نوع مقاله : مقاله پژوهشی
نویسندگان
1 واحد فناوری اطلاعات، شرکت فولاد آلیاژی ایران، یزد، ایران
2 کارشناسی ارشد هوش مصنوعی و رباتیکز، شرکت توسعه فولاد آلیاژی ایرانیان، یزد، ایران
3 استاد، دانشکده مهندسی کامپیوتر، دانشگاه یزد، یزد، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
This paper addresses the problem of detecting defective round steel billets during their rolling motion. In the finishing hall of the Iran Alloy Steel Factory, one of the essential operations applied to round billets is straightening. However, no system currently exists to automatically assess the quality of the straightening machine’s output. To address this challenge, videos of billets rolling on the discharge table after leaving the straightening machine are recorded. After preprocessing and illumination balancing, the cross-section of each billet in every frame is detected using the Circular Hough Transform method. The motion trajectory of the billet in each video is then determined, and six motion-related features are extracted for analysis. These features include Trajectory Length, Sum of Squared Deviations, Sum of Squared Vertical Deviations, Vibration of Standard Deviation, Wavelet Vibration Energy, and Standard Deviation of Acceleration. The values of these features are computed with normalization based on sampling per unit length. Experimental results demonstrate that features such as Vibration of Standard Deviation, Sum of Squared Vertical Deviations, and Sum of Squared Deviations show strong discriminative power in separating healthy billets from defective ones. Although the false negative rate is somewhat high, the proposed method achieves an overall accuracy of about 85%, which is considered acceptable for practical industrial application.
کلیدواژهها [English]
)