A Bio-Inspired Adaptive Probability IVYPSO Algorithm with Adaptive Strategy for Backpropagation Neural Network...

Area of Science:

• Civil Engineering
• Materials Science
• Computational Intelligence

Background:

• Accurate prediction of high-performance concrete (HPC) strength is vital for structural integrity and sustainable construction.
• Complex interactions within HPC constituents challenge traditional predictive models, leading to inaccuracies.
• Existing methods struggle with high dimensionality and noisy data, risking overfitting and local optima.

Purpose of the Study:

• To develop a novel bio-inspired hybrid optimization model, AP-IVYPSO-BP, for enhanced HPC strength prediction.
• To address the limitations of conventional models in capturing nonlinear and multi-factorial relationships in HPC.
• To improve the accuracy, robustness, and generalization of concrete strength prediction models.

Main Methods:

• Integration of the ivy algorithm (IVYA) with particle swarm optimization (PSO) and an adaptive probability strategy.
• Optimization of backpropagation neural network (BPNN) weights and biases using the AP-IVYPSO algorithm.
• Training and validation on a dataset of 1030 HPC mix samples.

Main Results:

• AP-IVYPSO-BP demonstrated superior performance compared to traditional BPNN, PSO-BP, GA-BP, and IVY-BP models.
• Achieved high accuracy on the test set with R² of 0.9542, MAE of 3.0404, and RMSE of 3.7991.
• The model effectively balanced global exploration and local exploitation, mitigating convergence issues.

Conclusions:

• The proposed AP-IVYPSO-BP model offers a highly accurate and reliable solution for predicting HPC compressive strength.
• This bio-inspired approach provides practical value for civil engineering applications and materials design.
• The model's enhanced predictive capabilities contribute to more robust and sustainable construction practices.