Under the continuous advancement of new urbanization, the complexity of building structures has been increasing, and the traditional safety management model relying on manual inspection and single-point monitoring has been unable to meet the monitoring requirements for structural performance under intelligent construction conditions. This paper addresses the heterogeneity of multi-source monitoring data and noise interference, and constructs an analysis framework integrating preprocessing, bidirectional temporal modeling of structural perception, and multi-objective optimization. It introduces a multi-scale attention mechanism and an improved particle swarm algorithm to jointly optimize key parameters. Based on three types of engineering monitoring data sets, experiments show that this method reduces the health degree prediction error to 0.074, with an accuracy rate of 93.8% and an F1 value of 92.7%, which is approximately 2% higher than the optimal comparison model. It also maintains good robustness and real-time performance even under noise enhancement and missing monitoring points, and has application value for building structure safety monitoring in the context of intelligent construction.
Povzetek: For the structural safety monitoring under the conditions of intelligent construction, this paper constructs an integrated analysis and optimization framework for multi-source monitoring data, integrating hierarchical coding, time series modeling and multi-objective optimization, to achieve the joint assessment of structural health and status level. The engineering measurement results show that this method reduces the health degree prediction error to 0.074 and increases the accuracy rate to 93.8%. It still has good robustness and real-time performance even in the scenarios of enhanced noise and missing monitoring points, providing technical support for the full life cycle monitoring and intelligent operation and maintenance decision-making of building structures.
