Optimizing concrete crack detection an echo state network approach with improved fish migration optimization.

Area of Science:

• Civil Engineering
• Artificial Intelligence
• Materials Science

Background:

• Concrete structures are susceptible to cracks from various factors, impacting safety and structural integrity.
• Accurate concrete crack detection is crucial for infrastructure assessment, but existing models face challenges like overfitting and computational complexity.
• Current methods often lack the efficiency and robustness required for reliable crack identification and classification.

Purpose of the Study:

• To propose an optimized model for concrete crack detection and classification.
• To enhance the accuracy and efficiency of concrete crack identification using artificial intelligence.
• To address the limitations of existing concrete crack detection models.

Main Methods:

• An enhanced fish migration optimization (IFMO) algorithm was developed.
• An optimized echo state network (ESN) model was integrated with IFMO for improved performance.
• The proposed ESN/IFMO model was evaluated on the SDNET2018 dataset.

Main Results:

• The ESN/IFMO model demonstrated superior performance in concrete crack detection.
• Accuracy was increased by 3.75-8.19% compared to state-of-the-art models (DL, DINN, AlexNet, CNN, LSTM).
• The F1 score saw an improvement of 5.14-12.55% over existing approaches.

Conclusions:

• The ESN/IFMO model offers a more effective solution for concrete crack identification.
• The optimized ESN arrangement significantly boosts detection accuracy.
• This approach shows strong potential for real-world infrastructure safety assessment.