The implementation of computerised condition monitoring systems for the detection cutting tools' correct installation and fault diagnosis is of a high importance in modern manufacturing industries. The primary function of a condition monitoring system is to check the existence of the tool before starting any machining process and ensure its health during operation. The aim of this study is to assess the detection of the existence of the tool in the spindle and its health (i.e. normal or broken) using infrared and vision systems as a noncontact methodology. The application of Principal Component Analysis (PCA) and Discrete Wavelet Transform (DWT) combined with neural networks are investigated using both types of data in order to establish an effective and reliable novel software program for tool tracking and health recognition. Infrared and visual cameras are used to locate and track the cutting tool during the machining process using a suitable analysis and image processing algorithms. The capabilities of PCA and Discrete Wavelet Transform (DWT) combined with neural networks are investigated in recognising the tool's condition by comparing the characteristics of the tool to those of known conditions in the training set. The experimental results have shown high performance when using the infrared data in comparison to visual images for the selected image and signal processing algorithms.

Abstract-This paper outlines a comparative study to compare between using the power of the spindle and the infrared images of the cutting tool to design a condition monitoring system. This paper compares the two technologies for the development of a tool condition monitoring for milling processes. Wavelet analysis is used to process the power signal. Image gradient and Wavelet analyses are used to process the infrared images. The results show that the image gradient and wavelet analysis are powerful image processing techniques in detecting tool wear. The power of the motor of the spindle has shown less sensitivity to tool conditions in this case when compared to infrared thermography.

No abstract