The present study shows the development of in-process tool condition monitoring systems utilizing signal decomposition technique, statistical data analysis, and artificial neural networks system. Two systems; (1) the system based on the multiple regression, (2) the system based on artificial neural networks with back-propagation learning algorithms were developed;The raw signals obtained from two sensors (tri-axial accelerometer and AE sensor) with different machining parameters and tool conditions were examined and decomposed into six components by utilizing a wavelet transformation. The most significant components of each signal were found by statistical method and implemented to develop two in-process tool monitoring systems;Before the multiple regression system was developed, a statistical process was performed to eliminate the effects of machining parameters from the signals of the accelerometer and AE sensor. The prediction performance improved 12.6% from the process;In order to maximize the benefit of artificial neural networks system in tool monitoring systems, a novel approach was performed in this study. A great number of networks structures were tested systemically to find an optimized structure for the artificial networks tool condition monitoring system. The technique provided benefits of not only saving time but also testing all possible structures more accurately compared with the traditional manual trial-and-error methodology;The developed statistical multiple regression tool condition monitoring system showed 90% accuracy, and the developed artificial neural networks tool condition monitoring system showed 97% accuracy from 151 tests with the reject flank wear size of 0.00787 inch (0.2 mm) or larger;The successful development of the tool condition monitoring systems can provide a practical tool to reduce downtime related with tool changes and minimize the amount of scrap in metal cutting industry. Implications of the study and recommendations for further research were provided.