key: cord-0077639-sz6juvm7
authors: Faraj, Rabar H.; Mohammed, Azad A.; Omer, Khalid M.; Ahmed, Hemn Unis
title: Soft computing techniques to predict the compressive strength of green self-compacting concrete incorporating recycled plastic aggregates and industrial waste ashes
date: 2022-05-02
journal: Clean Technol Environ Policy
DOI: 10.1007/s10098-022-02318-w
sha: 38e3b5ae5e8e6514e5741888e2d8b62a644c68be
doc_id: 77639
cord_uid: sz6juvm7

ABSTRACT: Rapid urbanization and industrialization with corresponding economic growth have increased concrete production, leading to resource depletion and environmental pollution. The mentioned problems can be resolved by using recycled aggregates and industrial waste ashes as natural aggregate and cement replacement in concrete production. Incorporating different by-product ashes and recycled plastic (RP) aggregates are viable options to produce sustainable self-compacting concrete (SCC). On the other hand, compressive strength is an essential characteristic among other evaluated properties. As a result, establishing trustworthy models to forecast the compressive strength of SCC is critical to saving cost, time, and energy. Furthermore, it provides valuable instruction for planning building projects and determining the best time to remove the formwork. In this study, four alternative models were suggested to predict the compressive strength of SCC mixes produced by RP aggregates: the artificial neural network (ANN), nonlinear model, linear relationship model, and multi-logistic model. To do so, an extensive set of data consisting of 400 mixtures were extracted and analyzed to develop the models, various mixture proportions and curing times were considered as input variables. To test the effectiveness of the suggested models, several statistical evaluations, including coefficient of determination (R(2)), scatter index, root mean squared error (RMSE), mean absolute error (MAE), and Objective (OBJ) value were utilized. Compared to other models, the ANN model performed better to forecast the compressive strength of SCC mixes incorporating RP aggregates. The RMSE, MAE, OBJ, and R(2) values for this model were 5.46 MPa, 2.31 MPa, 4.26 MPa, and 0.973, respectively. GRAPHICAL ABSTRACT: [Image: see text]

Innovative and high-performance materials are necessary to develop sustainable and long-lasting infrastructure. Because concrete is the most commonly used construction material, researchers and the construction industry have made it as durable as possible by introducing new materials and methods. Among developed high-performance cementbased materials, one can mention self-compacting concrete (SCC). The development of SCC can be regarded as a historical achievement in the construction and concrete industry because of its numerous benefits compared to traditional concrete composites. It is a kind of high-performance concrete that can flow efficiently without the need for any compaction or mechanical work . This concrete provides high passing and filling ability performance compared to other cement-based composites without causing segregation and bleeding . The passing ability and flowability characteristics of SCC can be obtained through the addition of superplasticizer (high range water reducer) along with incorporating different supplementary cementitious materials (SCMs) such as ground granulated blast furnace slag (GGBFS), metakaolin, silica fume, and fly ash to maintain the stability and proper cohesiveness during the flow of the mixture (Deilami et al. 2019) . The most significant advantages of SCC compared to traditional concrete include shorter periods for construction, minimizing the labor costs, and improved compaction in the structure, especially at very congested locations when a high amount of rebars are present in the concrete elements.

Aggregate comprises about 65-80% of the concrete volume, and it can provide such excellent properties of concrete as strength, permeability, volume stability, workability, and durability (Faraj et al. 2019) . Considerable quantities of fine and coarse aggregates are required to produce large amounts of concrete for global consumption (Spiesz et al. 2016) . Large amounts of waste can be avoided by reusing recycled materials to prepare new concrete. Environmental issues about aggregate mining and waste disposal and aggregate shortages on construction sites can be overcome by this strategy (Saikia and Brito 2013) .

Due to the current need for human activities, many different types of plastics are produced every year. However, the majority of the plastic types are produced for a single-use. Therefore, if not recycled correctly and due to the very low biodegradability of plastic waste, they cause environmental pollution.

The recycling of plastic wastes can be regarded as one of the optimum solutions for reducing plastic waste's influence on the environment in terms of energy consumption and natural resource, waste disposal, global warming, and environmental pollution. The reuse of plastic waste in the building sector is an ideal option for disposing of plastic waste among the various forms of recycling management approaches (Sadrmomtazi et al. 2016) . As previously mentioned, to decrease the consumption of natural aggregates, recycled plastic (RP) aggregates can be used effectively to produce cement-based materials such as SCC. In the literature, different types of plastics such as Polyethylene Terephthalate (PET) , expanded polystyrene (EPS) , polypropylene (PP) (Yang et al. 2015) , and High Impact polystyrene (HIPS) (Chunchu and Putta 2019) were used in the production of SCC. RP aggregates used in the previous investigations were replaced by volume to substitute fine or natural coarse aggregates. Generally, the mechanical properties of SCC, especially its strength was decreased with elevating the percentage of RP aggregate content; however, this negative effect of RP aggregates can be diminished through incorporating other supplementary cementitious materials like silica fume, GGBFS, and fly ash. Moreover, to maintain the stability of SCC containing RP aggregates during its flow, a high amount of paste is required to thoroughly coat the granular materials (aggregates) (Faraj et al. 2019) . Therefore, incorporating other cementitious materials to decrease the amount of cement consumption is highly recommended.

On the other hand, it is well proved that the manufacture of cement necessitates a significant amount of energy and, at the same time, contributes to the direct or indirect emission of a large quantity of total carbon dioxide (about 7%) into the environment (Yu 2019) . Furthermore, one ton of cement requires around 2.8 tons of raw materials; this is a resource-depleting process that uses various natural raw materials such as limestone and shale to manufacture clinkers cement (Guo et al. 2010 ). By-products or other waste materials, on the other hand, will increase land occupancy and destroy the environment. Consequently, reusing waste materials has a minimal impact on the environment (Shahbazpanahi et al. 2021) .

To decrease the negative environmental impacts of cement production, replacing it with other cementing materials that have less impact on the environment is essential. This can help to produce more sustainable and cleaner concrete with improved performance. GGBFS, silica fume, and fly ash are various by-product powders that can be used as cement replacements for SCC production. These by-product ashes can be used effectively with the combination of RP aggregates in the SCC mixtures to improve the mechanical features of SCC mixtures that were reduced by incorporating RP aggregates (Sadrmomtazi et al. 2016) .

Compressive strength (CS) is an essential characteristic in constructing engineering structures among the different properties of SCC. Other durability and mechanical parameters are related to CS and can be estimated from indirect relationships with CS (Neville 1995) . Multiple cubic and cylinder samples are currently made and evaluated at various curing durations to figure out the CS of SCC in practice. In general, work on a construction site should not continue until the CS test results are obtained at a particular age, such as 28 days. This causes construction projects to be delayed; the testing process is also timeconsuming and costly (Shariati et al. 2020) .

Because changing mix proportions and components can have a considerable impact on the characteristics of SCC, determining its CS without conducting experimental tests has always been one of the challenges in concrete technology (Shariati et al. 2020) . This is especially noticeable in SCC, where pozzolanic materials like GGBFS, limestone powder (LP), fly ash, and silica fume has been used to replace cement partially, and RP aggregates are incorporated to replace natural aggregates.

Since CS is sensitive to mixture proportions and depends on several parameters, more improved approaches should be employed to reduce the necessity for experimental tests in the laboratory as much as possible and afford engineers with more straightforward methods and mathematical formulas for forecasting experimental outcomes. Soft computing techniques might be regarded as an appropriate solution in this regard. The most significant advantage of these approaches is that they may be used to generate alternatives and solutions for linear and nonlinear issues when mathematical models cannot easily describe the relationship between the problem's relevant factors (Gao et al. 2019) . Artificial intelligence approaches for evaluating and predicting the mechanical characteristics of cement-based materials are a hot topic in the cement-based composites research field to provide the construction industry implementation new methods and techniques. Moreover, some researchers have applied machine learning methods to evaluate and forecast the compressive strength of different concrete types. In this regard, multi-linear regression analysis and ANN models are extensively used among various dissimilar models to estimate the CS of concrete .

A previous study conducted by Duan et al. (2013) used 168 datasets from the literature to forecast the CS of recycled aggregate concrete, and they found that an artificial neural network (ANN) is a promising algorithm to predict the CS of recycled concrete using various mixtures proportion and types of recycled aggregates. In the same context, Deshpande et al. (2014) revealed that the ANN model performed better than the model tree and linear regression analysis in predicting the CS of concrete with recycled aggregate concrete. Mohammed et al. (2020a) established multiscale approaches to simulate the CS of concrete containing high quantities of fly ash. In their study, 450 samples were utilized for modeling. The qualifications were developed using five distinct modeling techniques (Linear regression, nonlinear regression, Multi-logistic regression ANN, and M5P-tree). They depicted that the M5P-tree and ANN models could forecast the CS very well in terms of lower RMSE and MAE values and higher R 2 values. A similar approach was used to predict the effect of large volume fly ash on the CS of cement-based mortars at various curing periods and w/c ratios, with findings comparable to the prior work (Salih et al. 2020) . Mohammed et al. (2020b) also employed ANN and nonlinear approaches to predict the rheological performance and CS of nano clay-modified cement paste. They concluded that the nonlinear approach is the best-performed method to estimate the flow behavior and CS of cement paste, and it outperforms the ANN model. They also stated that, among numerous independent factors, nano clay content was the most critical factor in determining CS and cement paste rheological behavior.

Despite the widespread usage of RP aggregates in the SCC, studies regarding the prediction of CS of SCC incorporated RP aggregates are very scarce to be efficiently implemented by the construction industry. Furthermore, the construction industry's growing desire for innovative building materials with unique characteristics to extend the service life of concrete structures necessitates the development of creative models for forecasting the behavior of these new materials. As a result, the main goal of this study is to assess and quantify the impact of a broad range of mixture proportions on the CS of SCC from an early age (7 days) to late curing (400 days), including RP aggregates content, binder content (Limestone powder or GGBFS or fly ash or silica fume or their combination), natural fine and coarse aggregate content, w/b ratio, and superplasticizer (SP) content. Using 400 data from past investigations, several model approaches such as multi-logistic, ANN, linear, and nonlinear regression models were used to predict the CS of SCC, including RP aggregates.

The current study aims to develop, describe, and provide multiscale models for predicting the CS of sustainable and eco-friendly SCC containing RP aggregates. Extensive experimental data, including 400 tested results with varying RP aggregate concentrations, w/b ratios, and curing regimes, were utilized in addition to the various modeling methods to achieve the following goals: (i) to perform statistical analysis and investigate the impact of mix ingredients such as RP aggregates, natural coarse and fine aggregates, binder, and SP dosage, as well as the curing time and w/b ratio on the CS of green SCC produced with RP aggregates and different by-product ashes; (ii) to ensure that the building sector may use the generated models without the need for any laboratory testing or analytical constraints; (iii) to choose and evaluate the accurate model to forecast the CS of sustainable SCC incorporated RP aggregates among various models (ANN, nonlinear, linear, and multi-logistic models) using different statistical evaluating tools. Furthermore, the study's primary contribution is that it provides mathematical models to forecast the CS of a new composite type, such as SCC incorporated RP aggregates, to be used efficiently by the construction industry.

Totally 400 experimental data from earlier studies were gathered and statistically evaluated before being divided into three groups. The first and more extensive group comprised 280 datasets utilized to create the models. Each with 60 data points, the second and third groups were utilized to test and validate the models (Qadir et al. 2019; Shariati et al. 2020) . Table 1 summarizes the SCC mixes database and the measured CS of SCC produced with various mix proportions and RP contents. During the first step of data extraction, the contents of cement, GGBFS, LP, silica fume, and fly ash were extracted separately, and then they were combined to form a binder content as one input variable. Moreover, Table 2 shows all 400 sample data extracted to develop the models. Several databases' search portals, such as Google Scholar, Web of Science, Scopus, and Science Direct, were utilized to conduct a comprehensive literature search as part of the database preparation process. The majority of prior studies discussing the effect of RP aggregates on the characteristics of SCCs were collected, and their data were retrieved based on the authors' searches. Table 2 shows the input data set, which includes the content of RP aggregates by weight, binder (B) content, w/b ratio (%), SP content, natural fine aggregate (FA) content, natural coarse aggregate (CA) content, curing time (t) by days, and measured CS (MPa). The given data set, which included the above-mentioned independent factors, was utilized to forecast the CS of SCC produced with RP using several approaches compared to the measured reported CS (MPa). Figure 1 depicts the procedure used in this investigation in terms of a flowchart. In addition, the following sections describe and explain the specifics, such as data gathering, analysis, modeling, and assessment.

In this part, a statistical study was performed to determine whether or not there are substantial correlations between input parameters and CS of SCCs. The statistical analysis was performed to illustrate that the CS of green SCC is affected by all mixture proportions, and a single input parameter cannot be used to estimate the CS directly. To do so, all input variables such as (i) RP aggregates content ( Fig. 2) , (ii) binder content (Fig. 3) , (iii) w/b (Fig. 4) , (iv) curing time (Fig. 5) , (v) SP content (Fig. 6) , (vi) natural FA content (Fig. 7) , and (vii) natural CA content (Fig. 8 ) were plotted and analyzed with actual CS; additionally, the normal distribution of obtained CS from previous studies is shown in Fig. 9 . All the figures mentioned previously demonstrated that strong correlations among CS and various input variables did not exist due to the low R 2 values for all correlations. In addition, statistical functions such as minimum, maximum, average, standard deviation, skewness, kurtosis, and variance were calculated and displayed in Table 3 to show the distribution of each variable. Regarding the kurtosis parameter, a high negative value represents the shorter tails of the distribution relative to the normal distribution, and a positive value represents the longer tails. A large negative value for the skewness parameter indicates a long left tail, while a positive value indicates a right tail.

The rheological and strength behavior of SCC mixtures is directly related to the type, shape, and surface texture of RP aggregates and the type of natural aggregate it replaced. Therefore, the type and properties of RP aggregates for all previous studies used for developing the models were extracted and presented in Table 4 . It is evident from the table that the RP aggregates are mostly incorporated into the SCC to replace natural fine aggregates. Ordinary Portland cement (OPC) Type I was the most common type used in producing SCC mixes, and it met the ASTM C 150 standard. With a specific gravity of 3.05 to 3.2, the fineness was in the range of 300 m 2 /kg to 400 m 2 /kg. Different byproduct waste ashes such as LP, GGBFS, silica fume, and fly ash with different proportions and properties, as shown Soft computing techniques to predict the compressive strength of green self-compacting concrete… Soft computing techniques to predict the compressive strength of green self-compacting concrete… Soft computing techniques to predict the compressive strength of green self-compacting concrete… Soft computing techniques to predict the compressive strength of green self-compacting concrete… The models provided in this work are utilized to forecast the CS of SCC and choose the best one that provides a superior estimate of CS compared to the reported CS from the original data. The following evaluation criteria were used to compare the forecasts of different models: The model had to be scientifically accurate, have a minor percentage error between observed and forecasted data, and have a higher R 2 value with lower RMSE, Objective (OBJ), MAE, and SI values.

The linear regression model (LR) (Zain and Abd 2009), as illustrated in Eq. (1), is the most general technique for predicting the CS of concrete:

where a and b r c and w = c , respectively, denote equation parameters, CS, and water/cement ratio. Other components and variables of SCC mixes incorporated RP aggregates, including curing time and other mix contents, are not included in the previous formula, although they impact the CS. Equation (2) is presented to incorporate all different mix proportions and factors that may affect CS to obtain more reliable scientific findings.

where RP stands for recycled plastic aggregate content (kg/ m 3 ), B stands for binder content (kg/m 3 ), w/b stands for water to binder ratio, t stands for curing time, SP stands for superplasticizer dosage, FA stands for natural fine aggregate, and CA stands for natural coarse aggregate. 

The following Eq. (3) can be implemented to build a nonlinear model in general Sarwar et al. 2019) To forecast the CS of normal SCC mixes and SCCs incorporated RP aggregates, the connection among various variables in Eqs. (1) and (2) can be expressed in Eq.

(3) to forecast the CS.

where B stands for the binder content, w/b stands for the water to binder ratio, t stands for curing time, SP stands for the superplasticizer dosage, FA stands for natural fine aggregate, CA stands for natural coarse aggregate, and RP stands for the recycled plastic content. Moreover, the model parameters are a, b, c, d, e, f, g, h, I, j, k, l, m, n, and o, calculated based on the least square method.

The MLR, also a regression procedure, can be employed when the expectable variable has a parameter greater than Fig. 1 The flowchart diagram process followed in this study Soft computing techniques to predict the compressive strength of green self-compacting concrete… two stages. MLR is a statistical approach that is comparable to multiple linear regressions. Equation (4) can be utilized to find the variance among predictable and independent variables.

Equation (4), on the other hand, has a drawback in that it cannot be used to forecast the CS of SCC without RP. Thus, the RP content in this model should be larger than zero (RP content [ 0%). The least-square method was implemented to find the parameters (a, b, c, d, e, f, g, and h) and model variables.

ANN is a powerful simulation software designed for data analysis and computation to think like a human brain in processing and analyses. This machine learning tool is widely used in construction engineering to predict several numerical problems' future behavior (Sihag et al. 2018; Mohammed 2018 ). The ANN model is generally divided into three main layers: input, hidden, and output. Each input and output layer can be one or more layers depending on the proposed problem. However, the hidden layer is usually ranged for two or more layers. Although the input and output layers generally depend on the collected data and the designed model purpose, the hidden layer is determined by rated weight, transfer function, and the bias of each layer to other layers. A multi-layer feed-forward network is built based on a mixture of proportions, weight/ bias, several parameters, including (RP, B, w/b, t, SP, FA, and CA) as inputs, and output ANN here is the CS of SCC. There is no standard approach to designing the network architecture. Therefore, the number of hidden layers and neurons is determined based on a trial and error test. One of the main objectives of the training process of the network is to determine the optimum number of iterations (epochs) that provide the minimum mean absolute error (MAE), root means square error (RMSE), and best R 2 -value that is close to one. The effect of several iterations on reducing the MAE and RMSE has been studied. The collected data set (a total of 400 data) has been divided into three parts for the training purpose of the designed ANN. About 70% of the collected data was used as trained data for training the network. The 15% of overall data was used to test the data set, and the rest of the remaining data was used to validate the trained network . The designed ANN was trained and tested for various hidden layers to determine optimal network structure based on the fitness of the predicted CS of SCC containing RP aggregates with the CS of the actual collected data. It was observed that the ANN structure with three hidden layers, ten neurons, and a hyperbolic tangent transfer function (as shown in Fig. 10 ) was a best-trained network that provides a maximum R 2 and minimum both MAE and RMSE (shown in Table 5 ). The General Equation of the ANN model is shown in Eqs. 5, 6, and 7.

From linear node 0: Soft computing techniques to predict the compressive strength of green self-compacting concrete… From sigmoid node 1:

From sigmoid node 2:

Different evaluating metrics such as coefficient of determination (R 2 ), scatter index (SI), OBJ, root mean squared error (RMSE), and mean absolute error (MAE), were implemented to analyze and assess the effectiveness of the suggested models, which can be computed using the formulae below: 

From the formulas above, yp and tp are the expected and actual values of the path pattern, and t 0 and y 0 are the averages of the actual and forecasted values. Training, testing, and validating datasets are denoted as tr, tst, and val, respectively; also the number of patterns (collected data) in the associated dataset is denoted as n.

Except for R 2 , the optimum value for all evaluating factors is zero; nevertheless, R 2 has the optimum value of one. When it comes to the SI parameter, a model has (bad Soft computing techniques to predict the compressive strength of green self-compacting concrete… Fig. 10 The best ANN Model with three hidden layers and ten neurons performance) when it is [ 0.3, (fair performance) when it is between 0.2 and 0.3, (good performance) when it is between 0.1 and 0.2, and (great performance) when it is less than 0.1 (Li et al. 2013; Golafshani et al. 2020 ). In addition, the OBJ parameter was employed as an integrated performance parameter in Eq. (12) to measure the efficiency of the suggested models. Finally, positive and negative error margin lines were included in the model results to graphically illustrate how each model overestimates and underestimates the predicted results of CS compared to the actual values from the experiments. A positive value means the overestimated percentage of CS, while the negative value means the underestimated percentage of CS.

Relationships among forecasted and measured compressive strength

The LR model Figure 11a ,b, and c demonstrate the connection between estimated and actual CS of SCC mixes, including RP aggregates for all phases, including training, testing, and validating datasets. Based on the model parameters, the w/ b ratio and RP aggregate content substantially impact the compression strength of SCC incorporated RP aggregates. By optimizing the sum of error squares and the least square approach, which was performed in Excel using Solver to obtain the ideal value (a given value, minimum or maximum) for the Equation, the present model's weight of each parameter on the compression strength of SCC Soft computing techniques to predict the compressive strength of green self-compacting concrete… incorporated RP aggregates was found. The values of other equation cells in the worksheet were used to set limitations or restrictions on this object cell (Burhan et al. 2020 ). The following is the Equation for the LR model with various weight parameters (Eq. 13):

The w/b ratio has the most impact on lowering the CS among all factors, as seen in the Equation above. This may be consistent with the experimental findings reported in the literature. This model's R 2 , RMSE, and MAE assessment parameters are 0. 79, 10.73 MPa, and 8.97 MPa, respectively. Furthermore, as shown in Figs. 20 , 21, the current model's OBJ and SI values for the training dataset are 11.52 and 0.196, respectively.

Figures 12a, b, and c show the forecasted vs. real CS derived from past investigations of SCC mixes incorporated RP aggregates for training, testing, and validating datasets, respectively. Due to this model, the w/b ratio, FA, and CA concentration are the most significant factors influencing the CS of SCC mixes. Many experimental programs from previous research supported this, as shown in Table 1 , in which lowering the w/b ratio and altering the quantity of aggregates had a substantial impact on the CS of SCCs containing RP aggregates. The following is the suggested formula for the NLR model with various variable parameters (Eq. 14): 

The R 2 , RMSE, and MAE assessment parameters for this model are 0.77, 11.05 MPa, and 9.28 MPa, respectively. Furthermore, the current model's OBJ and SI values for the training dataset are 11.44 and 0.2, respectively. Figures 13a, b , and c demonstrate the comparison of predicted CS against actual CS extracted from previous studies of SCC mixes made with RP aggregates for all phases (training, testing, and validating datasets). As shown in previous studies reported in Table 1 , the most influential parameter that impacts the CS of SCC mixes, including RP aggregates, is the w/b ratio, similar to other models. Equation 15 may be used to forecast the CS of 

The R 2 , RMSE, and MAE assessment parameters for this model are 0.78, 8.67 MPa, and 6.83 MPa, respectively. Furthermore, the current model's OBJ and SI values for the training dataset are 8.46 and 0.167, respectively.

The authors investigated various hidden layers, neurons, momentum, learning rate, and iterations to achieve high ANN efficiency. Finally, they discovered that when the ANN contains three hidden layers, ten neurons on each side (as shown in Fig. 10 

As previously stated, five different statistical methods were implemented to classify the efficacy of the developed models, including SI, MAE, R 2 , RMSE, and OBJ. Compared to the LR, NLR, and MLR models, the ANN has a higher R 2 and lower RMSE and MAE values, as shown in Figs. 15, 16 , and 17 for R 2 values RMSE, and MAE, respectively. Figure 18 also compares model CS estimates for SCC mixtures, including RP aggregates based on all data. In addition, Fig. 19 Figure 20 shows the OBJ performance for various suggested models. The LR, NLR, MLR, and ANN models have 11.52, 11.44, 8.46, and 4 .26, respectively. The ANN model had an OBJ value of 63 percent less than that of the LR and NLR models, and it is also 50 percent smaller than that of the MLR model. This also shows that the ANN approach is more robust to forecast the CS of SCC mixes, including RP aggregates. Figure 21 shows that, except for the ANN model, the SI values for all stages and models were between 0.1 and 0.23, indicating good performance. However, the SI values for the ANN model were between 0 and 0.1, indicating that the ANN model performed excellently. Furthermore, like the other performance factors, the ANN model has lower SI values than other approaches. In the training phase, the ANN model has a 49% lower SI than Soft computing techniques to predict the compressive strength of green self-compacting concrete… the LR and NLR, and 35% lower SI value in the testing phase, and a 65.42% smaller SI in the validating. Furthermore, compared to the MLR model, the ANN had lower SI values in all stages, such as 40.11% lower in training, 17.58% lower in testing, and 55.42% lower in the validating. This also confirmed that the ANN is more capable and accurate when predicting the CS of SCC mixes, including RP aggregates than the LR, NLR, and MLR models.

A sensitivity analysis was conducted for the models to classify and evaluate the input parameter that mainly impacts the CS prediction of SCC containing RP aggregates (Mohammed et al. 2020a) . To do so, the MLR model was selected because all the input variables should have a value greater than zero in this model, and it means that the effect of different variables on the CS prediction could be more evident. All training data were incorporated throughout this analysis, and each time a single input variable was extracted. Parameters for interpretation, such as RMSE, R 2 , and MAE, were determined separately.

When the previously mentioned assessment values are changed and decreased significantly, it indicates the effectiveness of the input parameter. Table 6 illustrates the outcomes of the sensitivity study. The findings show that similar to the normal SCC, w/b ratio and curing times (t) are the most significant and dominant parameters in forecasting the CS of the SCC mixes, including RP aggregates. In this research, the w/b ratio for the collected data varied from 0.25 to 0.45; increasing the w/b ratio dramatically reduces the CS of SCC mixes containing RP aggregates. Moreover, increasing the curing time caused a considerable enhancement in the CS of SCC mixtures. All experimental findings presented in Table 1 support this.

Environmental and economic assessment of SCC incorporating RP aggregates As previously mentioned, the continuous growth of plastic production caused a significant environmental threat. In 2020, despite the global pandemic (COVID-19), the world produced 367 million tons of plastics, of which 55 million tons were produced in Europe; of this, 23% ended up in the waste stream (Plastics Europe 2021). Figure 22 illustrates the world production of plastic from 2016 to 2021. It is evident from the figure that the annual production rate of plastic was increased significantly, with an increase of 10% since 2016. Besides, many countries' landfill areas are restricted due to population growth, necessitating the search for a new way to dispose of plastic waste. Moreover, the use of natural coarse and fine aggregates in the development of new projects depletes natural resources, and natural aggregate utilization and extraction have a detrimental influence on the ecosystem. As a result, the concrete industry should establish a new strategy for sustainable development using this waste material as an aggregate. One of the potential uses of RP is its incorporation in SCC as a partial or total substitution of natural aggregates. The utilization of RPs from various products in the concrete industry is an attractive and safe mode of disposal that can resolve the majority of environmental issues caused by plastic wastes. The RP aggregates give SCC unique characteristics not found in normal SCC, such as lower unit weight and better ductility. The partial replacement of natural aggregates with RP aggregates reduces the density of SCC due to the low specific gravity of RP aggregates compared to the natural aggregates; thus, the possibility of producing lightweight SCC is higher. The benefits of lightweight SCC incorporating RP aggregates over normal SCC include the reduced size of the structural components, reduction in dead load of the structure, improved dynamic load response, more energy absorption and toughness, less reinforcing steel, greater design flexibility, and an overall decrease in the construction costs (Basha et al. 2020) . Another significant change when RPs are included in SCC is improved thermal insulation, one of the essential criteria for energy saving. A previous study reported that a retail building constructed with concrete that incorporated a high amount of RPs was shown to use 40% less energy for heating and cooling than a similar building made with normal concrete (Elzafraney et al. 2005) .

Accurate and dependable models to forecast the CS may result in considerable cost and time savings for the construction industry. It is possible to draw the following findings due to the analysis and modeling using the data gathered from prior research to forecast the CS of SCC mixes, including RP aggregates and different by-product ashes at 400 various mixed proportions:

• In the production of SCC mixes, the average proportion of RP utilized was 17.035 kg/m 3 . Furthermore, the proportion of RP aggregates replacement with natural aggregates (Fine or coarse aggregates) varied from 0 to 138 kg/m 3 . The curing period for data from various experimental investigations varied from 7 to 400 days. • The LR, NLR, MLR, and ANN models were established in this research to forecast the compression strength of SCC mixes. According to the various evaluation criteria, the ANN model outperformed others with higher R2, smaller OBJ value, smaller RMSE, smaller SI, and less MAE values for all phases. • Except for the ANN model, the SI values for all models and stages were between 0.1 and 0.23, suggesting good performance. However, the SI values for the ANN model were between 0 and 0.1, suggesting that the ANN model performed excellently. • The OBJ value of the ANN model is 63% smaller than that of the LR and NLR models, and it is also 50% smaller than the MLR model. This also shows that the ANN is more accurate and capable of forecasting the CS of sustainable SCC mixes, including RP aggregates. • Concerning the curing time, it was discovered via a sensitivity analysis that it is the most significant parameter for predicting the CS of SCC mixes, including RP aggregates and different by-product ashes.

Increasing the curing period of SCC mixes with or without RP aggregates resulted in a substantial improvement in the CS. • The overall findings and analysis showed that specific quantities of RP aggregates and industrial by-product ashes might be utilized effectively in green SCC manufacture to affect the strength and other aspects of SCC favorably. On the other hand, the optimal RP content is highly dependent on the shape, type, and surface texture of RP aggregates.

Funding This research received no funding.

Conflict of interest The author has no conflicts of interest to declare that are relevant to the content of this article.

Data availability All data are provided in the manuscript.

Ethical approval The contents of this manuscript are not now under consideration for publication elsewhere; The contents of this manuscript have not been copyrighted or published previously. The contents of this manuscript will not be copyrighted, submitted, or published elsewhere, while acceptance by the Journal is under consideration.

Consent to participate Not applicable.

Consent for publication Not applicable. Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Systematic multiscale models to predict the compressive strength of fly ash-based geopolymer concrete at various mixture proportions and curing regimes

Normal and high-strength lightweight selfcompacting concrete incorporating perlite, scoria, and polystyrene aggregates at elevated temperatures

Development of highperformance self-compacting concrete using waste recycled concrete aggregates and rubber granules

Mechanical and thermal properties of lightweight recycled plastic aggregate concrete

Enhancing the fresh and hardened properties of the early age concrete modified with powder polymers and characterized using different models

Rheological and strength behavior of binary blended SCC replacing partial fine aggregate with plastic E-waste as high impact polystyrene

An experimental study on the durability and strength of SCC incorporating FA, GGBS and MS

Artificial neural network and nonlinear models for gelling time and maximum curing temperature rise in polymer grouts

Modeling compressive strength of recycled aggregate concrete by artificial neural network, model tree and nonlinear regression

Prediction of compressive strength of recycled aggregate concrete using artificial neural networks

Development of energyefficient concrete buildings using recycled plastic aggregates

Mechanical, fracture and durability properties of self-compacting high strength concrete containing recycled polypropylene plastic particles

Systematic multiscale models to predict the compressive strength of self-compacting concretes modified with nanosilica at different curing ages

Development of a novel soft-computing framework for the simulation aims: a case study

Statistical variations and new correlation models to predict the mechanical behaviour of the cement mortar modified with silica fume

Predicting the compressive strength of normal and high-performance concretes using ANN and ANFIS hybridized with grey wolf optimizer

Strength and drying shrinkage properties of self-compacting concretes incorporating multi-system blended mineral admixtures

Compressive strength and microstructural characteristics of class C fly ash geopolymer

Influence of polyethylene waste on some fresh and mechanical properties of selfcompacting concrete

Lightweight selfcompacting concrete incorporating perlite, scoria, and polystyrene aggregates

General models for estimating daily global solar radiation for different solar radiation zones in mainland China

Durability and compressive strength of high cement replacement ratio self-consolidating concrete

Production and optimization of eco-efficient self-compacting concrete SCC with limestone and PET

Artificial Neural Network and NLR techniques to predict the rheological properties and compression strength of cement past modified with nanoclay

Soft computing techniques: systematic multiscale models to predict the compressive strength of HVFA concrete based on mix proportions and curing times

Experimental and statistical evaluation of rheological properties of self-compacting concrete containing fly ash and ground granulated blast furnace slag

Properties of concrete'', five, and final Edition

Plastics-the facts 2021. An analysis of european plastics production, demand and waste data

Characterizing and modeling the mechanical properties of the cement mortar modified with fly ash for various water-to-cement ratios and curing times

Strength and durability assessment of self-compacted lightweight concrete containing expanded R

The combined effects of waste polyethylene terephthalate (PET) particles and pozzolanic materials on the properties of self-compacting concrete

Waste polyethylene terephthalate as an aggregate in concrete

Systematic multiscale models to predict the effect of highvolume fly ash on the maximum compression stress of cementbased mortar at various water/cement ratios and curing times

Modeling the rheological properties with shear stress limit and compressive strength of ordinary Portland cement modified with polymers

Feasibility study on the use of tagouk ash as pozzolanic material in concrete

A novel hybrid extreme learning machine-grey wolf optimizer (ELM-GWO) model to predict compressive strength of concrete with partial replacements for cement

Modelling of impact of water quality on recharging rate of storm water filter system using various kernel function based regression

Utilization of waste glass in translucent and photocatalytic concrete

Performance of self-compacting concrete containing different mineral admixtures

Properties of self-compacting lightweight concrete containing recycled plastic particles

Application of nanomaterials in alkali-activated materials

Multiple regression model for compressive strength prediction of high: performance concrete