Alibrahim Bashar, Habib Ahed, Habib Maan
Department of Civil Engineering, Eastern Mediterranean University, via Mersin 10, Famagusta, North Cyprus, Türkiye.
Research Institute of Sciences and Engineering, University of Sharjah, Sharjah, United Arab Emirates.
Sci Rep. 2025 Jan 15;15(1):1989. doi: 10.1038/s41598-024-84325-z.
Concrete compressive strength is a critical parameter in construction and structural engineering. Destructive experimental methods that offer a reliable approach to obtaining this property involve time-consuming procedures. Recent advancements in artificial neural networks (ANNs) have shown promise in simplifying this task by estimating it with high accuracy. Nevertheless, conventional ANNs often require deep networks to achieve acceptable results in cases with large datasets and where generalization is required for a variety of mixtures. This leads to increased training durations and susceptibility to noise, causing reduced accuracy and potential information loss in deep networks. In order to address these limitations, this study introduces a novel multi-lobar artificial neural network (MLANN) architecture inspired by the brain's lobar processing of sensory information, aiming to improve the accuracy and efficiency of estimating concrete compressive strength. The MLANN framework employs various architectures of hidden layers, referred to as "lobes," each with a unique arrangement of neurons to optimize data processing, reduce training noise, and expedite training time. Within the study context, an MLANN is developed, and its performance is evaluated to predict the compressive strength of concrete. Moreover, it is compared against two traditional cases, ANN and ensemble learning neural networks (ELNN). The study results indicated that the MLANN architecture significantly improves the estimation performance, reducing the root mean square error by up to 32.9% and mean absolute error by up to 25.9% while also enhancing the A20 index by up to 17.9%, ensuring a more robust and generalizable model. This advancement in model refinement can ultimately enhance the design and analysis processes in civil engineering, leading to more reliable and cost-effective construction practices.
混凝土抗压强度是建筑和结构工程中的一个关键参数。能够提供可靠方法来获取该性能的破坏性试验方法涉及耗时的程序。人工神经网络（ANN）的最新进展显示出通过高精度估计来简化这项任务的前景。然而，传统的人工神经网络在处理大数据集以及需要对各种混合料进行泛化的情况下，通常需要深度网络才能取得可接受的结果。这导致训练时间增加且易受噪声影响，从而降低了深度网络的准确性并可能导致信息丢失。为了解决这些局限性，本研究引入了一种受大脑叶状处理感官信息启发的新型多叶人工神经网络（MLANN）架构，旨在提高估计混凝土抗压强度的准确性和效率。MLANN框架采用了各种隐藏层架构，称为“叶”，每个叶都有独特的神经元排列，以优化数据处理、减少训练噪声并加快训练时间。在本研究范围内，开发了一个MLANN，并对其预测混凝土抗压强度的性能进行了评估。此外，还将其与两种传统情况进行了比较，即人工神经网络（ANN）和集成学习神经网络（ELNN）。研究结果表明，MLANN架构显著提高了估计性能，均方根误差降低了高达32.9%，平均绝对误差降低了高达25.9%，同时A20指数提高了高达17.9%，确保了更强大且更具泛化性的模型。模型优化方面的这一进展最终可以加强土木工程中的设计和分析过程，带来更可靠且更具成本效益的施工实践。
