نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی کارشناسی ارشد مکانیک سنگ، دانشکده مهندسی معدن، دانشگاه صنعتی امیرکبیر، تهران، ایران
2 دانشیار مکانیک سنگ، دانشکده مهندسی معدن، دانشگاه صنعتی امیرکبیر، تهران، ایران،
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
This study investigates the effectiveness of non-destructive ultrasonic testing for detecting and classifying damage in cement mortar specimens with different mix designs. To this end, a set of specimens with varying cement content (CC) were prepared and tested in both undamaged and damaged states. Ultrasonic signal data, after preprocessing and the extraction of statistical and time–frequency features, were used as inputs to three machine-learning algorithms: K-Nearest Neighbors (KNN), Naïve Bayes, and Random Forest. Analysis results indicated that the Naïve Bayes classifier, owing to its ability to handle independent and uncorrelated features of the ultrasonic data, outperformed the other methods and yielded the highest classification performance, achieving an accuracy of 99.00 ± 3.16% and a recall of 97.50 ± 7.91%. The main innovation of this research is the combination of machine-learning approaches with non-destructive methods to analyze energy variations in ultrasonic signals and to enable early damage detection in cement mortars with different cement content ratios. This data-driven, reproducible analytical framework not only improves the accuracy and reliability of damage identification but is also extensible for health monitoring and continuous assessment of concrete structures at practical scales, serving as an effective tool to enhance the durability and safety of civil infrastructure.
کلیدواژهها [English]
)