Tool condition monitoring of titanium alloy end mill process based on artificial intelligence model using transfer learning
- 문서 유형
- 연구출판물 · 학술대회 발표논문
- 등재
- Scopus
- 피인용
- Scopus 0
저자
초록
Titanium alloy has low thermal conductivity during cutting, so a lot of heat is generated, which leads to a short tool life and difficult tool management. When machining titanium alloy, it is important to detect the condition of the tool and determine the appropriate replacement time when a large amount of machining is required. Therefore, an AI-based analysis model was applied to monitor the end mill machining process status of titanium alloy. Many cases have been reported in which AI models combining CNN and LSTM directly applied time series data to design AI models with high accuracy and reliability. But AI models learned through deep learning have a limitation in that the accuracy for input values that deviate from the learned conditions are low. To overcome this, it is necessary to build data for all possible conditions and learn through deep learning, but this is realistically difficult. Recently, efforts have been made to solve the problems of data shortage or increased learning time by methods such as transfer learning or hyperparameter tuning that maintain the weights of existing AI models and change some layers. To obtain acceleration sensor data required for deep learning, we performed a machining experiment on titanium alloy using an end mill tool. To identify the relationship between the sensor signal and the tool condition, we periodically measured tool wear using an optical microscope. Based on the acquired dataset, we developed an AI model suitable for tool monitoring that classifies the tool condition according to the machining signal using deep learning combining CNN and LSTM. In addition, we applied and reviewed a method to improve the classification accuracy of the tool condition according to new machining conditions using transfer learning based on this. Through this, we understood transfer learning and reviewed solutions for solving the problem.
키워드
Artificial intelligence modelEnd millTitanium alloyTool monitoringTransfer learning
식별자
- Scopus EID
- 2-s2.0-105052008639



로그인
연구물 목록으로