Research on intelligent semi-active control algorithms and seismic reliability based on machine learning

Abstract

Abstract Aiming to address the shortcomings of existing semi-active control algorithms with poor robustness and the limited generalization ability of current evaluation methods based on deterministic analysis, a novel approach based on probability density evolution is proposed. This method is designed to assess the seismic reliability, enabling a more comprehensive evaluation of the control effectiveness of aqueduct structures. Building upon this, an intelligent semi-active control algorithm leveraging machine learning is introduced. The algorithm is further validated through engineering case studies to investigate semi-active control strategies in response to random seismic events. The results show that the seismic reliability of the machine learning-based semi-active control algorithm is significantly higher than that of the uncontrolled state for the same failure threshold under random seismic effects.

Keywords

• Semi-active control
• Probability density evolution
• BP neural network
• Modified Optimization Algorithm for Sand Cat Colonies
• Seismic reliability