Development of an Artificial Intelligence Model for Tool Wear eva‎luation and Remaining Life Estimation in the Drilling Process Using Machine Vision

9/18/2025 12:00:00 AM

Drill tool wear in the drilling process is a major challenge in the industry, as excessive wear leads to reduced dimensional accuracy, increased cutting forces, and ultimately, potential damage to the workpiece. Various factors, such as cutting speed, feed rate, tool material, and environmental conditions, affect the extent of wear. Machine vision, with advanced techniques like texture analysis, edge detection, and artificial neural networks, can automatically identify worn areas on the tool and predict the wear progression. This approach enables optimization of machining conditions and prolongs tool life, leading to improved productivity and reduced maintenance costs. Detecting and predicting tool wear is crucial for cost reduction and preventing damage to both the tool and the workpiece. In this study, artificial intelligence algorithms based on machine vision were developed and employed to analyze visual data collected from the drilling process under various machining conditions, classifying them into different wear categories. The method provides an estimate of tool life, indicating the number of additional holes that can be drilled with a specific bit under defined conditions. For this research, the YOLO and U-Net algorithms were used, achieving accuracy rates above 90%, and long short-term memory (LSTM) networks were applied for tool life estimation, yielding results with an RMSE error of less than 5.13 and an MAE of 3.50.

هوش مصنوعي , سوراخكاري , بينايي ماشين , يادگيري عميق , يادگيري ماشين , ماشين‌كاري سنتي , يولو , يونت , شبكه‌هاي حافظه بلند-كوتاه مدت

Artificial intelligence , Drilling , Machine vision , Deep learning , Machine learning , Traditional machining , YOLO , Unet , LSTM

Rasoul Vahedi Gandomani

Dr. Behnam Davoudi and Dr. Mohammad Shahbazi