Predicting the Modulus of Elasticity of Recycled Aggregate Concrete Using an Artificial Neural Network Model in The Gambia

The Elastic Modulus (E) is an important mechanical property of materials universally defined as the ratio of stress to strain below the proportional limit. Greater values of elasticity, E, imply smaller strains resulting from larger stresses and, in concrete, indicate greater stiffness or more strength [1]. The elastic modulus of concrete, Ec, is not easily determined as concrete is not an elastic material and does not follow Hooke's law. Ec must be determined in designing structures as it shows the ability of concrete to withstand stresses and not deform elastically, that is, undergo deflections, or produce strains [2, 3]. Ec is also needed to satisfy requirements of ultimate and serviceability limit states [1]. Determining the elastic modulus of concrete involves using a Universal Testing Machine (UTM) or a Compression Testing Machine, and using the compressive strength, Cs, and/or stress-strain graph obtained, to calculate Ec. (Compressive strength is a measure of the compressive forces or loads a material can bear before cracking or fracturing.) This is a tedious, time-consuming process and is not always accurate as there are many formulas available to use [3].

As buildings age, they get to a point where they are deemed unsafe and demolished. Structures may also be demolished for other reasons and this demolition produces construction waste. To catapult into a more environmental and economic industry, the Gambian construction industry can look into how recycled aggregates (RAs) from construction waste can be reused in new upcoming projects/buildings, especially Recycled Concrete Aggregates (RCA). This RCA is derived by separating concrete rubble from the mortar content, rubble, bricks, tiles, sand, dust, timber, plastics, metals, cardboard, and paper, of construction and demolition waste, and crushing it to required sizes [4]. Determining the Ec of concrete made with RCAs proves to be a difficult task as RCAs are sourced from different places and the properties of the original concrete are not always known. As the world becomes more aware and concerned about our environment, reusing RAs helps to reduce the number of debris and construction waste that takes up land space and leaches into the ground, making it agriculturally unusable even in the event of a transfer of materials in the future [5]. Environmental risks of RA causing pollution include Chromium (Cr) leaching.

In the past decade, the advancement of Artificial Intelligence (AI) has been useful in application to various human problems, including civil engineering ones. Data-Driven Models (DDMs) are heuristic techniques that have been employed in the concrete industry and successfully predicted the mechanical properties of several types of concrete. Artificial Neural Networks (ANNs) have been particularly outstanding in predicting these properties as compared to other DDMs such as Multivariate Adaptive Regression Splines (MARS), Grasshopper Optimisation Algorithm (GOA), Extreme Learning Machine (ELM), Random Forest (RF), M5 Prime Model Tree (M5p), etc. [6]. ANNs are also much simpler to use. ANNs may be defined as computational models that can perform pattern recognition and retention of information-based knowledge by imitating brain activity. The brain is a complex organ; ANNs run a simplified version of its operations. Thus, ANNs are made up of an input layer, one or more hidden layers, and an output layer. Each layer can have a different number of nodes (i.e., artificial neurons) [7].

It is important to understand the mechanical properties of recycled aggregate concrete (RAC) in reusing it in the construction industry to ensure the quality of buildings produced is not affected by the replacement. The elastic modulus of concrete is one of the most important mechanical properties. Thus, this study seeks to accurately predict the elastic modulus of Recycled Aggregate Concrete in The Gambia using an ANN model, and this is achieved by predicting the compressive strength of Recycled Aggregate Concrete (RAC) in The Gambia, comparing the compressive strength of the Natural Aggregate Concrete (NAC) with that of RAC, comparing the elastic modulus of NAC with that of RAC and predicting the applicability of ANNs in solving further problems within the civil engineering field.

Several RAs have been used to replace natural aggregates (NAs) in concrete mixes such as recycled mortar, recycled bricks, etc. This study limits itself to the use of Recycled Concrete Aggregates (RCAs). The RCAs replaced 100% of the NAs in the experimental component of the study as coarse aggregates for the RAC mix. Fine Aggregates were not replaced in the study and all mixes were designed for normal-weight concrete. The ANN for the study was constructed only for the RAC using TensorFlow [8], Keras API [9], and run on Google Colab [10]. About 90 papers from previous studies were sourced manually in compiling the dataset for the model. The output parameter considered was the elastic modulus of RAC and the study's dataset had a total of 11 input parameters which did not include factors such as the type/nature of curing, the type of natural aggregate, and the moisture state of the RCA.