<p>Structural health monitoring (SHM) systems are essential to ensure safety, reliability, and extended lifespan using sensors, data acquisition technologies, and machine learning algorithms. Artificial intelligence algorithms are used to accurately and early assess the condition of concrete buildings to ensure safety and prevent disasters. This paper proposed and implemented a real-time anomaly detection system that consists of a transmitting unit, a gateway unit, and a novel anomaly detection approach called the low-pass filtered regression approach for anomaly detection (LPFRA-AD) to identify the damage severity accurately. The transmitting unit comprises multiple sensors linked to the building structure to read the structure’s status. LoRa SX1278 Ra02 communication technology transfers the information from the transmitter side to the gateway unit. The gateway processes the data and transmits it to the cloud through Wi-Fi. The proposed LPFRA-AD approach consists of a low-pass filter to remove irrelevant structure signals and the multiple linear regression (MLR) technique to analyze and estimate the damage intensity. The system is verified by exposing the proposed building structure to several peak ground acceleration (PGA) intensities and comparing the results using MAE, RMSE, and R² with the MLR technique and other algorithms. The results showed that the presented method is more accurate and reliable, with RMSE and R² equal to 0.0248 and 0.954, compared to 0.107 and 0.075 for the traditional MLR technique. Moreover, the ThingSpeak IoT platform is established with a gateway unit to store the structure’s signals. It relates to a pre-trained adopted approach for enabling real-time anomaly detection.</p>