شناسایی الگوهای ناهنجار داده‌های شتاب‌سنج در پل‌های کابلی خودایستا به روش یادگیری نظارت شده و با استفاده از تصویر میدان زاویه‌ای گرمیان

Identifying structural damage before an accident occurs is crucial in structural health monitoring (SHM). SHM involves monitoring the condition of structures to detect anomalies and potential damages. Accelerometers are key tools in SHM for collecting structural data. However, these sensors can experience abnormalities due to various factors such as noise, sensor failure, or environmental changes. If these abnormalities are not detected, they can compromise the entire SHM system, leading to the misinterpretation of sensor failures as structural failures. Therefore, developing efficient methods for detecting and correcting sensor anomalies is essential.
This paper evaluates different types of sensor anomalies using real bridge benchmark data. The proposed method employs supervised machine learning with a convolutional neural network (CNN) and Gramian angular field (GAF) images for anomaly detection. The results demonstrate the method’s efficiency and accuracy, achieving a 93% success rate. The research assesses the performance of the proposed method on normal data and various anomalies, showing excellent performance for normal data and certain anomalies, such as trend anomalies. Implementing this method can significantly improve the accuracy of anomaly detection in SHM systems.