Axial capacity of GFRP-concrete-steel composite columns and prediction based on artificial neural network

GFRP-concrete-steel composite columns (GCS) are formed by combining different components with excellent mechanical properties. The GCS exhibits good corrosion resistance and can serve as a pile foundation platform for offshore wind turbines. An Artificial Neural Network (ANN) is utilized to swiftly predict the loading capacity of GCS columns with various geometrical parameters. The prediction error is primarily controlled within 10%. The thickness of GFRP tubes (t(G)), the inner diameter of GFRP tubes (D-G,D-i), the thickness of steel tubes (t(s)), the outer diameter of steel tubes (D-S,D-o), the strength of GFRP tubes (f(G)), the strength of concrete (f(C)), and the strength of steel tubes (f(S)) were used as independent variables to determine the key factors affecting the loading capacity of GCS columns. The impact of the number of input variables on the final prediction accuracy is analyzed. The proportional influence of different parameters on the loading capacity of GCS column is investigated using the Garson algorithm, among which, the thickness of the GFRP tube (t(G)) accounts for the largest proportion, approximately 24.57%. The suitability of the artificial neural network for predicting the loading capacity of GCS columns with various geometries is revealed. The results indicate that the ANN can be utilized for high-precision compressive strength prediction, and the thickness of the GFRP tube (t(G)) is the most influential factor on the loading capacity of GCS columns.