1. Industry demand
During the cutting process of the machine tool, the degree of tool wear will directly affect the quality of the machining accuracy. The current industry common practice is to change the tool in advance to ensure product quality, but the production cost is relatively higher. If the tool change is postponed, the broken and collapsed tools will cause equipment failure and defective output. Therefore, by establishing tool wear life prediction technology, the process of industrial upgrading can be accelerated.
2. Technology Introduction
(1)Introduction of HUNTER tool record system
Through the Industry Bureau of the Ministry of Economic Affairs, our company promotes introducing AI application value-added programs in the supply chain of the small and medium manufacturing enterprise, which successfully develop tool life diagnosis and prediction systems. This project uses CNC turning and milling compound center machine, the processing key components of the index plate - turret as an example, collect the processing equipment parameters and use tool measuring equipment to measure three main processing tools, plain cutter, fraise, and fine boring cutter, tracking its changing characteristics including information such as cutter length, diameter, and image. Through radio frequency identification technology (RFID), the original measurement data is written into the HUNTER system to establish the tool usage history, and transparent digital and efficient production management can be carried out through the HUNTER system.
(2)Tool life diagnosis and prediction model
Combining Convolutional Neural Networks with Recurrent Neural Networks, Convolutional Recurrent Neural Networks Long and Short-term Memory (ConvLSTM) architecture extracts database feature information.
The information includes tool diameter, tool length wear value, tool diameter wear value, tool wear area, tool image wear level, turret processing times, and last processing. Carrying out neural network model training and prediction can build a tool life prediction model.
(3)Technical integration
The project imports the tool life prediction model into the HUNTER system. This system automatically substitutes database information for tool life analysis and prediction. It also provides managers to confirm the timing of tool replacement.
Besides, it avoids waste and reduces defective products caused by changing tools earlier and damaged tools. Hunter system would improve the efficiency of tool use through automatic write-back processing to adjust machining parameters.
3. Project benefits
Through the AI tool diagnosis and prediction system, it is possible to predict the tool wear of each machining and estimate the number of times that the tool can be machined.
It also establishes a standard for the number of tool machining times to use tools effectively and reduce damages of products. AutoCam estimates that it will produce at least 7,000 turrets (processing 168,000 holes) each year, which can save tool costs about 1,130,800 NTD per year.

線上展網址：
https://tievirtual.twtm.com.tw/iframe/16434558-ade7-4bd4-8af7-67c323dbfc3c?group=f36efffd-9ee6-4962-aa09-72f6797d2449&lang=en