Research on Key Technologies of Embedded Systems in Building Health Monitoring and Structural Safety Assessment

Structural Health Monitoring (SHM) is crucial for ensuring the safety and longevity of civil infrastructure by providing real‐time data on the condition of structures. Existing SHM systems struggle for simulating dynamic situations and intricate geometry, while using Partial Differential Equations (PDEs) and Principal Component Analysis (PCA) has high computing costs, difficulties in deploying sensor networks, and the possibility of errors in real‐time data processing. Many systems still rely on basic algorithms, highlighting the need for more sophisticated approaches to enhance the accuracy and efficiency of damage detection, particularly when dealing with noisy or variable environmental conditions. This gap was addressed by integrating embedded systems, sensor networks, and advanced Machine Learning (ML) techniques, specifically the Kernelized Support Gradient Vector Machine (KSGVM), for real‐time Building Health Monitoring (BHM) and structural safety assessment. The system incorporates accelerometers, strain gauges, and temperature sensors, all integrated with Raspberry Pi for accessing and processing sensor data. It utilizes publicly available experimental datasets, consisting of time‐series data gathered from concrete structures under various operational conditions. These datasets include measurements of acceleration, strain, and temperature, and undergo preprocessing techniques such as noise filtering, data normalization, and missing data imputation using interpolation to ensure high‐quality input for the ML models. Linear Discriminant Analysis (LDA) is employed for feature extraction, while Recursive Feature Elimination (RFE) is used for feature selection. The system, developed and simulated using Python and libraries such as scikit‐learn and TensorFlow, demonstrated robust damage detection performance. KSGVM significantly outperformed traditional ML algorithms in terms of classification accuracy (0.9882), F1‐score, precision and recall. Additionally, the KSGVM model showed strong performance in predicting damage severity, and the system proved resilient to noisy sensor data, making it suitable for real‐time SHM applications in dynamic and diverse environmental conditions such as buildings, monitoring bridges, and other critical infrastructure.