOM removal. Synthesized adsorbent showed maximum adsorption at a 1:1 M ratio of aluminum sulfate and PB in DI, tap, and surface water. The pseudo 2nd order kinetics model described the Cs + adsorption by HAOPs-PB more accurately that indicating physiochemical adsorption. Adsorption of Cs+ showed an increasing trend with higher HAOPs-PB concentration, while high pH also favored the adsorption. Maximum NOM adsorption is found at a higher HAOPs-PB dosage and a neutral pH value. Furthermore, the proposed GEP model shows outstanding performance for Cs+ adsorption modeling, whereas a modified-GEP model presents promising results for NOM adsorption prediction for testing dataset by learning the relationship between inputs and output with R2 values of 0.9348 and 0.889, respectively.

Predicting the Compressive Strength of Concrete Containing Fly Ash and Rice Husk Ash Using ANN and GEP Models.

Climate change has become trending news due to its serious impacts on Earth. Initiatives are being taken to lessen the impact of climate change and mitigate it. Among the different initiatives, researchers are aiming to find suitable alternatives for cement. This study is a humble effort to effectively utilize industrial- and agricultural-waste-based pozzolanic materials in concrete to make it economical and environmentally friendly. For this purpose, a ternary blend of binders (i.e., cement, fly ash, and rice husk ash) was employed in concrete. Different variables such as the quantity of different binders, fine and coarse aggregates, water, superplasticizer, and the age of the samples were considered to study their influence on the compressive strength of the ternary blended concrete using gene expression programming (GEP) and artificial neural networking (ANN). The performance of these two models was evaluated using R2, RMSE, and a comparison of regression slopes. It was observed that the GEP model with 100 chromosomes, a head size of 10, and five genes resulted in an optimum GEP model, as apparent from its high R2 value of 0.80 and 0.70 in the TR and TS phase, respectively. However, the ANN model performed better than the GEP model, as evident from its higher R2 value of 0.94 and 0.88 in the TR and TS phase, respectively. Similarly, lower values of RMSE and MAE were observed for the ANN model in comparison to the GEP model. The regression slope analysis revealed that the predicted values obtained from the ANN model were in good agreement with the experimental values, as shown by its higher R2 value (0.89) compared with that of the GEP model (R2 = 0.80). Subsequently, parametric analysis of the ANN model revealed that the addition of pozzolanic materials enhanced the compressive strength of the ternary blended concrete samples. Additionally, we observed that the compressive strength of the ternary blended concrete samples increased rapidly within the first 28 days of casting.

Soğan Dilimlerinin Kuruma Özelliklerinin Modellenmesinde Gen İfade Programlamasının (GEP) Kullanımı (Allium Cepa)

In this study, the drying properties of onion slices were experimentally investigated under the different drying temperatures of 45, 50, 55, and 60°C for different thicknesses of 2, 3, 4, and 5mm. The drying characteristics of the onion slices were significantly influenced by drying temperature. Thin slices with higher temperature dried in the shortest time while thick slices with low temperature took longer to dry. In modeling drying curves, the moisture ratio values of the onion slices were compared with five models commonly used in the literature. In addition, Gene Expression Programming (GEP) was used to model the drying characteristics of the onion slices, and mathematical formulas were derived to calculate moisture ratio values. The results indicated that the moisture ratio values predicted by all models agreed with the experimental moisture ratio values for onion slice samples at different temperatures. The scientific findings we obtained here showed that the two model provided a better simulation of onion slice drying kinetics than other models in different experimental temperature and thickness ranges. Also, the GEP model was able to usefully determine the moisture ratio of onion slices with appropriate accuracy in a shorter time without the need for complicated formulas.

Machine learning based computational approach for crack width detection of self-healing concrete

Concrete structures frequently experience the phenomena of crack development. The researchers used certain healing agents to boost the frequently observed autogenous crack-healing capacity of concrete. Although various artificial intelligence (AI) techniques have been used to forecast a number of concrete properties, the application of AI to forecast self-healing capacity of engineering cementitious composites (ECC) is relatively rare. For this purpose, three gene expression programming (GEP) models were created to predict the potential of admixture-based concrete to self-heal. A total of 619 data points were extracted from the literature with four contributing factors i.e., amount of fly ash (FA), silica fume (SF), limestone powder (LP), and crack width before self-healing (CWB) in order to forecast the crack width after self-healing (CWA). The data points were divided into training (70%) and testing (30%) sets. Various performance indicators were employed to assess the efficacy of the derived GEP models, which includes coefficient-of-correlation (R), relative-root-mean-square-error (RRMSE), root-mean-squared-error (RMSE), root-mean-squared-logarithmic-error (RMSLE), Nash-Sutcliff efficiency (NSE), root-squared-error (RSE), mean-absolute-error (MAE), performance-index (PI), and objective-function (OBF). The best optimized GEP model (SHC-GEP 1), was found with R = 0.938, NSE = 0.944, RMSE = 2.799 µm, MAE = 3.72 µm, RMSLE = 0.006 µm, RSE = 0.124 µm, and RRMSE = 0.439 µm, in the testing phase. While the OBF was less than 0.2 (i.e., 0.119), indicating that the model is free from overfitting issue. In contrary, the conventional linear regression model fails to meet this criterion and gives many negative predicted values. Moreover, the sensitivity analysis results indicate that the CWB has the greatest impact on the CWA. Also, the parametric trends between input variables and CWB, are in-line with the previous literature, indicating the robustness of the established model. Additionally, the suggested GEP model may be used as a cutting-edge alternative for forecasting the final crack width after concrete self-heals, helping engineers to assess the crack-reduction capability. Furthermore, it is recommended to explore the crack healing capabilities of other supplementary cementitious material like bagasse ash, wheat straw ash, and soda glass powder, subjected to modeling their healing ability using AI techniques.