Full length article
Fuzzy-metaheuristic ensembles for predicting the compressive strength of brick aggregate concrete

https://doi.org/10.1016/j.resconrec.2021.105443Get rights and content

Abstract

Several million tons of demolition and construction (D&C) wastes are being produced worldwide. Brick waste is one of the eminent D&C wastes and several models have been performed on recycling brick waste to produce environmentally friendly concrete. This study develops three fuzzy-metaheuristic ensembles, based on adaptive neurofuzzy inference system (ANFIS) with a fuzzy c-mean clustering approach to forecasting the compressive strength of brick aggregate concrete (BAC). Such models incorporate ANFIS with particle swarm optimization (PSO), genetic algorithm (GA), and firefly algorithm (FFA). For the model development, 132 datasets were used for standalone and hybrid models. All models were trained and tested with 80% and 20% of the datasets, respectively. A k-fold cross-validation method has been applied to validate the generalization accuracy of the established models. A sensitivity analysis is also used for evaluating the influence of each input variable on the proposed techniques. Among the different designed and trained models, the results revealed that the hybrid ensemble models are more successful than ANFIS based fuzzy c-mean clustering approach in forecasting the 28-days compressive strength of BAC. The developed ANFIS-PSO model yields better prediction compared with the other two hybrid GA and FFA models so the correlation coefficient (R2), a root means square error (RMSE), and the mean absolute error (MAE) were acquired as 0.913, 0.057, and 0.032, respectively. A sensitivity analysis indicated that the content of cement and the specific gravity of brick aggregates played a significant role in the model output.

Introduction

Rapid urbanization certainly results in an enormous construction of new buildings together with the demolition of the old building. In China, approximately 1.8 billion tons of demolition and construction (D&C) wastes were produced in 2017, and as such, it is considered the largest producer of D&C wastes in the world (Ma et al., 2019). A huge deal of brick waste has been produced, approximately 0.4 billion tons of the total amount of D&C wastes are produced annually during the different D&C activities (Fig. 1). The D&C wastes disposal or landfills inevitably generates some environmental impacts (Shaban et al., 2019a). Thus, researching an economic and environmental technique to handle the disposal of such wastes has become a significant feature for developing a sustainable environment (Liang et al., 2019, Shaban et al., 2021; Shaban et al., 2019b). It has been proposed that the brick waste could be recycled to substitute the natural aggregate (NA) in concrete (Li et al., 2016, Yang et al., 2020). The utilization of recycled brick not only helps in decreasing the landfill problem, but it is also economically useful furthermore being eco-friendly. Previous studies (Bazaz and Khayati, 2012; Dang et al., 2018; Jankovic et al., 2010; Wong et al., 2018; Yang et al., 2011) indicate that the recycled brick aggregate can be utilized to replace the coarse and fine NA in mortar and concrete. However, the higher water absorption and porous structure of brick aggregate (BA) cause inferior workability for fresh concrete and badly affect the several properties of the resultant concrete. The lack of knowledge about the behavior of brick aggregate concrete (BAC) and the limited available standards are also considered other issues that hinder the use of BA in concrete. Although, Cachim (2009) observed that the BA replacement level of up to 20% had no noticeable bad impact on the splitting tensile and compressive strengths of concrete. Yang et al. (2011) indicated that the flexural strength was not affected by a 50% substitution percentage of coarse BA. Adamson et al. (2015) revealed that the chloride penetrability resistance of BAC reduced with increasing the content of BA. Results agreed with Yang et al. (2011) that detected the clear durability reduction of BAC by increasing the content of BA in concrete. This inconsistent behavior of BAC is due to the different physical properties of BA and the proportion of concrete mix (Yang et al., 2020). It is necessary to realize the relation between the proportions of concrete mix and the mechanical properties of BAC. Accordingly, the application of artificial intelligent (AI) techniques has been rising steadily for the development of predicting the concrete behavior to encourage the large scale using of recycled brick aggregate in concrete and thus adopt its use in the construction industry.

Recently, AI techniques such as regression, neural network (ANN), random forecast (RF), and adaptive neurofuzzy inference system (ANFIS) have been approved due to their high precision in predicting the various properties of different concrete kinds through finding a correlation between parameters (Khademi et al., 2016; Zhang et al., 2019; Gao et al. 2020; Cai et al., 2020; Lyu et al., 2020, Shariati et al., 2020a). Among the existing AI models, ANFIS gives a more reasonable estimation for the compressive strength of concrete (Khademi et al., 2015). Tesfamariam and Najjaran (2007) proposed a neuro-fuzzy model to forecast the mixture design of normal concrete. Bilgehan and Turgut (2010) conducted a comparative study for determining the concrete compressive strength using ANN and ANFIS models. Nehdi and Bassuoni (2009) used a fuzzy logic method to predict the durability of concrete. Sadrmomtazi et al. (2013) determined the compressive strength of lightweight concrete through ANFIS. Khademi et al. (2016) summarized that the ANFIS model can be utilized in the optimization of the mix design of recycled aggregate concrete and the case of higher precision requirements. Despite the significant performance presented by ANN and ANFIS models, the influence of the clustering approach on the metaheuristic models is not clearly investigated. Hence, there is a necessity for predominant optimization algorithms to avoid the limitation of the standalone models such as local minima and poor generalization.

The metaheuristic optimizations, e.g., particle swarm optimization (PSO), firefly algorithm (FFA), and genetic algorithm (GA) are considered as a powerful population-based approach which capable to resolve the disconnected and improve the generalization performance of artificial techniques (Elbaz et al., 2019, Shariati et al., 2020b; , Elbaz et al., 2019, Elbaz et al., 2020 a, Elbaz et al., 2020 b; Nguyen et al., 2019). Furthermore, the seeking for an optimization method is vital to fulfilling the best design with minimum objective function by varying the design parameters whereas satisfying design constraints. To avoid uncertainties and imprecisions that exist in traditional models, different optimization models are more convenient. To date, there are no existing models considering the fuzzy clustering approach to forecast the compressive strength of BAC based on ANFIS-GA, ANFIS-FFA, and ANFIS-PSO models.

This study aims to propose three fuzzy-metaheuristic ensembles, based on GA, PSO, and FFA to forecast the compressive strength of BAC through the amount of concrete component and the properties of BA. For this purpose, comprehensive data were collected during laboratory experiments of the previous studies to use in compressive strength predictive model development. For evaluating the performance of the developed algorithms and dataset accuracy, a k-fold cross-validation approach was utilized. Based on the statistical error analyses, the performance of the developed models was validated and compared with those from the ANFIS model. Finally, to assess the significance and weight of each parameter to forecast the compressive strength, a sensitivity analysis was performed.

The remainder of the present study is structured as follows: the used mathematical tools are discussed in Section 2. Section 3 introduces the development and implementation of the utilized AI models. The data collection and the input parameters used in this study for forecasting the 28-days compressive strength of the BAC are explained in Section 4. Section 5 shows the results of the developed models and the performance of the models, while Section 6 concludes this study.

Section snippets

Methodology

In order to forecast the compressive strength of BAC, the accomplishments performed in the present study are illustrated in Fig. 2. The total process is classified into four main parts; (i) the collection of data samples and thus, determining the input parameters and the corresponding output (ii) the determination of the AI method, (iii) the optimized metaheuristic ensemble models, and (iv) the evaluation process and sensitivity analysis. Three metaheuristic algorithms, namely GA, PSO, and FFA

Developing soft computing techniques

Each input in ANFIS generally includes several membership functions (MFs) and each MF becomes a maximum somewhere. Overall, there are no specific formulas to forecast the type and number of MFS (Elbaz et al. 2019; 2020a). Therefore, the MFs were calculated by the trial and error assumption. In order to increase the accuracy of ANFIS and optimize the position of MFs through the training process, the metaheuristic optimizations namely; GA, PSO, and FFA are used. In the beginning, all the datasets

Results and discussion

In this study, all the datasets were normalized in the range of [0, 1] for simplifying the modelling process based on the following Eq.Xn=(XXmin)(XmaxXmin)where: Xn, Xmin and Xmax are normalized, minimum, and maximum data for each parameter, respectively.

The datasets were also randomly divided into two sets, i.e., 80% datasets for model construction and 20% for testing the developed models (Elbaz et al., 2020 a, Elbaz et al., 2020c). For implementing the established models, MATLAB software

Conclusions

This study presents an attempt to provide a new way for forecasting the compressive strength of brick aggregate concrete. In this regard, three fuzzy-metaheuristic optimization algorithms are developed to predict the compressive strength of BAC through the amount of concrete components and the properties of BA. Five-fold cross-validations were conducted as the validation technology and the optimal performance of the learning algorithms was observed. Major conclusions are observed as follow:

  • 1

    The

CRediT authorship contribution statement

Wafaa Mohamed Shaban: Conceptualization, Methodology, Software, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Project administration. Jian Yang: Conceptualization, Methodology, Validation, Resources, Supervision. Khalid Elbaz: Methodology, Software, Validation, Investigation, Writing - review & editing, Project administration. Jianhe Xie: Funding acquisition. Lijuan Li: Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

The authors would like to thank the supports from Science and Technology Program of Guangzhou, China [Grant No. 201704030057].

References (45)

  • A. Rohani et al.

    A novel soft computing model (Gaussian process regression with K-fold cross validation) for daily and monthly solar radiation forecasting

    Renew. Energy

    (2018)
  • W.M. Shaban et al.

    Effect of pozzolan slurries on recycled aggregate concrete: Mechanical and durability performance

    Constr. Build. Mater.

    (2021)
  • W.M. Shaban et al.

    Properties of recycled concrete aggregates strengthened by different types of pozzolan slurry

    Constr. Build. Mater.

    (2019)
  • M.T. Uddin et al.

    Effects of maximum size of brick aggregate on properties of concrete

    Constr. Build. Mater.

    (2017)
  • B. Vakhshouri et al.

    Prediction of compressive strength of self-compacting concrete by ANFIS models

    Neurocomputing

    (2018)
  • C.L. Wong et al.

    Potential use of brick waste as alternate concrete-making materials : a review

    J. Clean. Prod.

    (2018)
  • J. Yang et al.

    Concrete with recycled concrete aggregate and crushed clay bricks

    Constr. Build. Mater.

    (2011)
  • J. Yang et al.

    Properties of concrete containing strengthened crushed brick aggregate by pozzolan slurry

    Constr. Build. Mater.

    (2020)
  • W.-H. Zhou et al.

    Study of the volumetric water content based on density, suction and initial water content

    Measurement

    (2016)
  • J.B. Bazaz et al.

    Properties and performance of concrete made with recycled low-quality crushed brick

    J. Mater. Civ. Eng.

    (2012)
  • D.T Bui et al.

    Fine-tuning of neural computing using whale optimization algorithm for predicting compressive strength of concrete

    Eng. Comp.

    (2019)
  • H. Cai et al.

    Intelligent building system for 3D construction of complex brick models

    IEEE Access

    (2020)
  • Cited by (0)

    View full text