Optimization of Green Concrete Containing Fly Ash and Rice Husk Ash Based on Hydro-Mechanical Properties and Life Cycle Assessment Considerations

The development of sustainable concrete in achieving the developmental goals of the United Nations in terms of sustainable infrastructure and innovative technology forms part of the focus of this research paper. In order to move towards sustainability, the utilization of the by-products of agro-industrial operations, which are fly ash (FA) and rice husk ash (RHA), in the production of concrete has been studied. Considering the environmental impact of concrete constituents, multiple mechanical and hydraulic properties of fly ash (FA) and rice husk ash (RHA) concrete have been proposed using intelligent techniques; artificial neural network (ANN) and evolutionary polynomial regressions (EPR). Also, an intelligent mix design tool/chart for this case under study is proposed. Multiple data points of concrete materials, which were further reduced to ratios as follows; cement to binder ratio (C/B), aggregate to binder ratio (Ag/B), and plasticizer to binder ratio (PL/B) were used in this exercise. At the end of the protocol, it is observed that the constituents’ ratios are dependent on the behavior of the whole, which can be solved by using the proposed model equations and mix design charts. The models performed optimally, as none showed any performance below 80%. However, ANN, which predicted Fc03, Fc07, Fc28, Fc60, Fc90, Ft28, Ff28 & Fb28, S, Ec28 & K28, and P with an accuracy of greater than 95% each with average error of less than 9.4% each, is considered the decisive technique in predicting all the studied concrete properties, including the life cycle assessment potential of the concrete materials.


Introduction
The worldwide construction sector has profited from the world's fast-growing economy during the twentieth century and continues to achieve new heights [1]. Sustainability in the building industry is among the most crucial needs of developing countries because of limited natural resources and growing carbon dioxide emissions from massive cement and concrete manufacturing. Due to the breakdown of utilized CaCO3 in the cement clinker and the burning of fuel, cement production is responsible for emitting approximately 5 to 7% of carbon dioxide emissions worldwide [2][3][4][5][6][7]. As testing, the rebound hammer test was used, and a mathematical relationship was developed between the compression and rebound values of RHA concrete, which showed a strong correlation.
Current studies and allied sectors have started to concentrate on climate change and environmental problems. As a result, numerous methods and indicators have been created to better identify the primary causes of these issues and to take steps to mitigate their effects. For the environmental assessment of "green" concrete mixes, a systematic Life Cycle Assessment (LCA) method is needed. Despite the importance of environmentally friendly infrastructure actions, the rising costs of building materials and other building supplies, and the reality that even before concrete is poured in a structure, its effects are imprinted for several years [40], a life-cycle framework is especially important for a systematic detection and characterization of the environmental impacts of both traditional and "green" concrete production. LCA is a method for statistically and comparably analyzing the environmental effect of a product/process [41]. The use of SCMs instead of clinker/cement has been shown to have a lower environmental effect [17]. Despite various researches focusing on ecologically friendly concrete mixes utilizing FA, the authors are unaware of any LCA of rice hull ash concrete. Rice hull ash in concrete is not widely used in the U.S. since nowadays it is a novel product that requires additional investigation into its environmental implications. Because RHA improves early strength development, which is an issue with FA concrete, it can be used instead of OPC in the concrete mixture [42]. Figure 1 summarized the structural and environmental benefits of using FA and RHA in concrete. The development of intelligent models to optimize the utilization of the combined effect of rice husk ash (RHA) and fly ash (FA) in concrete would have solved the inconsistency in the right proportions to achieve the required strengths for a more sustainable infrastructure. Furthermore, a clearer understanding of the environmental impact of the inclusion of RHA and FA in concrete as partial replacements for cement would have equally informed future research, design, and construction of concrete structures on the optimal amount of replacement needed to achieve sustainable strength without further endangering the environment.
In this study, intelligent models and mix design tools/charts have been proposed for rice husk ash and fly ash blend concrete data points under environmental impact considerations for the hydro-mechanical properties (slump, compressive strength, splitting tensile strength, flexural strength, bonding strength, elastic modulus, and permeability) of concrete.

Green Concrete
During the last century, the building and construction industry have relied heavily on concrete, the most frequently utilized and efficient man-made substance [43][44][45][46][47]. Concrete constructions can be customized to have any required qualities and come in just about any required form. This is conceivable due to the robustness, foldability, and less life cycle cost of the material [48][49][50][51]. The primary conventional binder for concrete manufacture is Ordinary Portland Cement (OPC). By 2050, cement consumption is predicted to exceed 6 Gt/year [52][53][54]. OPC manufacture requires a significant amount of fuel and natural resources. It is a multi-stage industrial process that produces massive amounts of greenhouse gases such as CO2 [55][56][57]. Numerous environmentalists criticize the manufacturing industry for pollutant emissions, claiming that it contributes for around 5-7% of worldwide CO2 emissions. For every kilogram of cement created, around 0.8-0.9 kg of CO2 is released [58,59]. There is worldwide agreement on the need to reduce the carbon footprint of the cement industry via substantial substitution of cement in concrete. As a result, a solution for the longterm sustainability of concrete is required, as is the development of a new binder capable of overcoming the barriers associated with the manufacture and implementation of OPC [60][61][62].
Supplementary cementitious materials (SCMs) constitute substances that may be used in lieu of OPC in certain situations. Many industrial and agricultural waste products are now being employed as SCMs on a global scale. Fly ash (FA), rice husk ash (RHA), volcanic ash (VA), electric arc furnace slag (EAFS), zeolite (ZLT), metakaolin (MK), and silica fume (SF) are the most frequently utilized SCMs [63,64]. Different industrial wastes have been employed to enhance the mechanical and durability qualities of concrete throughout the years. This also contributes to reducing reliance on OPC. Among the many types of industrial waste, FA is widely utilized and has been seen as a viable substitute for OPC [65][66][67]. FA is a coal combustion byproduct that is also collected from coal-fired thermal power plants. It has been used in place of OPC or clinker cement as a partial substitute for a long period of time. The amount of FA in the combination is limited to 15%-20% by weight of the OPC [68][69][70]. The concrete produced with highvolume FA has such a low per-unit cost of manufacture. Attributed to the prevalence of SiO2 and Al2O3, FA has a unique pozzolana characteristic. SiO2 and Al2O3 combine with calcium hydroxide during the hydration process, generating additional calcium silicate hydrate (CSH) and calcium aluminate hydrate (CAH). The concrete appears to be getting denser when CSH and CAH are present. Additionally, it plugs voids in the concrete and strengthens the matrix. It contributes to the reduction of the difficulties associated with sustainable building, including its environmental effects [71].
Alternatively, it should be mentioned that RHA has grown to become one of the most important SCMs in the world. Among the SCMs derived from rice crop agricultural waste is RHA, which is derived from the waste of the rice crop [72,73]. Rice grains are wrapped with rice husks (RH), which are then used as a fuel in rice mills to boil the paddy that has been harvested. As a silica-containing material, it has the potential to be utilized effectively as an SCM in the production of concrete [74]. As a result, RHA may be used as a cementitious material with efficiency in certain applications. RHA does not emit a significant quantity of carbon dioxide into the atmosphere. It has the potential to be employed as a structural concrete. Not only does it add to the overall strength of the concrete, but it also contributes to the long-term durability of the concrete's qualities [75].

Evolutionary Polynomial Regression (EPR)
Genetic Algorithm (GA) is famous optimization technique depends on simulating the evolution process of biological creatures. It depends on one simple rule "The most fitting creature will survive". To apply this principal on optimization, there must be a pool of solutions for the considered problem, a fitting criterion and a procedure to generate new solutions by mixing the existing ones. Biological creatures transfer their data to the next generation as an arranged series of genes called "Chromosome", similarly, (GA) presents the solution (chromosome) as an arranged list of steps (genes). (EPR) is a direct application of (GA), it depends on optimizing the number of terms of the "Traditional Polynomial Regression (TPR)". (TPR) is a well-known mathematical regression technique uses "Least Squared Error" principal to find the optimum coefficient values of a certain polynomial function to fit a certain dataset. The considered polynomial may be single or multi variables depending on the considered problem configuration (dataset). The chosen polynomial degree (its highest power) depends on the complexity of the considered problem, first degree polynomial (linear) may be used for simple problems, for more complicated ones, second degree (quadratic), third degree (cubic) of higher degrees may be required. The number of polynomial terms dramatically increases with increasing the variable numbers and polynomial degree, for example, a two variables second degree polynomial have only six terms (X2+Y2+ XY+X+Y+C), while three variables third degree polynomial have 20 terms and four variables forth degree polynomial have 70 terms and so on. As the number of polynomial terms increases, it becomes more difficult to apply less practical. Hence, (EPR) technique aims to optimize the (TPR) by eliminating the less important terms and keep only the most effective ones using (GA) technique. So, the population (solutions) consists of a set of polynomials, the fitting criteria is the "Sum of Squared Errors (SSE)" and the chromosome consists of a list of polynomial terms, the length of the chromosome is the chosen number of terms. Cycles after cycle, the most important terms accumulate in the survival chromosomes and the less important ones deleted.

Artificial Neural Network (ANN)
(ANN) is an umbrella of a wide range of (AI) techniques that depends of mimicking the behavior of biological neurons. They all consist of nodes (cells or neurons) and links to connect the nodes, but they have different neurons arrangement and connection patterns. "Multi-Layer Perceptron (MLP)" is one of the earliest and most (ANN) types. It is the commonly used type for regression problems. It consists of a number of nodes arranged in layers, the first layer is called "Input layer" and it is used to receive the input values, while the last layer is called "Output layer" and it is used to deliver outputs values. Between the input and the output layers there are a number of intermediate layers called "Hidden layers", they are responsible of predicting the outputs from the inputs. (MLP) must have one hidden layer at least. Each node in a certain layer is connected to all the nodes in the previous and the next layers by links, but the nodes of each layer are not connected to each other. Each link has an importance factor called "Weight" and each node has a triggering formula called "Activation Function", this could be any nonlinear function, but the most popular ones are the sigmoid, the hyper-tan and the ramp functions, they are responsible of the nonlinear capability of the (ANN). Due to the variation in ranges of input values, all inputs must be scaled to a unified range, this process is called "Standardization" if the input variance is divided by its standard deviation (SD) and called "Normalization" if the inputs are scaled between (0 to 1) and called "Hyper normalization" if the inputs are scaled between (-1 to 1). The scaled inputs propagate from the input layer to the output layer through the hidden layers. The output of certain node is the result of applying its activation function on the summation the node inputs multiplied by corresponding links' weights. After the output layer, the outputs must be de-scaled to its original renege. Any (ANN) model must be trained using a given dataset, during the training process the weight values of the model's links are adjusted to predict the correct outputs from the inputs. There many training techniques could be used to find the optimum values for links' weights such as "Back Propagation (BP)", "Gradually Reduced Gradient (GRG)" and "Genetic Algorithm (GA)".

Life Cycle Assessment
Climate change is attributed to the increase in greenhouse gas emissions (GHG) in the atmosphere. Approximately 54 giga-tons of carbon dioxide (CO2-eq) were recorded as the total annual GHGs emissions in 2017. Concrete is one of the main contributors to these emissions and concrete usage and demand for cement in the construction industry is expected to be increased in the future due to the rapid urban development in emerging countries. Annually about 20 billion tons of concrete are produced worldwide, and Portland Cement (PC) production is responsible for 7-10% of total CO2 emissions. When considering the construction and building sector, approximately 80% of the GHG emissions and energy consumption are generated during the operation phase of the buildings. Many studies have been conducted and introduced new technologies, policies, and mitigation measures/techniques to reduce GHG emissions during the operation stages [76]. However, around 10-20% of GHG emission and energy consumption is from material manufacturing, construction, and demolition, with limited studies reported related to reducing GHG emissions and other environmental impacts during material production and construction stages [77]. Life Cycle Assessment (LCA) is an established and well-known method that can evaluate the environmental impacts of products throughout its life cycle. To date, limited studies have been conducted to compare the detailed environmental impact of using blended alkaliactivated binders against PC binders in the construction industry. Adapting the waste products (fly ash and RHA) by replacing cement in the construction industry reduces the environmental impact and cost obtained from source material manufacturing and solves problems associated with landfill disposal activities [78,79]. The utilization of such industrial and agricultural waste could be a massive benefit to the environment. Hence this study investigates the GHG emissions from cradle to end of building construction stage (cradle to end of construction), a life cycle impact and benefit assessment of the manufacturing stage (cradle to gate) and initial cost analysis of fly ash geopolymer concrete (100FA), alkali-activated fly ash-RHA blended concrete (hereafter, used as blended alkali-activated concrete) and comparing with PC concrete. The outcomes of this study will useful when adapting feasible, sustainable alternatives in alkali activated concrete manufacturing and construction considering the environmental and economic aspects [59,80].
As previously stated, the substitution of OPC with SCMs has long been considered as one of the most efficient methods of reducing greenhouse gas emissions. Additionally, it is essential mentioning that the amount of cement in the mix has a direct correlation with the concrete's durability, environmental impact, quality, and other characteristics. Scholars employed experimental methods to investigate the effect of SCMs on the mechanical properties of green concrete. Pitroda et al. [81] examined the mechanical properties of concrete by varying the fly-ash concentration. The fly-ash content was tested by substituting fly-ash for cement. When the qualities of conventional strength concrete were compared to those of concrete containing 10% fly ash, it was determined that the latter possessed superior properties. Once the fly-ash concentration approached ten percent, the concrete's mechanical properties deteriorated as the fly-ash amount grew. Saha [82] studied the use of class F fly ash as a partly cement replacement in concrete. Compared with the control samples, the fly ash specimens had an exceedingly low early compressive strength. Nonetheless, the pozzolanic activity improved the specimens' strength over time, whereas the reference samples' strength paused after 56 days of curing. Increasing the fly ash content in the mix reduced drying shrinkage. Using fly ash as a binder reduced the porosity of the concrete. As a result, the fly ash concrete's water sorptivity and chloride permeability were lowered. Giaccio & Malhotra and Feldman et al. [83,84] conducted extensive study on high-volume FA concrete. The experimental findings indicated that concrete containing a high proportion of Class F FA enhanced mechanical qualities, had extremely little chloride ion penetration, exhibits adequate resistance to repeated freezing and thawing cycles, and had no adverse effect when reactive aggregate was added. Bouzoubaâ et al. [85] observed that test results of concrete comprising class C and class F FA were equivalent to or superior to those achieved using ordinary concrete. As a consequence, FA concrete offers equivalent or superior properties in comparison with ordinary concrete while being less costly. Over the course of 28 days, Poon et al. [86] explored the compressive strength of concrete comprising 45% FA and discovered that the concrete attained a compressive strength of 80 MPa and exhibited lesser heat of hydration and chloride diffusivity when compared to OPC. The influence of plastic waste (PW) and graphene nanoplatelets (GNP) on the characteristics of high-volume fly ash (HVFA) concrete was explored by Adamu et al. [68], and the outcomes indicated that the response surface methodology was utilized for experiment design, data modeling, and process optimization. PW, HVFA, and GNP were used as variables, and the materials' strength and water absorption were used as responses. PW and HVFA both diminish strength and absorption, but GNP strengthens both of these features. Each suggested model was statistically substantial and had a high level of correlation [85][86][87].
Rice husk is among the most common agricultural wastes, obtained from the outer coating of rice grains throughout the milling operation. It is one of the most common crop wastes. Twenty percent of the 500 million tons of paddy produced across the globe comes from this region. More academics have thus concentrated their efforts on the impact of this ecologically friendly substance on the qualities of concrete as a result. For example, in 2005 Bui et al. [64] evaluated the effect of particle size on the strength of blended gap-graded Portland cement concrete made with RHA. This study investigated the strength activity index of mortars including residual RHA created during the burning of rice husk pellets and RHA derived via grinding residual RHA. The compressive strength and durability of concrete were examined when ground RHA was partially substituted for cement. Ameri et al. [87] implemented a study about concrete including RHA. According to the findings, concrete with RHA exhibited a significant increase in early compressive strength. Increases in RHA concentrations greater than 15% resulted in a decrease in compressive strength. This is due to the presence of an overabundance of silica in RHA which was not been reacted with other components. Iqtidar et al. [88] employed machine learning to forecast the qualities of RHA-containing concrete. This research evaluated the compressive strength of rice husk ash mixed concrete using 192 data points. Age, quantity of cement, RHA, superplasticizer, water, and aggregates were all input factors. This study used four approaches of soft computing and machine learning, including artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS), multiple nonlinear regression (NLR), and linear regression. The acquired data were evaluated using sensitivity analysis, parametric analysis, and the correlation factor (R2). The findings indicated that ANN and ANFIS outperformed other approaches. Amin et al. [63] employed sophisticated machine learning approaches to estimate the compressive strength of RHA. Six inputs were chosen based on the available literature: specimen age, percentage of rice husk ash, percentage of superplasticizer, aggregates, water, and quantity of cement. The results of machine learning techniques were compared to those of more conventional techniques such as linear and non-linear regressions. The performance of machine learning approaches was shown to be superior to that of conventional methods for calculating the compressive strength of RHA. Iftikhar et al. [89] used machine learning methods to anticipate and establish an empirical formula for the compressive strength of concrete containing RHA. Gene expression programming (GEP) and Random Forest Regression (RFR) were used in this investigation. The models were developed using a dependable database of 192 data points. The most significant factors in the creation of RHA-based concrete models were age, cement, RHA, water, superplasticizer, and aggregate. Models were evaluated using a variety of statistical variables.

Novelty and Innovation
In an effort to propose models; analytical and numerical in order to solve the problems of sustainability in concrete production and also those of repeated visits to laboratories prior to design and construction activities, databases of various sizes have been explored as well as different artificial intelligent techniques. In this present research work, an extensive literature search was done to collect a rich database comprising of concrete constituents, which include rice husk ash (RHA) and fly ash (FA), which form the focus of this work in terms of sustainable concrete devoid of or with less amount of greenhouse emission footprint. Also, these concrete constituents were reduced to ratios, which make easier unlike in previous works to handle and predict more accurate functions. In addition, the evaluation of the environmental impact of the concrete constituents was conducted to determine the life cycle assessment potential, which was incorporated into further research considerations in this work. Moreover, multiple AI techniques have been deployed in this research work to a step further to propose concrete design tools/charts, which works in synergy with the closed-form AI models to predict concrete design mixes of desirable strength and accompanying reduced environmental impact. With all these innovative considerations, this research work stands out from what has been researched previously in this field of sustainable and artificial intelligence applications.

Collection of Concrete Experimental Data
An extensive literature study was conducted, which gave rise to multiple points of concrete mixes containing the following; cement, rice husk ash, fly ash, water, plasticizer, fine aggregates, and coarse aggregates [90]. The following concrete hydraulic and mechanical properties were studied; air content, slump, hydraulic conductivity, compressive strength, splitting tensile strength, flexural strength, bonding strength and elastic modulus. In Figure 2, the theoretical framework of the research methodology and modelling operations are presented.
Where (B) is binder content (cement content + fly ash or rice husk ash content). For all collected concrete mixes, the additional cementitious material was either fly ash (FA) or rice husk ash (RHA) but not a combination of them. Also, the ratio between coarse and fine aggregates (CAg/FAg) is constant and equals to 2.0 in all collected design mixes, hence, both of them are equivalent to only one independent variable presented by total aggregate content (Ag). The collected records were divided into training set (85 records) and validation set (20 records). Table 1 includes the  complete combinatorial dataset, while Tables 2 and 3 summarize their statistical characteristics and the Pearson correlation matrix. Finally, Figure 3 shows the histograms for both inputs and outputs. This presents the internal consistency of the studied parameters, which shows their distribution frequency in the array of the concrete mix points. It can be observed that C/B and Ag/B have the more central distribution within the central bin of the database for the inputs while S has the most central distribution bin for the outputs.

Research Program
Two different models were used to predict the twelve outputs of concrete mix using the gathered dataset. The implemented techniques are "Artificial Neural Network (ANN)" and "Evolutionary Polynomial Regression (EPR)". Each implemented technique is based on different approach mimicking human brain for ANN, optimization of mathematical regression for EPR and simulating evolution of natural creatures for GA. However, for all techniques, their accuracies were evaluated in terms of "Sum of Squared Errors (SSE)", "Root of Mean of Squared Errors (RMSE)" and "Determination Coefficient (R 2 )". Table 2 illustrated the different correlations between considering inputs and outputs, based on these values the predicting models were constructed as shown in Figure 4. Where the concrete strengths (Fc03, Fc07,  Fc28, Fc60, Fc90, Ft28, Ff28 & Fb28) were predicted using the design mix ratios (C/B), (Ag/B) & (PL/B), while the predicted Fc28 is used besides the design mix ratios to predict the values of (S, Ec28 & K28). Finally, the predicted Fc28, S, Ec28 in addition to the design mix ratios are used to predict (P) values. The next sections present the results of each technique and its accuracy metrics.

General Remarks on the Concrete Mixes' Strength and Environmental Impact
The concrete mixes collected from literature was reduced to the following ratios; cement to binder (cement + fly ash + rice husk ash) ratio (C/B), aggregate (FAg + CAg) to binder ratio (Ag/B), and plasticizer to binder ratio (PL/B) and the behaviour of the hydro-mechanical properties of the concrete were observed under life cycle assessment (P) considerations. It can be observed that none of the concrete constituents' ratios reacted independent of the other. But more emphasis is placed on the binders and their ratios (C/B and PL/B) as they constitute the contribution to the global warming potential (GWP) of cement use in concrete production, transportation and utilization during construction activities. It is observed that a reduction in the C/B ratio brings about an equivalent reduction in P, though it is not consistent with all the points of the database. This indicates that there is a portion of environmental impact contribution from the Ag/B ratio [35,85]. This further implies that aggregates (Ag) equally consume energy and release considerable amount of greenhouse emission during the production process (quarrying) and transportation. However, closed-form equations, intelligent models and intelligent mix design tools have been proposed for fly ash/rice husk ash blend in concrete to solve this optimal materials selection and sampling riddles. Furthermore, this development can be employed to solve infrastructural design and problems dealing with concrete structures as well as concrete containment liner systems which need a compressive strength as low as 0.2MPa and hydraulic conductivity (k) of 1E-09m/s for singles liner systems and 1E-07 for double liner systems [25]. It can also be noted that air content was not considered in the parametric study and the models' execution. This was because voids have minor effect on concrete strength, for normal weight, non-air-entrained concrete; the void ratio is ranged between 0.5% and 2.5% with average value of 1.5+/-1%, that affect the strength by +/-5% as per ACI-212 and E4-12.

Using ANN Technique
As shown in Figure 4, the predictive model consists of three sub-models. The first is for concrete strengths; the second is for concrete physical properties; and the third is for environmental impact factor. The three sub-models were developed using the ANN technique with a layout of (3:8:8), (4:8:3), and (6:8:1), respectively. All the sub-models used the traditional "Back Propagation (BP)" training algorithm, standardization method (Var/S.D.), and activation function (Hyper Tan). The used network layout is illustrated in Figures 5 to 7, while the weight matrices of each model are shown in Tables 4 to 6. The prediction error % values of the first sub-models were ranged between 6.9% and 9.4% with an average value of 8.2%, and the error % of the second sub-model were ranged between 5.9% and 7.6% with an average value of 7.0%, while the error % of the third sub-model was 6.8%. On the other hand, the R 2 values of the first submodel were ranged between 0.925 and 0.965 with an average value of 0.945, and for the second sub-model, the R 2 values were ranged between 0.878 and 0.949 with an average value of 0.894, and finally, the R 2 value of the third submodel was 0.960. This compares well with previous research works which utilized fly ash-silica combination and fly ash alone in concrete production [91][92][93][94][95][96], and had used other intelligent methods in their prediction of the concrete mechanical properties.      Figure 8 illustrates the relative importance values of each input parameter for P and it indicates that all input parameters have almost the same importance level except (Ag/B) which has a slightly higher importance. This conclusion confirms the proposed theoretical framework shown in Figure 4.

Using EPR Technique
Similarly, the developed EPR model was divided into three sub-models as per Figure 4. The first sub-model was concerned in predicting the concrete strengths using a set of eight optimized cubic polynomials. All polynomials have the same three inputs (C/B, Ag/B & PL/B), hence, each polynomial have 20 possible terms (10+6+3+1=20) as follows: GA technique was applied on these polynomials to select the most effective five terms of each polynomial to predict (Fc03, Fc07, Fc28, Fc60, Fc90, Ft28, Ff28 and Fb28). The outputs are illustrated in Equations 2 to 9 and its fitness is shown in Figure 9. The error% values were ranged between 9.2% and 10.9% with average value of 9.7%, while (R 2 ) values were ranged between 0.917 and 0.938 with average value of 0.928. This also compares well with previous research works, which utilized fly ash-silica fume combination and fly ash alone in concrete production [91,96], and had used other intelligent methods in their prediction of the concrete mechanical properties.
The second sub-model was concerned in predicting the physical properties of concrete using a set of three optimized cubic polynomials. All polynomials have the same four inputs (C/B, Ag/B, PL/B & Fc28), hence, each polynomial have 35 possible terms (20+10+4+1=35) as follows: GA technique was applied on these polynomials to select the most effective seven terms of each polynomial to predict (S, Ec28 & K28). The outputs are illustrated in Equations 11 to 13 and its fitness is shown in Figure 9. The error % values were ranged between 8.6% and 10.5% with average value of 9.3%, while R 2 values were ranged between 0.831& 0.887 with average value of 0.868.
Finally, the third sub-model was concerned in predicting the environmental impact factor using six inputs cubic polynomials, which are (C/B, Ag/B, PL/B, Fc28, S and K28), hence, the polynomial has 84 possible terms (56+21+6+1=84) as follows: GA technique was applied on these polynomials to select the most effective nine terms of each polynomial to predict P. The output is illustrated in Equation 15 and its fitness is shown in Figure 9. The error% and (R 2 ) values were (21.4% -0.371).
The results of all developed models are summarized in Table 7. Figure 9 compare the accuracies of the developed models graphically. Fc03 ANN Figure 5, Table 4 Figure 9. Relation between predicted and calculated values using the developed models

The Developed Mix Design Aid
Although the developed predicting models for (Fc28) using (ANN) and (EPR) are accurate enough, they are still hard to implement in practical mix designs, especially for manual calculations [91][92][93][94][95][96]. Hence, concrete mix design tools were developed by substituting in the developed ANN model different combinations of input parameter values that varied at constant intervals. Figure 10 presents the corresponding binder content (B) and water-binder ratio (W/B) for each input parameter combination, while the calculated (Fc28) values were plotted on the contour charts shown in Figure  11. The key chart indicates the changes in mix design when moving toward each side of the design chart; moving vertically changes the concrete strength and plasticizer-binder ratio, while moving horizontally changes the binder content and water-binder ratio. Using this tool is very simple, requiring only the desire (Fc28) and cement-binder ratio (C/B). Just select the correct chart in Figure 11 based on the (C/B) value, then locate the required (Fc28) on the selected chart to determine both (Ag/B) and (PL/B). From the corresponded cells in Figure 10, both binder content (B) (ton/m 3 ) and water-binder ratio (W/B) could be determined. Finally, the absolute values of water, plasticizer, cement, (fly ash or rich husk ash), and fine and coarse aggregates [92,94] could be calculated by multiplying their ratios by the binder content, considering that the coarse aggregate content is twice the fine aggregate content.

Dec. W
For example, to design a concrete mix with (Fc28=60MPa) and (C/B=80%), the middle left chart in Figure 11 will be used, (Ag/B) is 3.5 and (PL/B) is 1%. From Figure 10, the binder content is 0.49 t/m 3 and (W/B) is 40%. Hence, 1.0 m3 of this mix contains 390kg cement, 100kg fly ash or rice husk ash, 195 litter of water, 4.9kg plasticizer, 570kg fine aggregate (sand) and 1140kg coarse aggregate.

Conclusions
This research presents three models using two techniques (ANN, and EPR) to predict concrete compressive strength at different ages (Fc03, Fc07, Fc28, Fc60, Fc90), splitting, flexural and bond strengths after 28 days (Ft28, Ff28 and Fb28), slump, elastic modulus and permeability after 28 days (S, Ec28 and K28) besides the environmental impact factor (P) using cement-binder ratio (C/B), total aggregate-binder ratio (Ag/B) and super-plasticizer-binder ratio (PL/B). Also, the research presented a concrete mix design tool developed based on the developed predictive models. The results of comparing the accuracies of the developed models could be concluded in the following points:  Based on the correlations between the considered variables, the two developed models were constructed using three sub-models, the first to predict the strengths, the second to estimate the physical properties, and the third to evaluate the environmental impact factor.
 ANN models showed the best accuracy, with average values of 91.8%, 93.0%, and 93.2% for the three sub-models, respectively. The total average accuracy was 92.3%. In spate of its high accuracy, it is too difficult to implement manually. The developed mix design tool provides a quick and easy alternative to the long and complicated calculations of ANN.
 The EPR model presented a less complicated and also less accurate alternative for the ANN, where its accuracies were 90.3%, 90.7%, and 78.6% for the three sub-models, respectively. The total average accuracy was 89.4%.
 The limited accuracy of the third EPR sub-model indicated that the relation between P and the considered inputs is too complicated to be captured by polynomial regression, even for the optimized one. On the other hand, the highly complicated third ANN sub-model was capable of capturing this relationship accurately.
 The sum of the absolute weights of each neuron in the input layer of the developed third ANN sub-model indicated that P is almost equally affected by all input values, which is confirmed by the developed formula in the third EPR sub-model that contains all the considered inputs.
 The GA technique successfully reduced the 20, 35, and 84 terms of the conventional polynomial regression cubic formula to only 5, 7, and 9 terms for the three sub-models, respectively, without having a significant impact on its accuracy.
 Like any other regression technique, the generated formulas are valid within the considered range of parameter values; beyond this range, the prediction accuracy should be verified.
 Figures 10 and 11 present the mapping for (Fc28) values corresponding to all possible mix combinations (within the limits of the available dataset). Studying these charts leads to the following additional remarks.
 This tool (charts) is limited to the following ranges: o Binder content, 340 to 630 kg/m 3 ; o Water-binder ratio, 31% to 54%; o Fly ash or rice husk ash-binder ratio, 0.0% to 40%; o Total aggregatebinder ratio, 2.5 to 5.5; o Coarse -fine aggregate ratio is constant and equals to 2.0; o Fc28, 20 to 105 MPa.  The maximum Fc28 values occur when the cement-binder ratio is 80%.
 Decreasing the aggregate-binder ratio increases the (Fc28) due to increased binder content and reduces the waterbinder ratio, and vice versa.
 Increasing the plasticizer-binder ratio increases the concrete strength (Fc28) regardless of the cement-binder ratio.
 Further studies may be carried out to predict the same outputs using a wider range of databases that include highand ultra-high-performance concrete, besides special types of concrete such as self-healing concrete.

Data Availability Statement
The data presented in this study are available in the article.