IoT-Enabled Bridge MonitorX: A Machine Learning Approach for Real-Time Damage Detection and Vibration Analysis

The damage of bridge infrastructure is a serious threat to the security of the people and constant monitoring is necessary to detect any damages at the initial stages. In this paper, the complex and affordable IoT-based bridge health monitoring system is presented, which combines several sensors and machine learning algorithms to detect anomalies and classify damage in real-time. It uses Arduino Uno microcontroller, which is attached to SW-420 vibration sensors, DHT11 temperature-humidity sensors, LM35 temperature sensors, LM150 strain gauges, and SX1278 LoRa wireless data transmission modules. Random Forest, Support Vector Machines (SVM), and Decision Trees are ensemble machine learning algorithms that are employed to maintain an analysis of the environmental and structural parameters on a continuous basis. The Isolation Forest algorithm proved to be highly efficient in detecting anomalies being able to differentiate standard structural behavior and possible damage. The comparative analysis of four classifiers (Random Forest, SVMRBF, SVM-Linear and Decision Tree) demonstrated strong results in all measures with accuracy, precision, and recall, and F1-score more than 85 percent. The importance score of vibration was the highest (0.483), then temperature, and strain measurements in the analysis of features and important parameter. The performance of the classification models was high as shown by the ROC curve analysis with AUC values of close to 1.0. It is a cost-efficient measure because it allows proactive maintenance measures, decreases the cost of inspection, and increases the safety of bridges by implementing continuous structural health monitoring.