Assessment of ML techniques and suitability to predict the compressive strength of high-performance concrete (HPC)

Abstract

Using industrial soil waste or secondary materials for making cement and concrete has encouraged the construction industry because it uses fewer natural resources. High-performance concrete (HPC) is recognized for its exceptional strength and sturdiness compared to conventional concrete. Accurate prediction of the compressive concentration of HPC is vital for optimizing the concrete mix design and ensuring structural integrity. Machine learning (ML) techniques have shown promise in predicting concrete properties, including compressive strength. This research focuses on various ML techniques for their suitability in predicting the compressive dilution of HPC. In this research, the Extended Deep Neural Network (EDNN) technique is used to analyze the strengths, limitations, and performance of different ML algorithms and identify the most effective methods for this specific prediction task. However, there is a problem with accuracy. Therefore, our research approach is the EDNN-centred strength characteristics prediction of HPC. In the suggested approach, data is initially acquired. Afterward, the data is pre-processed through normalization and removing missing data. Thus, the data are fed into the EDNN algorithm, which forecasts the strength characteristics of the particular mixed input designs. With the Multi-Objective Jellyfish Optimization (MOJO) technique, the value of weight is initialized in the EDNN. The activation function is the Gaussian radial function. In the experimental analysis, the implementation of the suggested EDNN is evaluated to the performance of the prevailing algorithms. When compared to current research methodologies, the proposed method performs better in this regard.