Hybrid MRGNN-HSTO approach for thermal and mechanical effects on the durability of steel fibre-reinforced concrete

Steel Fibre-Reinforced Concrete (SFRC) enhances structural strength and crack resistance through steel fibre reinforcement. Ensuring SFRC's performance under high strain rates and elevated temperatures requires robust testing and quality control. This research employs the Split Hopkinson Pressure Bar (SHPB) method to evaluate SFRC's mechanical behaviour under dynamic loading conditions. A novel hybrid approach integrating the Multi-Relational Graph Neural Network (MRGNN) with the Hazelnut Tree Search Optimisation (HSTO) is proposed to optimise predictive accuracy in SFRC analysis. MRGNN is used to predict the punching capacity of concrete slabs and the HSTO technique is used to optimise the weight parameter of the MRGNN method. The key goal of the proposed strategy is to enhance the predictive accuracy and optimisation of SFRC performance under high strain rates and elevated temperatures. The proposed strategy is the operational MATLAB platform. The proposed strategy achieves a significantly lower error rate of 0.016%. In terms of accuracy and compressive strength (CS), the proposed technique also excels, with an accuracy value of 0.94% and a CS of 4.5 MPa. These results validate the efficiency of the MRGNN-HSTO approach in optimising SFRC performance, making it a promising tool for structural durability assessment under extreme condition