Thermography Based Advanced Intelligent Tool Condition Monitoring System

Introduction Intensive development is taking place in the field of improvements of CNC machines. Taking into account small series and single item production, the turning process, despite the advancements in technology, still requires an operator. Tool wear monitoring is one of the reasons why operators are needed. The challenge in the field of tool wear is: “When should one change a cutting tool?” This is decided by a machine operator based on their expertise and informal knowledge. If the tool wear is not properly considered by the operator when replacing the cutting tool, one of two scenarios happens – either the tool is replaced too soon or too late, and both decisions have negative consequences. In practice, both scenarios occur regularly, especially due to less experienced operators, who cannot make the right decisions about cutting tool replacement, and due to the lack of a reliable and favourable system that would make good decisions autonomously. The main idea of the research project Detection of cutting tool wear and the decision to change a cutting tool will be automated, independent of the operator’s knowledge and experience. Problem identification - Turning and temperatures The turning process is characterized by high temperatures, which are also closely connected to the state of the tool wear. The Tool Condition Monitoring (TCM) system developed in this research project that determines whether the cutting tool is suitable for further machining or not will be based on thermographic images of the cutting tool. The existing temperature based TCM systems have focused on the absolute cutting tool temperature measurement, the system proposed in this research is different – it is based on the cutting tool temperature distribution (relative temperatures). The tool wear will be categorized based on the features of the image, so the whole thermographic image will be an input to the Neural Network. Specifically, a Convolutional Neural Network (CNN) will be used as a Deep Learning method. Objective of the proposed research The main objective To develop a reliable Tool Condition Monitoring (TCM) system based on a cutting tool’s IR image using Deep Learning methods. A large database, also known as big data, is essential for Deep Learning algorithms, because they require large amounts of data to train on in order to produce accurate results. However, there is no existing database of specific images for turning processes. Within the project, an extensive database of thermographic images will be created, which will be freely available to other scientists and will be a major contribution to the development of science in the field of tool condition monitoring using artificial intelligence. It is difficult to provide a large enough images database of turning with a worn tool. The GAN (Generative Adversarial Network) method will be used to create additional images based on the captured images. This will ensure a balanced database for all categories of tool wear. Short description of the work programme The entire project is divided into 4 work packages: WP1 - PRE-EXPERIMENTAL PHASE (D1.1 Detailed experimental plan) WP2 – EXPERIMENT (D2.1 Image database, D2.2 Extended image database, D2.3 Big data) WP3 – GAN (D3.1 GAN-based extended database) WP4 – CLASSIFICATION (D4.1 Trained model for classification, D4.2 Deep NN) Conclusion The study aims to create a system that utilizes artificial intelligence to monitor tool wear during turning processes. The validation of the method under ideal conditions has already been done, and it will now be further developed under actual industrial conditions, including cooling, and extended to various turning parameters. The findings will be relevant to other machining methods where heat is generated (e.g. milling).