Security Assessment Algorithm of Intelligent Buildings Based on Data Analysis

Traditional building safety assessment methods have problems such as single data dimension and insufficient building safety capabilities. This paper proposes an intelligent building safety assessment algorithm framework based on multi-source heterogeneous data fusion, collects structural response data including strain, displacement, and vibration frequency, and constructs a multimodal data set with spatiotemporal correlation; uses an improved variational autoencoder to extract deep features of the original signal, and designs a hybrid assessment model based on a dynamic Bayesian network and a long short-term memory network (LSTM). The mean error of damage location of the proposed algorithm is only 1.08 mm, the crack width error is reduced, and the material aging assessment error is controlled at 2.0–5.0%. This study provides a high-precision and explainable decision support framework for the safe operation and maintenance of intelligent buildings throughout their life cycle, and promotes the paradigm shift of building safety management from experience-driven to data-mechanism fusion.