Akademik Veri Yönetim Sistemi
9TH INTERNATIONAL ANTALYA SCIENTIFIC RESEARCH AND INNOVATIVE STUDIES CONGRESS, Antalya, Türkiye, 20 - 23 Kasım 2025, ss.579-589, (Tam Metin Bildiri)
Introduction and Purpose: Real-time monitoring of tool wear in turning is important for process reliability and economics. Since turned-surface images carry wear traces, feature-based approaches derived from machined surface images have increased in recent years; in most studies, analyses were conducted on images containing multiple feed marks. This study investigated whether computing the previously used grayscale mean (G) feature over a single feed mark (instead of multiple marks) improves its performance in reflecting wear.
Materials and Methods: Cutting tests were performed at five different cutting speeds, and three additional experiments were conducted with varied feed rates. For each test, the G feature was computed from a single feed mark in each image. The temporal trend of G during cutting was modeled linearly versus time, and goodness-of-fit was quantified by the coefficient of determination (Adj. R²). In addition, the dependence of wear on cutting speed was jointly modeled with a logarithmic term to assess the dynamic relationship.
Results: Across the five speed-series tests, using a single feed mark yielded a 17% increase in mean Adj. R² compared with the multi-mark computation. When the logarithmic trend with cutting speed was included, an additional 9.7% improvement in the dynamic fit was obtained. In the three experiments with varied feed rate, the mean Adj. R² increased by 111%.
Discussion and Conclusion: Computing G from a single feed mark was associated with a more consistent relation to the wear trend than the multi-mark approach. The joint log-speed modeling further strengthened this relation, indicating that the single-mark strategy can serve as a simple indicator for real-time tool-wear monitoring in turning.