Visual Inspection of Drill Cutting Tool Using Deep Learning

Visual inspection of drill cutting tools is an important step in any production process that requires precision and speed. Due to the limits of manual visual inspection using the human eye in terms of precision and focus, a system has been developed that can do visual inspections using artificial intelligence with deep learning algorithms. The company that produces drill cutting tools must maintain quality to avoid sending rejected items to customers. In the manufacturing process, drill cutting tools are made through various processes and monitored during the inspection process by the QC (Quality Control) department. In the inspection process, the QC team still has shortcomings such as visual acuity due to differences in light intensity at different times, reduced concentration when working close to the end of the workday, and varying precision among each QC member. So sometimes there are items that pass QC inspection and its sent to customers, which can affect customer satisfaction. Therefore, a visual inspection system was created that include an algorithm for detecting defects using deep learning. Detection using the deep learning method is expected to have high accuracy and faster detection. In the design process using the CNN (Convolutional Neural Network) algorithm with MobileNet V2 architecture, it starts with image capture, followed by annotation on the drill cutting tool part, then training for 40,000 steps, and after training, testing is conducted. This solution is expected to accurately detect defects, thereby maintaining product quality.