Integrating Dual-Parameter Soil Classification with CPT-Driven Machine Learning for Site Investigation of Offshore Wind

Accurate geological characterization is crucial in offshore engineering projects. This study aims to develop an advanced methodology that extends the application of Cone Penetration Testing (CPT), providing a more precise and detailed classification of subsurface soils in the locations without coring. Then, the methodology integrates advanced techniques to achieve precise geological characterization. It initiates with a dual-parameter clustering analysis, a process pivotal in uncovering nuanced geological insights. Submerged unit weight and undrained shear strength, recognized as primary indicators, are meticulously examined in the designated coring sites. The K-means algorithm, a unsupervised machine learning technique, is employed in this clustering analysis. It iteratively refines the classification by strategically grouping data points based on their proximity in the dual-parameter domain. Then, a Convolutional Neural Networks (CNN) architecture is designed to synthesize the outcomes of the clustering analysis with the extensive dataset obtained from Cone Penetration Testing (CPT). This network undergoes a training process using a combination of current program data and historical datasets from various regions. This integrated approach capitalizes on the strengths of both Kmeans clustering and neural networks. While Kmeans excels in identifying patterns and grouping data points in a two-dimensional space, neural networks are adept at learning complex relationships within data. By combining these methodologies, the study achieves a synergistic effect, resulting in a highly accurate and detailed classification of subsea soils. Compared to conventional CPT classification methods, our approach demonstrates superior precision and granularity. By incorporating dual-parameter clustering and machine learning, the methodology refines geological classification and description, providing a more comprehensive understanding of the physical properties of subsea soils.