Four-limb CFST latticed columns seismic performance: experimental and ANN predictions

Area of Science:

• Structural Engineering
• Seismic Performance Analysis
• Computational Mechanics

Background:

• Concrete-filled steel tubular (CFST) lattice columns are crucial structural components.
• Accurate seismic performance prediction is vital for structural safety and design.
• Existing artificial neural network (ANN) models have limitations in predicting hysteresis curves.

Purpose of the Study:

• To investigate the seismic performance of four-limb CFST lattice columns under horizontal low-cycle reciprocating loads.
• To enhance the predictive accuracy of seismic performance using an optimized ANN.
• To develop a novel prediction model by integrating quantum computations with the sparrow search algorithm (SSA).

Main Methods:

• Experimental testing of four-limb CFST lattice columns with varying slenderness and axial load ratios.
• Development of an ANN model to predict seismic performance.
• Optimization of ANN weights and thresholds using the sparrow search algorithm (SSA) enhanced with quantum computations (QMESSA).

Main Results:

• Experimental results showed distinct hysteresis curve shapes (bow, inverse S, pike) indicating varied plastic capacity and failure modes.
• The QMESSA effectively optimized the ANN, leading to accurate predictions of load-displacement hysteresis curves.
• Predicted damage variables from QMESSA showed strong agreement with experimental data and conventional damage models.

Conclusions:

• The seismic performance of four-limb CFST lattice columns is significantly influenced by slenderness and axial load ratios.
• QMESSA demonstrates superior performance in optimizing ANNs for predicting the seismic behavior of CFST lattice columns.
• The proposed QMESSA offers a reliable and accurate method for seismic performance assessment of these structural elements.