Real-time detection of cutter wear on industrialmachining machines using signal processing methods

Industry 4.0 is a revolution based on automation and digitalization in production processes. In this context, artificial intelligence approaches such as Machine Learning (ML) and Deep Learning (DL) are integrated with big data analytics and the Internet of Things (IoT) to create predictive maintenance systems. This increases the traceability, management, and efficiency of production processes, while reducing errors and failures and utilizing resources more effectively. With this perspective, "predictive maintenance", which plays a critical role in monitoring the performance of equipment, makes it possible to maintain machines and devices based on information and predictions based on sensor data analysis. Every machine and device used in industry has internal and external sensors. Various sensors perform real-time measurements and make it possible to create a digital image - a map - of the production process. These sensors provide control and assurance of the work and the device. The information obtained from the sensors is stored and then processed to make predictions. In my study, by evaluating the data obtained in real time with a vibration sensor from a CNC machine in an industrial facility, appropriate signal processing methods were investigated to prevent cutting tool breakage in CNC machines. A data set was created to examine the wear and breakage of cutting tools by taking a period of the part processed by the CNC machine in real time. As a result of different analyses performed in the MATLAB environment with the generated dataset, precise predictions about the current state of the tool could be made with RMS (root mean square) values, and certain threshold values were defined. If these thresholds are exceeded, the system can automatically warn the user and inform that the tool needs to be replaced. Thus, operator-dependent control processes are reduced, and automatic and uninterrupted production environments are created. In addition, the data received from the vibration sensors were also evaluated with analysis methods such as Fast Fourier Transform (FFT), Hilbert-Huang Transform (HHT), Continuous Wavelet Transform (CWT), and S-Transform (ST). In this way, the wear process of the tool, pre-fracture behavior and characteristic changes in the signals were analyzed in detail. This automated warning system reduces dependency on operators, enabling continuous and uninterrupted manufacturing processes. Furthermore, the collected vibration data were analyzed using Fast Fourier Transform (FFT), Hilbert-Huang Transform (HHT), Continuous Wavelet Transform (CWT), and S-Transform (ST) methods. These analyses provided in-depth insights into tool wear progression, pre-failure behavior, and characteristic changes in vibration signals. As a result of the analysis, it was seen that the 'Total RMS Values of All Parts' method gives much better results in wear-fracture detection. With this method, a 28% increase in value was observed between the first machined part and the pre-fracture part. In addition, it was determined that vibrations above 0.4g RMS value with linear increase can be accepted as a warning point. This data acquisition could be performed in 90ms and total analysis in as little as 0.3sec (300ms). The developed system was both compared with similar methods in the literature and tested experimentally. The system has the potential to be used in unmanned production environments such as dark factories. In this respect, it is not only an academic contribution, but also provides high value in terms of efficiency and reliability in industrial applications. As a result, predictive maintenance decisions become more reliable with the interpretation of this data, and the system's ability to provide warnings in real time prevents production downtime. Thus, while production quality is increased, cutting tool life is also optimized. Since the cutting tool can be detected very quickly in case of sudden breakage and before breakage in case of wear, lost time with tool sharpening or replacement is minimized and defective parts are prevented. This system can be considered as an important step in the transition to unmanned production.