**1. Introduction**

Numerous tests are performed to measure concrete performance, but compressive strength is frequently considered the most significant [1]. Compressive strength tests

**Citation:** Yuan, X.; Tian, Y.; Ahmad, W.; Ahmad, A.; Usanova, K.I.; Mohamed, A.M.; Khallaf, R. Machine Learning Prediction Models to Evaluate the Strength of Recycled Aggregate Concrete. *Materials* **2022**, *15*, 2823. https://doi.org/10.3390/ ma15082823

Academic Editor: Krzysztof Schabowicz

Received: 4 March 2022 Accepted: 27 March 2022 Published: 12 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

offer good insight into the concrete's diverse properties. The compressive strength of concrete is directly or indirectly connected to a number of mechanical and durability properties [2]. Flexural strength is also a key characteristic to consider when designing structural concrete, since it has an effect on the flexural cracking, shear strength, deflection properties, and brittleness ratio of the concrete [3]. The compressive and flexural strength of recycled aggregate concrete (RAC) are reliant on a number of variables, including the mechanical and physical properties of the recycled aggregate used, as well as the microstructure of the resulting matrix [4]. Typically, RAC has an inferior compressive and flexural strength compared to natural aggregate concrete, owing to insufficient bonding between the aggregate and the old mortar, fractures and cracks in the recycled aggregate formed during the recycling procedure, and the presence of low-permeability mortar connected to the recycled aggregate [5–7]. Furthermore, the characteristics of RAC are reliant on the amount of recycled aggregate substituted and the moisture content [8,9]. The strength of RAC varies according to the recycled aggregate replacement ratio, the water– cement ratio (w/c), the recycled aggregate's moisture content, and the recycled aggregate's physical and mechanical properties [9,10]. When w/c is held constant, experimental data suggest that recycled aggregate replacement content has a significant effect on the strength of RAC [11,12]. When natural aggregate is totally replaced with recycled aggregate, the compressive strength of RAC can be reduced by up to 30% [13,14]. Similarly, other researchers discovered a drop in compressive strength of between 12% and 25% with 100% recycled aggregate incorporation [15,16]. It was discovered that the age of the waste concrete from which the recycled aggregate was manufactured had a substantial impact on the strength of the RAC [17]. Moreover, the strength of the parent concrete from which recycled aggregates are produced affects the strength of the RAC [18]. Hence, there are several factors that influence the strength of RAC, and to study their combined impact through experimental investigations is challenging. In contrast, using computational methods might better examine the combined influence of these factors on the strength of RAC.

The practice of developing models for forecasting the strength of concrete is ongoing in order to reduce unnecessary test repetitions and material waste. There are several prominent models for modeling concrete properties, such as best fit curves (based on regression analysis). However, due to the nonlinear behavior of concrete [19,20], regression models generated using this technique may not accurately represent the underlying nature of the material. Additionally, regression methods may understate the effect of constituent materials in concrete [21]. Artificial intelligence techniques such as machine learning are some of the more contemporary modeling techniques that have been used in the area of civil engineering. These approaches use input parameters to model responses, and the output models are validated by experimentation. For construction applications, machine learning algorithms estimate concrete strength [22–26], bituminous mixture performance [27], and concrete durability [28–30].

This study concentrates on the use of machine learning methods to forecast the compressive and flexural strength of RAC. Two distinct ensemble machine learning techniques were used—gradient boosting and random forest—and the effectiveness of both methods was evaluated using correlation coefficients (R2) and statistical checks. Moreover, k-fold analysis and error distributions were used to determine the validity of each technique. The reason for selecting the ensemble machine learning method was that the literature reported their better performance compared to individual machine learning methods, such as support-vector machines and artificial neural networks [31–33]. This research is interesting in that it predicts both the compressive and flexural strength via two ensemble machine learning methods, while experimental studies require considerable human effort, the cost of experimentation, and time for material collection, casting, curing, and testing. Since a number of factors—including w/c, recycled aggregate replacement ratio, parent concrete strength, water absorption of the recycled aggregate, density of the recycled aggregate, etc.—influence the strength of RAC, their combined impact is hard to study through

an experimental approach. Machine learning methods are capable of determining their combined impact at a reduced effort. Machine learning methods require a dataset, which may be collected from past studies, since many investigations have been undertaken to determine material strength, and such a dataset might be utilized for training the machine learning models and forecasting the material properties. The purpose of this work is to ascertain the most appropriate machine learning method for the compressive and flexural strength estimation of RAC based on the results forecast and the effects of various parameters on the strength of RAC.
