Compressive Strength of Geopolymer Concrete Composites: Modeling and Comprehensive Systematic Review


 The growing concern about global climate change and its adverse impacts on societies is putting severe pressure on the construction industry as one of the largest producers of greenhouse gases. Given the environmental issues associated with cement production, geopolymer concrete has emerged as a sustainable construction material. Geopolymer concrete is cementless concrete that uses industrial or agro by-product ashes as the main binder instead of ordinary Portland cement; this leads to being an eco-efficient and environmentally friendly construction material. Compressive strength is one of the most important mechanical property for all types of concrete composites including geopolymer concrete, and it is affected by several parameters like an alkaline solution to binder ratio (l/b), fly ash (FA) content, SiO2/Al2O3 (Si/Al) of the FA, fine aggregate (F) and coarse aggregate (C) content, sodium hydroxide (SH) and sodium silicate (SS) content, ratio of sodium silicate to sodium hydroxide (SS/SH), molarity (M), curing temperature (T), curing duration (CD) inside the oven and specimen ages (A). In this regard, a comprehensive systematic review was carried out to show the effect of these different parameters on the compressive strength of the fly ash-based geopolymer concrete (FA-GPC). In addition, multi-scale models such as Artificial Neural Network (ANN), M5P-tree (M5P), Linear Regression (LR), and Multi-logistic Regression (MLR) models were developed to predict the compressive strength of FA-GPC composites. For the first time, in the modeling process, twelve effective parameters including l/b, FA, Si/Al, F, C, SH, SS, SS/SH, M, T, CD, and A were considered the modeling input parameters. Then, the efficiency of the developed models was assessed by various statistical assessment tools like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Scatter Index (SI), OBJ value, and the Coefficient of determination (R2). Results show that the curing temperature, sodium silicate content, and ratio of the alkaline solution to the binder content are the most significant independent parameters that influence on the compressive strength of the FA-GPC, and the ANN model has better performance for predicting the compressive strength of FA-GPC in compared to the other developed models.


Introduction
It is widely known that the production of Portland cement needs a considerable amount of energy and at the same time contributes to generating a huge volume of the total carbon dioxide (around 7%) to the atmosphere directly and indirectly, the heating of limestone releases CO2 directly which is called calcination (50%), while the burning of fossil fuels to heat the kiln indirectly results in CO2 emissions, this is around 40 percent of cement emissions and finally around 10% for quarrying and transporting [1][2][3]. In addition, around 2.8 tons of raw materials are needs for the manufacture of one ton of cement; this is a resource-exhausting process that consumes a large number of natural resources such as limestone and shale for the production of clinkers for cement [4]. Furthermore, approximately one trillion liters of mixing water is required to be used in the concrete industry annually [5]. In the same context, the cement industry is the most energy-intensive construction material after aluminum and steel manufacture. In a typical cement plant alone, around 110-120 kWh is used for each ton of produced cement [6]. However, cement-based concrete is still the main important material used in the construction industries worldwide [7]. Therefore, a highly efficient application of renewable and nonrenewable raw materials is essential for the economic development [8]. Sustainable development of a novel material to replace the Portland cement has become increasingly substantial as the globe goes on to face serious environmental problems [9,10].
A convenient, suitable replacement to conventional concrete is geopolymer technology that was developed first by Davidovits in France, 1970 [11]; the ancient Roman civilizations have used geopolymer for building their monumental and castles in ancient times [12]. Geopolymers are members of the family of inorganic alumino-silicate polymer synthesized from alkaline activation of various aluminosilicate materials or other industrial or agro by-product materials rich in silicon and aluminum like fly ash (FA), ground granulated blast furnace slag (GGBFS), metakaolin (MK), palm oil fuel ash (POFA), rice husk ash (RHA) [13]. The microstructure of geopolymer materials is amorphous, and their chemical constituents are similar to the natural zeolitic materials [14]. The chemical reaction between alkali solution and base material containing aluminosilicate is called geopolymerization process; the final product of the geopolymerization process gives a three-dimensional polymeric chain and ring structure Si-O-Al-O bonds as shown in the Scheme.1. [15], with an empirical formula of {Mn[-(SiO2)z-AlO2]n.wH2O}, when; M is an alkali action, n is the percent of polymerization, and w is the content of water [16]. In addition, the chains in aluminosilicate could be in the form of poly-(sialate) with the ratio of Si to Al is equal to 1.0 (-Al-O-Si-chain), poly (sialate-siloxo) with the ratio of Si to Al is equal to Fig.1. the schematic graphical representation that shows the transition of fly ash to fly ash-based geopolymer cement/concrete [53] One of the most important mechanical properties of all types of concrete composites, including FA-GPC, is the compressive strength (fc′), and it can be found by following the standard test methods of ASTM C39 [54] BS EN 12390-3 [55]. fc′ gives a general performance about the quality of the concrete [56]. However, in the structural design and construction field, the compressive strength of the concrete at 28 days is essential. Therefore, achieving an authoritative model for predicting the compression strength of GPC is essential regarding the possibility of changing or validating the GC mix proportions [57,58]. By selecting appropriate mixing proportions, economic and efficient designs will be accomplished. Therefore, various researches have been tried to shorten the time of choosing a proper mix of proportions to get the targeted properties; among them is modeling with developing empirical equations. There are different ways to model construction materials' characteristics, including statistical techniques, computational modeling, and nowadays developed techniques such as regression analysis [59,60]. A variety of factors affect the fc′ of the FA-GPC; this leads to different compression strength results; consequently, predicting fc′ is a challenging task for researchers and engineers. Therefore, there is a need for numerical and mathematical models [61]. Due to the good ability of machine learning regarding prioritization, optimization, forecasting and planning were widely used in the various engineering fields [57]. In the literature, machine learning systems were used to model the various characteristics of different types of concrete composites such as compression strength of green concrete [62], splitting tensile and flexural strength of recycled aggregate concrete [63], modulus of elasticity of recycled concrete aggregate [64,65], the fc′ of high volume fly ash concrete [66], the fc′ of eco-friendly GPC containing natural zeolite and silica fume [67], splitting tensile strength of fiber-reinforced concrete [68], the fc′ of self-compacting concrete modified with nanosilica [69], and so on.
In the literature, there is a lack ofsystematic and comprehensive reviews to show and measure the effects of several mixture proportion parameters and different curing regimes on the compressive strength of FA-GPC. Therefore, in this study, for the first time influence of several parameters like an alkaline solution to binder ratio (l/b), fly ash (FA) content, SiO2/Al2O3 (Si/Al) of the FA, fine aggregate (F), and coarse aggregate (C) content, sodium hydroxide (SH) and sodium silicate (SS) content, the ratio of sodium silicate to sodium hydroxide (SS/SH), molarity (M), curing temperature (T), curing duration (CD) inside the oven and specimen ages (A) were investigated to show their influences on the fc′ of the FA-GPC, and then multi-scale models such as Artificial Neural Network (ANN), M5P-tree (M5P), Linear Regression (LR), and Multi-logistic Regression (MLR) models were developed to predict the fc′ of the FA-GPC composites at different mixture proportions.   (iv) to discover the most authoritative model to predict the compression strength of FA-GPC from four  Table 3. The models used twelve input parameters, which not allowed authors to use more datasets in the developed models. For instance, those researches were ignored if the mix proportions and any other model parameters of the research were not provided. The larger group of a dataset, which included 340 datasets, was used to create the models. The second group consists of 85 datasets used to test the proposed models, and the last group, which includes 85 datasets, was used to validate the provided models [57,69]. The data collection, comprehensive review, and modeling work are summarized in a flow chart, as depicted in Fig.2.

Chemical composition of fly-ash (SiO2/Al2O3) (Si/Al)
In the literature, a variety of fly ashes with slightly different chemical compositions, specific surfaces and specific gravity were used to prepare the GPC. Based on the ASTM C618 [29], ash with the summation of their SiO2+Al2O3+Fe2O3 greater than 70% can be known as fly ash. To investigate the effect of mixed compositions on the fc′ and microstructure of FA-GPC an experimental laboratory research work has been conducted by Thakur and Ghosh [125]. They reported that the compression strength was improved almost linearly with Si content up to 4, and then it was decreased, as illustrated in Fig.3. This result was attributed to the lower content of CaO and low activity of fly ash [126,127]. MPa strength increment between three and 365 days [102]. These results revealed that there is a continuation in the geopolymerization process which leads to improvement in the compressive strength of different types of FA-GPC. This output is in contrast to the past studies on FA-GPC, which have recorded little subsequent improvement in the fc′ for the heat curing conditions [70,128,129].
In addition, another research study had been carried out on the mechanical characteristics of geopolymer concrete which they used three different fly ashes with different chemical compositions. In the controlled heat curing conditions and constant mixture proportions, fc′ values were 32.1, 41, and 36.6 MPa for the Si/Al of the fly ash of 2.9, 2.78, and 2.6, respectively [104]. Scanning Electron Microscopy (SEM) examined the results behind different compressive strength values. They reported that the samples made from the fly ash with Si/Al of 2.78 have the highest degree of reacted fly ash spheres, which participated in the greatest compression strength. This result was reported by other researchers who claimed that the total reacted Si/Al is crucial to the progress in the polymerization of the FA-GPC process [70,128]. In addition, it was observed that with the increase of Si/Al ratio, compression strength was improved [80,130]. Similar observations have been made by Thokchom et al. [131], who founded that with the increase in Si/Al ratio of binder source materials, residual compressive strength of the GPC was improved when the specimens exposed to different temperature degrees, as shown in Fig. 4. The GPC with 2.2 ratios of Si/Al still retained nearly 63% of its compression strength even after exposure to 900 ℃, while the GPC mixture prepared with 1.7 ratios of Si/Al indicated the residual compressive strength of 50%. This result was further investigated by SEM tests which revealed that the GPC specimens prepared with 1.7 Si/Al ratio have entire disruption of the matrix due to sintering of phases inside the samples; however, interconnected matrix the most of the initial pores being blocked in the heating process. While GPC made with Si/Al of 2.2 presents a relatively undisturbed matrix except at little places [131].
Lastly, an experimental research program was performed to investigate the mechanical properties of high early strength FA-GPC. They used five different fly ash source materials with various Si/Al ratio and CaO content. Their results indicated that the highest fc′ was recorded for a mixture with 2.18 of Si/Al and 4.96 CaO content [107], and from the

The alkaline solution to the binder ratio (l/b)
The alkaline solution is the summation of the sodium hydroxide and sodium silicate content, while the binder content is the total weight or volume of the fly ash or other source binder materials in the mixture proportions of the GPC. Based on the findings of Aliabdo et al. [79], when the chemical admixture content, additional water content, SS/SH, and M were kept constant at 10.5 Kg/m 3 , 35 Kg/m 3 , 0.4, and 16, respectively, the fc′ of the FA-GPC was increased with the increment in the l/b ratio up to 0.4; then the effect reverses as shown in Fig.6. This increment in the compression strength of the FA-GPC was 52%, 78%, and 68% for mixes with 0.35, 0.4, and 0.45 alkaline liquid to fly ash ratio, correspondingly, as compared with that mix that has 0.3 alkaline liquid to fly ash ratio. Similar observations have been made by Shehab et al. [74], who found that the higher percentage of l/b improves the fc′ of the FA-GPC at the age of 7 and 28 days. In the same context, experimental work was performed to investigate the effect of different parameters on the mechanical properties of FA-GPC, and they noticed that the fc′ was significantly increased as the l/b increased up to 0.55 and beyond that negatively affected. The respectively when the other variable parameters are constant at 100 ℃ curing condition. Also, they observed that when the curing temperature is 80 ℃ and 60 ℃, the compressive strength was the same as before, which improved up to 0.55 of l/b ratio [98].
In contrast to the results mentioned above, some studies reported that with increasing l/b ratio will cause a reduction in the fc′ of the FA-GPC, for instance, an experimental research study have been performed to investigate the behavior of low-CaO fly and bottom ash-based geopolymer concrete with different l/b ratio cured at ambient temperature. They observed that the fc′ of the geopolymer concrete at 28 days is 18.8, 27.2, and 34.3 MPa at l/b ratio of 0.5, 0.35, and 0.3, respectively [80]. This result argued to the fact that water content in the reaction medium of the GPC mixture was increased as any increase in the l/b ratio which leads to decreasing friction between the particles and, as a consequence reducing the fc′ of the GPC [133]. However, the same GPC with l/b ratio of 0.25 is the only exception to this general trend, as a slight reduction in the fc′ was actually observed due to the lack of suitable workability of the GPC mixture in the fresh state, which caused placement problems during concrete casting compared to the other l/b ratio mixtures, therefore, as a result, affected the compressive strength negatively [80].
Similar results can also be found in other studies even though the different alkaline solution to binder ratio was used [26,77,114]. In the same context, Fang et al. [118] claimed that the increment of l/b ratio strongly influences the early-age compression strength of FA-GPC but no significant effect on the later ages of the GPC. This result argued that decreasing l/b ratio will lead tohat decreasing l/b ratio accelerating in the alkaline activation process of fly ash-slag geopolymer concrete due to the decrease of consistency GPC mixture [134]. In this case, the Calcium Aluminate Silicate Hydrate (C-A-S-H) gel and Sodium Aluminate Silicate Hydrate (N-A-S-H) gel can be generated quickly in the geopolymer concrete mixture with low l/b ratio and, as a consequence, participated in the development of early-age compressive strength of fly ash-slag geopolymer concrete [135,136].
Lastly, it was observed that the compressive strength decreased with increasing in the l/b ratio for the lower sodium hydroxide concentrations (molarity) cured at ambient temperature; for instance, when the molarity is 8 M, the compression strength was 11, 7.6, and 7.5 MPa at l/b ratio of 0.4, 0.5, and 0.6, respectively. However, when the molarity increased to 14 M, the fc′ of FA-GPC improved to 18.1, 21.5, and 21.5 MPa, at 0.4, 0.5, and 0.6 l/b ratios, correspondingly [117], as shown in Fig.7. Furthermore, the increase in the l/b ratio from 0.5 to 0.6 results in the decrease in the fc′ of the GPC by about 6.1%, 8.8%, 13.8%, 22.2%, and 14% for the molarity of the sodium hydroxide concentration of 8, 10, 12, 14, and 16 M, correspondingly. This result argued to the fact that increasing molarity of the alkaline liquid will lead to the presence of more solid part compared to the water content, which significantly influences the geopolymerization process and, as a consequence, compression strength was improved [137].

Superplasticizer Dosage and Extra Water
Superplasticizer and water content are two key factors that govern the workability behavior of the FA-GPC and hence affect the hardened characteristics of the geopolymer concrete. The alkaline solution, which is consists of sodium hydroxide and sodium silicate, is more viscous than water; hence their use in the GPC makes the mixture more sticky and cohesive than the traditional concrete [138]; therefore, extra water and superplasticizer are used to improve workability in GPC mixture.
To investigate the effect of superplasticizer dosage on the fc′ of FA-GPC, Hardjito et al. [139] set an experimental program using different dosages of high-range water-reducing admixtures. Their results revealed that workability of the FA-GPC was improved with the inclusion of superplasticizer on the one hand; on the other hand, superplasticizer addition to the geopolymer, concrete mixture has very little effect on the compression strength up to nearly 2% of fly ash by mass as shown in Fig.8. After 2% of superplasticizer dosage, the value of compressive strength was decreased with increasing the superplasticizers dosage; for instance, the fc′ was decreased by 19% when the dosage of superplasticizer increased from 2% to 3.5%. Moreover, according to the findings of this research study [139], with the increment of extra water content to the GPC mixture, compressive strength was significantly declined at different curing temperatures. This result is same as the results of Barbosa et al. [140] who their works were carried out on the geopolymer concrete pastes. Furthermore, experimental research work has been conducted to investigate the influence of extra water addition on the workability and compressive strength of FA-GPC. Their results indicated that with increasing the amount of extra water content to the geopolymer concrete mixture, the workability was increased and improved, but the fc′ was decreased. For instance, the fc′ was decreased by 32%, 42%, and 71% when the extra water content increased from 0.25 to 0.30, 0.35, and 0.40, respectively [100].
This result attributed to the fact that the evaporation of the water from the GPC, leaving pores and cavities within the geopolymer concrete matrix, when the geopolymer concrete specimens curing at the high temperature inside ovens, and present the extra water may influence the alkalinity environment of the FA-GPC matrix that could cause decreasing the rate of the geopolymerization process between the alkaline materials and source material (fly ash) of the FA-GPC composites [71]. In the same context, a research study has been done to study the effect of extra water and superplasticizer content on the fc′ of the FA-GPC. They conclude that with the increment of extra water content in the geopolymer concrete mixture, compression strength was decreased as shown in Fig.9 at the age of 7 and 28 days, this decline in the compressive strength was not greater than 10% up to 30 kg/m 3 of extra water, while, reduction in the compressive strength was increased to 24% due to use of 35kg/m 3 of extra water as compared to 10 kg/m 3 of extra water content for the FA-GPC [79]. Similarly, they noticed that the fc′ of the FA-GPC was slightly decreased as the superplasticiser content increased. For instance, they reported that the value of the compressive strength was decreased by 4.2%, 8.6%, and 24% for the FA-GPC with 5, 7.5, and 10.5 kg/m 3 superplasticiser dosage compared to the same mixture with 2.5 kg/m 3 admixture content [79]. Furthermore, Josef and Mathewb [98] reported that the fc′ of fly ash-based geopolymer concrete is decreased with the increment of water to geopolymer concrete solids ratio; this reduction in the compressive strength is nearly linear for all values of l/b ratio as presented at Fig.10.  Lastly, experimental research work has been carried out to investigate the effect of superplasticizer and water to binder ratio on the fc′ of FA-GPC. They observed that with increasing the superplasticizer dosage and water to binder ratio, the compressive strength was significantly decreased. For example, the fc′ was declined by 41% and 50% at 13.92 and 48 kg/m 3 superplasticiser content, respectively, compared to 8.64 kg/m 3 content of superplasticiser. Moreover, the fc′ was decreased to 46.2 and 27.2 MPa at 45.6 and 60 kg/m 3 extra water content, correspondingly, compared to the 40.8kg/m 3 of extra water content [105]. Similar results have been made by Vora and Dave [114], who observed that the compression strength of FA-GPC was decreased as superplasticiser and water to binder ratio of the concrete mixture increased. This is attributed to the fact that the extra water in the GPC mixture leads to generate large gel crystals with trapped water inside, and then, once the entrapped water evaporated in the mixture, it produces a highly porous matrix, as a consequence, it causes a decrease in the fc′ and increases the absorption capacity of the GPC [141]. Overall, high amount of water and superplasticiser content than the solid parts in the geopolymer concrete mixture cause to a reduction in the compression strength of the GPC composite as a result of decreasing contact among activating solution and the source reacting material [26,142].

Fly ash content
Fly ash (FA) is widely used as a source material for making geopolymer concrete due to its low cost, abundance availability, and higher potential for preparation geopolymers [25,26]. The content of fly ash in the mixture proportions of different fly ash-based geopolymer concrete for the collected data varied from 254 to 670 kg/m 3 . The FAs have different chemical compositions with different specific gravity range from 1.95 to 2.54.
An experimental laboratory research work has been carried out to investigate the influence of different fly ash contents on the bond and compressive strength of FA-GPC; they claimed that with the increment in the fly ash content, bond and compression strength of the GPC were improved. The maximum increment in the fc′ of three different fly ash content for five different fly ash binder source materials (ER, MP, BW, GL, and CL) were 19%, 23%, 17%, 36%, and 25%, respectively, when the fly ash content changed from 300 kg/m 3 to 500 kg/m 3 [71]. This result (based on their SEM tests) argued the fact that higher content of fly ash in the geopolymer concrete mixture gives denser and compacted microstructure to the geopolymer concrete matrix. Moreover, the particles of fly ash facilitate movement among the aggregate particles owing to the spherical shape and smooth surface of the particles of fly ash [143]; therefore, reducing the fly ash content decreases the capability of the FA-GPC components to consolidate and compact properly, as a consequence bond and compression strength were decreased in one hand, on the other hand, the volume of fine fraction particles in the geopolymer concrete matrix increased as the fly ash content increased, thus in turn fill the voids and pores between the aggregate particles and hence compressive strength was improved [71]. Similar results of increasing fc′ of FA-GPC with increasing fly ash content were reported at both heat curing and ambient curing regimes [74,96]. For instance, the value of compression strength changed from 21 MPa to 42 MPa, as the fly ash content increased from 300 to 400 kg/m 3 [96].
In addition, according to the findings of Singhal et al. [73], the compressive strength of FA-GPC was improved as the content of fly ash increased. For example, at the molarity of 16 M, the compression strength was increased by 11% and 32%, when the fly ash content was increased from 350 kg/m 3 to 375 and 400 kg/m 3 , correspondingly, at the ambient curing age of 7 days, and increased by 15% and 24% at the age of 28 days. This improvement in the fc′ of the GPC with the increment of fly ash content was reported for other sodium hydroxide molarities, as shown in Fig.11, at the ambient curing age of 28 days. This result may be attributed to the fact that fly ash is the main source of aluminosilicate source materials in the geopolymer concrete mixture, which silica and alumina increased as the amount of fly ash content increased; thus, they affect the reactions in the polymerization process, which in turn, C-A-S-H and N-A-S-H gels increased, and finally, compression strength was improved [73]. In the same context, a research study have been conducted to investigate the properties of FA-GPC, they used different contents of fly ash and they observed that the compressive strength of the fly ash-based geopolymer concrete was increased as the fly ash content increased. For example, the compressive strength was 25.44, 36, and 48 MPa, at 356, 408, and 444 kg/m 3 of fly ash content, respectively. These findings are attributed as the same as mentioned before [94]. Similarly, Ramujeea and PothaRajub [77] observed that the compressive strength was increased as fly ash content increased as fly ash content increased in the geopolymer concrete mixture. Overall, most of the researches revealed that the fc′ of FA-GPC increased with the increment of fly ash content in the geopolymer concrete mixture, and it is obvious that the fly ash with more fineness and glassy phase is more reactive, which leads to accelerating geopolymerization rate and as a consequence produces a high strength geopolymer concrete [20,144,145].

Aggregate Content
Aggregates in the geopolymer concrete mixtures are the same as a conventional concrete mixture which consists of fine and coarse aggregates. In past studies, river and crushed sand with a maximum particle size of 4.75 mm and a specific gravity of 2.60-2.75 were employed as fine aggregate. Its grade also met the requirements of ASTM C 33 [146]. On the other hand, natural gravel or artificial crushed stone with the nominal aggregate size of 20 mm was used in the previous research as the coarse aggregate to prepare the coarse aggregate to prepare the coarse aggregate to prepare the fly ash-based geopolymer concrete mixtures. Based on the collected datasets from different fly ash-based geopolymer concrete mixture proportions, coarse aggregate content was varying between 394 to 1591 kg/m 3 .
A research study had been carried out to show the effect of total aggregate content on the fc′ of FA-GPC at different molarity and curing temperatures. They used five different volume fractions of total aggregate contents from 74% to 82%. They observed that the compression strength of the FA-GPC was increased with increasing total aggregate content up to 78%, and they revealed that this result might be argued to the inadequate binding of the aggregate and the matrix phase in the fly ash-based geopolymer concrete mixture. In addition, they reported that the highest compressive strength was achieved for the geopolymer concrete mixture with 78% of total aggregate content at 60 ℃ curing condition as shown if the Fig.12, but this percent of aggregate content leads to a reduction of the workability by 37.5% as compared to the GPC mixture with 76% of aggregate content, so they prefer to use 76% of total aggregate content having the molarity of 12 M and cured at 90 ℃ as it yields low reduction in the compressive strength (about 2.6%) without hindering the slump value of the FA-GPC mixture [75]. Effect of total aggregate content on the fc′ of FA-GPC at the age of 3, 7, and 28 days [75] In addition, experimental research work has been conducted on the effect of aggregate characteristics on the mechanical and absorption properties of fly ash-based geopolymer mortars. They used three different types of aggregates: river sand, crushed sand, and combined river and crushed sand. They reported that the geopolymer mortar mixtures' compression strength was between 28.2 to 47.8 MPa at the age of 1 day when the molarity was 12 M, the ratio of sodium silicate to sodium hydroxide was 2.5, and the specimens cured at 90 ℃ for 24 hr. It was revealed that the geopolymer mixture with crushed sand had higher compression strength as compared to the other aggregates. This result was attributed to the fact that the crushed sand has a rough surface texture with an angular shape which gives a greater surface to volume ratio and, as a consequence, provides better bond properties between the aggregates and the source material pastes [147]. Also, they reported that the highest compressive strength was recorded for the crushed sand with a coarser grade (2-4) mm compared to the other grades, as shown in Fig.13. Similar results have been reported by Mane and Jadhav [148], who observed that the crushed sand gives higher compression strength as compared to the river sand. Furthermore, they reported that the utilization of granite as a coarse aggregate provides better fc′ to the FA-GPC in compassion to the coarse basalt aggregates. In the same context, another study has been carried out by Nuaklong et al. [149] on the influence of recycled concrete aggregates and crushed limestone aggregates on the FA-GPC properties. They claimed that there is a chance to use recycled concrete aggregates to produce FA-GPC within the 7-days compression strength of 30-38 MPa; however, this value is slightly smaller than those of FA-GPC with crushed limestone aggregates, which has the compressive strength of 38-41 MPa.
Similar results can also be found in other studies, even though different mixture proportions were used [89]. These results argued to the fact that the granite slurry act as a filling agent which fills the voids and pores of the geopolymer concrete and hence made the geopolymer concrete dense, as a consequence lead to increase the fc′ of the GPC until 40% replacement of sand, beyond that lead to decrease in the compression strength because of the high percentages of fine materials in the GPC mixtures [150].  [150] Furthermore, according to the findings of Embong et al. [86] who investigated the effect of replacement of coarse granite aggregate by limestone with different percentages. Their experimental woks were carried out by substitute the portion of granite coarse aggregate by (0%, 25%, 50%, 75%, and 100%) with limestone in the GPC mixtures. They noticed that the replacement of limestone has a greater effect on the fc′ of the FA-GPC in the ambient curing condition as compared to the oven curing condition. For example, an improvement in the compression strength of 35.3%, 19.5%, and 14.15% was achieved for the replacement level of 25%, 50%, and 75%, respectively, compared to the control GPC mixture without any limestone content. This result attributed to the formation of extra C-A-S-H gels, providing a solid structural framework in the geopolymer concrete [86]. Also, the extra dissolution of Si element in the fly ash to generate C-A-S-H gels tackles the drawbacks of the low reactivity in ambient curing conditions [151]. However, a 10.2% reduction in the fc′ was reported for the replacement level of 100% of limestone due to lower aggregate packing density provided by uniformly graded limestone in the GPC mixtures [86]. On the other hand, the replacement of granite with limestone in the oven curing condition provides an improvement in the fc′ just up to 25% replacement level, and beyond that, reduction in the compressive strength was recorded as shown in Fig.15.  [86] Lastly, research work was performed to show the effect of aggregate content on the fresh and mechanical properties of FA-GPC. They used different aggregate contents and various fine aggregate to total aggregate ratios. They concluded that the fc′ of the GPC mixtures increased with the increment in the total aggregate content up to 70%, and beyond that, it was declined. Also the compression strength was improved by increasing the fine aggregate to total aggregate ratio up to 0.35% and then it was decreased. So, it is evident that for a given sort of coarse and fine aggregate, there is a limit proportion of fine aggregate and total aggregate content that provided the highest compression strength for the FA-GPC [98].

(Na2SiO3/NaOH) Ratio
Generally, sodium hydroxide (NaOH) and sodium silicate (Na2SiO3)  The reduction in the fc′ of the FA-GPC at the curing age of 28 days was 22.5% and 29.5% at 0.4 and 0.5 of SH to SS ratio, correspondingly, as compared to the SH to SS ratio of 0.3. Similar results can also be found in other studies even though different SS/SH was used [94,112,123,152]. Furthermore, an experimental research study has been carried out to investigate the effect of various parameters on the performance of FA-GPC. It was observed that the compressive strength was increased with the increment of SS to SH ratio up to 2.5, and beyond that, decline in the fc′ was reported as shown in Fig.17. [98]. This result argued to the fact that the microstructure of the geopolymer concrete changes due to the quantity of sodium silicates content, while the reduction in the compressive strength was attributed to the fact that there is not a sufficient amount of sodium hydroxide present in the mixture to completion of dissolution process during the formation of geopolymer [153,154], or due to the excess OHconcentration in the GPC mixture [25]. On the other hand, some researchers believed that the excess of sodium content could form sodium carbonate by atmospheric carbonation, and this may disrupt the polymerization process, and as a result, compressive strength was decreased [140]. In the same context, another study has been carried out to investigate the effect of SS/SH ratio on the fc′ of the FA-GPC. They used six different SS/SH ratios (0.5, 1.0, 1.5, 2.0, 2.5, and 3.0), and they reported that with the increment of SS/SH compressive strength was increased up to the ratio of SS/SH of 2.5 and then it was decreased [155]. On the other hand, according to the findings of some researchers, the influence of SS/SH on the compression strength of geopolymer concrete was not obvious [99,118].
In contrast to the above-mentioned results, few studies reported that the compression strength decreased as the ratio of SS/SH increased. For instance, a research study was conducted to show the effect of alkaline activators on the mechanical characteristics of FA-GPC at the ambient curing regime. Their results show that with the increase in the SS/SH ratio from 1.5 to 2, compressive strength was decreased by 5.2%, 7.6%, 7.6%, 10.8%, and 12.8% at the molarities of 8, 10, 12, 14, and 16, correspondingly.
Similar trends were true when the ratio of SS/SH changes from 2 to 2.5, as depicted in Fig.18 [117].
This result argued to the reason that the increment in the SS/SH ratio would lead to a decrease the amount of sodium hydroxide solution and hydroxide ions (OH -), which subsequently decreased the formation of N-A-S-H gels, which it is the main 3D network that directly affect the microstructure of the geopolymer concrete and as a consequence fc′ was decreased [156,157]. Similarly, a reduction in the compression strength of FA-GPC was reported [114] as SS/SH ratio increased from 2 to 2.5, as well as a reduction in the compression strength of FA-GP mortar was reported as the ratio of SS/SH increased from 1.0 to 1.5, 2.0 and 2.5 [158].

Sodium Hydroxide (NaOH) Concentration (Molarity (M))
The concentration of sodium hydroxide is one of the key parameters that affect the performance of FA-GPC; therefore, a wide range of researches has been conducted to investigate the mentioned phenomena. According to collected datasets from the literature, the sodium hydroxide concentration  claimed that the fc′ was improved as the molarity of sodium hydroxide increased [88,114,122,123].
According to the findings of Aliabdo et al. [79], the compression strength was increased as the molarity increased in the FA-GPC mixture up to 16 M, and then it was decreased as depicted in Fig.19. Also, they reported that the optimum concentration of NaOH was 16 M for 48 hr of curing at 50℃. Similar results was observed by Chindaprasirt and Chalee [82], who used different molarities of sodium hydroxide (8, 10, 12, 14, 16, 18, and 20), and they reported that the maximum compression strength was 32.2 MPa at 16 M in the age of 28 days at the ambient curing conditions. In the same context, a research study has been carried out to demonstrate the influence of different NaOH molarities (10, 13, and 16 M) on the fc′ of FA-GPC at elevated temperatures (200, 400, 600, and 800℃). Their results indicated that the fc′ was higher for those specimens with higher molarities (13 and 16 M) at all temperature changes than those specimens containing 10 M NaOH solutions as illustrated in Fig.20; on the other hand, interestingly, they observed that the rate of compressive strength loss after 600℃ is also high in the FA-GPC mixtures that have a greater concentration of NaOH solution [83].
In addition, a research study has been conducted to show the effect of different molarity of sodium  Si particles, and as a result, the greater is the compression strength of GPC mixtures [161].
Generally, there are three different curing regimes to cure FA-GPC composites, namely ambient curing, oven curing, and steam curing regimes. In the following paragraph, the effect of these curing conditions on the fc′ of FA-GPC was provided. The majority of researches used oven curing conditions to cure the FA-GPC as compared to other curing condition types.
A research study has been carried out to show the effect of different oven curing temperatures on the compressive strength of FA-GPC. Their results revealed that the fc′ was increased as the oven curing temperatures increased, as illustrated in Fig.22. But this improvement in the compression strength did not increase substantially beyond the curing temperature of 60℃ [139]. In addition, a research study has been conducted to show the effect of different curing conditions on the mechanical properties of FA-GPC. Their results reported that the fc′ of heat curing regime was higher than those specimens cured at ambient curing conditions as depicted in Fig.23   Furthermore, experimental laboratory research works have been conducted to investigate the effect of ambient and oven curing conditions on the mechanical properties of FA-GPC composites. It was observed that the fc′ improvement for oven curing conditions was greater than those specimens cured under ambient curing regimes, as shown in Fig.24. This result attributed to the fact that the process of geopolymerization is accelerated as the temperature of the GPC mixtures increased [105]. In the same manner, Joseph and Mathew [98] revealed that the compression strength of FA-GPC was increased with increasing oven curing temperatures up to 100℃, and beyond that, it was decreased as shown in Chindaprasirt et al. [163] reported that the compression strength was improved as the oven curing temperature increased.  correspondingly [121]. On the other hand, another research study has been conducted to show the effect of ambient, hot gunny sack, and external exposure curing regimes on the performance of FA-GPC. They observed that the compression strength improvement was greater for external exposure curing condition and then for the ambient condition, while the worth curing condition was recorded for the hot gunny suck curing regimes [124]. The duration of curing FA-GPC specimens inside ovens is one issue that some researchers deal with.
For instance, a research study has been carried out to investigate the influence of different mix compositions on the compression strength and microstructures of FA-GPC composites. It was observed that the fc′ improved with increment in curing duration inside an oven at a constant temperature. The highest compressive strength (40.8 MPa) for FA-GPC was obtained with 48 hrs of thermal curing. Also, they reported that further increment in the period of curing did not give an appreciable improvement in the compression strength of the FA-GPC, as illustrated in Fig.27 [125]. Similar results have been reported by Hardjito et al. [139]. On the other hand, Joseph and Mathew [98] claimed that the compression strength of FA-GPC was increased as the curing time inside the oven increased at a constant temperature. This strength gain is proportional to the duration of curing and a very small strength gain could be obtained after 24 hrs, as shown in Fig.28; this result may be attributed to the fact that most of the geopolymerization process would have been completed within 24 hrs.  6 Statistical assessment Below sufficient information regarding each variable considered as the input parameter is present. More information on each statistical criterion was reported by Silva et al. [164].

Alkaline solution/binder (l/b)
According to the dataset, which contains 510 data samples from past researches, the l/b ratio of the FA-GPC was varied from 0.25 to 0.92 with an average, variance, standard deviation, skewness, and kurtosis of 0.5, 0.01, 0.1, 1.21, and 2.88, respectively. Skewness is belonged to distortion or asymmetry in a symmetrical normal distribution in a dataset. If the curve is moved to the right or the left side, it is stated to be skewed. Also, skewness could be quantified as an impersonation of the range to which a given distribution differs from a normal distribution. For instance, the skew of zero value was measured for normal distribution, while, right skew is an indication of lognormal distribution [165]. Figure 29 is presented the relationship between the compression strength of geopolymer and l/b.

Fine aggregate content (F)
In past studies, river and crushed sand with a maximum particle size of 4.75 mm and a specific gravity

Sodium silicate (SS)
Based on the dataset, which contains 510 data samples from literature, the content of SS was varied between 48 to 342 kg/m 3 . The constituents of the SS were SiO2, Na2O, and water. The range of SiO2 was varying from 28 to 37%, Na2O was in the range of 8 to 18%, and the percent of water in the SS was in the range of 45 to 64%. The statistical analysis for the collected data of SS revealed that the average content of SS in the FA-GPC was 123.4 kg/m 3 , the standard deviation was 36.2 kg/m 3 , the variance was 1313, skewness was 2.89, and kurtosis was 12.8. (Fig.35).

SS/SH
Referring to the collected data, the ratio of Na2SiO3 to NaOH was varied from 0.

Compression strength (fc′)
The measured fc′ of the 510 collected data from the literature studies was shown in Table 3

Modeling
Based on the coefficient of determination (R 2 ) and statistical analysis, there are no direct relationships between the fc′ and the constituents of the FA-GPC, as shown in Fig.29 to Fig.40.
As a result, four different models are proposed to examine the effects of the above mixture proportions on the fc′ of FA-GPC, as shown below.
The models suggested in this paper are used to forecast the fc′ of the FA-GPC and choose the optimum solution that delivers a better estimate of fc′ than the experimentally determined fc′. All the collected datasets were randomly split into three parts: training, testing,, and validating datasets [57,69]. 340 training dataset is used to train the LR, MLR, ANN, and M5P-tree model and obtain the optimal weights and biases, while 85 testing dataset is used to confirm the fulfillment of the proposed models. Moreover, 85 validating datasets are used to explore the generality of the models and prohibition of the over-fitting problem in the case of classical training algorithms. Comparisons of model predictions were made using the following evaluation criteria: The model should be scientifically accurate, with a smaller proportion of error between observed and projected data, as well as lower RMSE, OBJ, SI, and higher R 2 values.

Linear regression model (LR)
One of the most common methods to predict the fc′ of concrete is the linear regression model (LR) [166], as shown in Eq.1, and it is considered as a general form of the linear regression model [66,69,165].

Multilogistic model (MLR)
Same as the former models, a multi-logistic regression analysis model was carried out for the collected datasets, and the general form of the MLR is shown in Eq.3 based on the research studied that had been conducted by Mohammed et al. [66] and Faraj et al. [69]. MLR is used to clarify the difference between a nominal predictor variable and one or more independent variables.  While a, b, c, d, e, f, g, h, i, j, k, l, and m are the model parameters.

Artificial Intelligence network (ANN)
ANN is a powerful simulation software designed for data analysis and computation to think as a human brain in terms of processing and analyses. This machine learning tool is widely used in construction engineering for predicting the future behavior of several numerical problems [60,167,168]. ANN model is generally divided into three main layers, which are input, hidden, output layers. Each of the input and output layers 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 are usually depending 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 pro-portions, weight/bias, several parameters including (l/b, FA, Si/Al, …) as inputs, and the output of ANN here is the compressive strength. 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 iteration (epochs) that provide the minimum mean absolute error (MAE), and root mean square error (RMSE) and best R-value that close to one. The effect of several epochs on reducing the MAE and RMSE has been studied. The collected data set (a total of 510 data) has been divided into three parts for the training purpose of the designed ANN. Around 70 th percent of the collected data was used as a trained data for training the network, 15 th percent of overall data was used for testing the dada set, and the rest of the remaining data was used to validate the trained network [167]. The designed ANN was trained and tested for various hidden layers to determine optimal network structure based on the fitness of the predicted compression strength of FA-GPC with the fc′ of the real collected data. It was observed that the ANN structure with two hidden layers, twenty neurons, and a hyperbolic tangent transfer function was a best-trained network that provides a maximum R 2 and minimum both MAE and RMSE (shown in Table 4). As a part of this study, an ANN model has been From linear node 0: From sigmoid node 1: From sigmoid node 2:

M5P-tree model
The M5P model tree is a reconstruction of Quinlan's M5 algorithm [169] that is based on the conventional decision tree with the addition of a linear regression function to the leaves nodes. The decision tree is a representation of the algorithms by a tree form trained through a data to form nodes.
The nodes constituting the decision tree are divided into three types namely; root nodes, internal nodes, and leaves nodes. Nodes are interconnected to each other through branches until the leaves reached [170,171]. entering that node. The attribute that maximize the reduction of estimated error at each node is used to evaluate any task of that node. As a result of this division in the M5P tree, a large tree likes structure that leads to overfitting will be created. In the followed step, the enormous tree is trimmed, and the With the exception of the R 2 value, zero is the optimal value for all other evaluation parameters.
However, one is the highest benefit for R 2 . When it came to the SI parameter, a model has bad performance when it is > 0.3, acceptable performance when it is 0.2 SI 0.3, great performance when it is 0.1 SI 0.2, and great performance when it is 0.1 SI 0.1 [57,69,172]. Furthermore, the OBJ parameter was employed as a performance measurement parameter in Eq.12 to measure the efficiency of the suggested models.

MLR model
The developed models for the MLR model with various variable parameters presented in Eq.14. In the MLR model, like other developed models, the curing temperature, sodium silicate content, and alkaline liquid to the binder ratio were the most significant independent variables that affect on the fc′ of the FA-GPC that is matched with the experimental works presented in the literature [71,73,79,96,98,105,108,112,113,121,122,139]. The relationships between the predicted and measured fc′ of the training data set for the FA-GPC was shown in Fig.45a. Further, same as the previous model, this model was checked by two sets of data (testing and validating dataset) to show their efficiency for other data out of the model data (training data); the results show that this model can be used to predict the fc′ of FA-GPC just by substitute the independent variables into the developed equation as shown in Fig.45b and Fig.45c.    Figures 47a, b, and c. The studied datasets have a +15% and -20% error line for the training data, +10% and -20% error lines for testing data, and +15% and -10% for the validating datasets, which it is better than the other developed models. Furthermore, this model has a better performance compared to other models to predict the fc′ of the FA-GPC based on the value of OBJ and SI that illustrated in Fig. 43 and Fig. 44, also, the value of R 2 = 0.9647, MAE = 2.5853 MPa, and RMSE = 2.332 MPa. Finally, the residual compression strength for the ANN model was shown in Fig.49 for the predicted and measured fc′ by using all datasets.   [71,73,79,98,139]. Fig.51 indicates the tree-shaped branch relationship, and the model (Eq.15) parameters are summarized in Table 5, and based on the linear tree registration function, the model variables will be selected.      increases the compressive strength of fly ash-based geopolymer concrete, although the increase in strength may be insignificant for curing at more than 60°C and for periods longer than 48 hrs.
Therefore, for heat curing regimes, temperatures between 50-80℃ and curing time of 24 hr are widely accepted values used for successful geopolymerization process. In addition, among the curing condition methods (oven, steam, and ambient), oven curing techniques has a better influence on the compressive strength of fly ash-based geopolymer concrete composites.
 As a result of the comprehensive systematic review that had been carried out in this study on the factors that affect on the compressive strength of FA-GPC, different mixture proportions, curing conditions, and age of the concrete specimens influence the compressive strength. Therefore, developing multi-scale models to predict this important property of the geopolymer concrete by considering a wide range of input parameters is essential regarding knowing the effect of each parameter on the compressive strength of the FA-GPC as well as modeling will be helpful for the concrete and construction industry regarding saving in time, energy, cost-effectiveness, and it gives guidance about scheduling for the construction process and removal of formwork elements.

Conclusions
Based on the extensive literature review and discussions made in this study, the following conclusions can be reached:

I.
Geopolymer concrete with acceptable compressive strength values could be produced by using fly ash as source binder materials.

II.
, The alkaline solution to the binder ratio (l/b), has a significant impact on the compressive strength of the fly ash-based geopolymer concrete. Some researches believed that the compressive strength was improved as the l/b increased. While, many researchers reported a reduction in the compressive strength as the l/b was increased.

III.
Increasing water content or extra water to the fly ash-based geopolymer concrete will lead to decreasing the compressive strength of the geopolymer concrete. While superplasticizer content improves the compressive strength of the fly ash-based geopolymer concrete composites up to a limited value around 2% of fly ash content.

IV.
strength of fly ash-based geopolymer concrete increases as the ratio of SS/SH increased up to around 2.5, and then it decreased.

V.
It was suggested to use the molarity of sodium hydroxide in the range of 10-16M to produce the fly ash-based geopolymer concrete mixtures with acceptable compressive strength behavior.

VI.
Among the curing methods, the heat curing regime is the best one for getting early and high compressive strength of fly ash-based geopolymer concrete.

VII.
It was suggested to use the oven curing temperatures between 50-80℃ and curing time of 24 hr for successful geopolymerization process as well as getting acceptable compressive strength of fly ashbased geopolymer concrete.

VIII.
Different models could be successfully used to predict the compressive strength of FA-GPC with different mixture proportions, curing regimes, and concrete ages.

IX.
All the used models LR, MLR, ANN, and M5P could be successfully used to develop predictive models for the compressive strength of the FA-GPC. Overall, the ANN model has better performance than the other two models.

Recommendation
Detailed investigations on the fresh and mechanical properties of fly ash-based geopolymer concrete can be found in the literature. However, studies which are focused on the other properties of this composite are still limited. In order for this composite to be acceptable by the construction industry, some durability properties such as water permeability, gas permeability, chloride resistance, fatigue performance, and freeze-thaw resistance should further be examined comprehensively.

Author Declarations
We wish to draw the attention of the Editor to the following facts which may be considered as potential conflicts of interest and to significant financial contributions to this work.
[OR] We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing, we confirm that we have followed the regulations of our institutions concerning intellectual property. We

Availability of data and materials
The data supporting the conclusions of this article are included with the article.