Title: Concrete compressive strength prediction modelling utilising ensemble-deep-learning framework

Authors: Mu'tasime Abdel-Jaber; Ma'en Abdel-Jaber; Rob Beale; Nisrine Makhoul

Addresses: Hourani Center for Applied Scientific Research, Al-Ahliyya Amman University, Al-Saro Street, Al-Salt, Amman, Jordan ' Al Hussein Technical University, Bulding 23, King Hussein Business Park, King Abdullah II St 242, Amman, Jordan ' Faculty of Design, Technology and Environment Oxford Brookes University, Headington Campus, Oxford; OX3 0BP, UK ' École Spéciale des Travaux Publics, du Bâtiment et de l'Industrie, ESTP Paris, 28 Av. du President Wilson, 94230 Cachan, France

Abstract: The unique ensembled-deep-learning model used in this work is used to estimate the compressive strength of concrete automatically. These raw data will be pre-processed using a data-cleaning technique. Then, characteristics based on Yule's coefficient, Pearson's coefficient, and Percentile coefficient will be retrieved from the pre-processed data, along with statistical features. A new ensembled-deep-learning model will be developed using the extracted features to predict the concrete strength. A new hybrid optimisation approach called the hybrid poor rich owl algorithm (HPROA) is applied to adjust CNN's weight to increase the projected model's capacity for accurate prediction. A conceptual HPROA is the fusion of the common poor and rich optimisation (PRO) and owl optimisation will result in the hybrid optimisation model that is provided. The implementation is performed using the MATLAB software. Compared to already used methods, the suggested model's performance is evaluated and the obtained mean square error (MSE) is zero.

Keywords: compressive strength; concrete; ensembled-deep-learning model; data cleaning technique; multilayer perceptron; MLP; convolutional neural network; CNN; gated recurrent units; GRUs.

DOI: 10.1504/IJMOR.2026.153584

International Journal of Mathematics in Operational Research, 2026 Vol.34 No.1, pp.73 - 99

Received: 12 Sep 2023
Accepted: 22 Dec 2023
Published online: 18 May 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article