Assessment of soil classification based on cone penetration test data for Kaifeng area using optimized support vector...

Area of Science:

• Geotechnical Engineering
• Machine Learning Applications
• Computational Science

Background:

• Traditional soil classification methods are often expensive and time-consuming.
• Accurate soil analysis is crucial for the success of engineering projects.

Purpose of the Study:

• To develop and validate enhanced Support Vector Machine (SVM) models for soil classification.
• To improve the accuracy and efficiency of soil analysis using Cone Penetration Test (CPT) data.

Main Methods:

• Trained SVM models using 649 CPT datasets with cone tip resistance and sleeve friction as inputs.
• Applied 25 optimization algorithms to enhance SVM model performance.
• Validated models against an independent dataset of 208 CPT records.

Main Results:

• 23 optimization algorithms improved SVM classification accuracy.
• 18 algorithms surpassed the accuracy of the "Code for in-situ Measurement of Railway Engineering Geology."
• Thermal Exchange Optimization (TEO) algorithm yielded the most significant improvement, increasing accuracy by 10%.

Conclusions:

• Integrating evolutionary algorithms with SVM offers a more efficient and accurate method for soil classification.
• Optimized SVM models provide superior performance compared to existing engineering standards.
• This approach holds significant promise for advancing geotechnical engineering applications.