Artificial Intelligence Based Tool Condition Monitoring in Machining

Abstract

In a manufacturing environment, gradual and catastrophic failure of a cutting tool are common faults associated with a machining process. Left unmonitored these failures exhibit a high likelihood of triggering workpiece surface quality issues and reducing the overall productivity of the process. This research offers a comprehensive study of the design, development and implementation of a low-cost artificially intelligent (AI) tool condition monitoring (TCM) system for the turning operation of different workpiece material. As the characteristics of the signals from the fusion of multiple sensors differ from each individual sensor, fast Fourier transform (FFT) and wavelet transform (WT) were used to identify signal features which provided the most useful information about the cutting tool conditions. Pearson correlation coefficients (PCC) and principal component analysis (PCA) singled out the most sensitive features from the sensor fusion signals which play a pivotal role in monitoring gradual tool wear (flank wear) progression during machining. Experimental results indicated that the total harmonic distortion (THD), crest factor of the spindle motor current and certain spectral features from the vibration sensor signal had a significant correlation with cutting tool flank wear under different cutting conditions and when machining different materials. Furthermore, a data-driven artificial neural network (ANN) and an adaptive neuro-fuzzy inference system (ANFIS) were studied to investigate their prediction accuracy for cutting tool conditions (tool wear). The ANFIS prediction model outperformed the ANN prediction in terms of accuracy. As a result, the developed (trained and validated) ANFIS based TCM system was implemented in the material turning process. In addition, the adaptability of the supervised ANFIS based TCM system was shown to further increase the reliability of tool wear prediction.