Abstract: Application of pattern recognition technique to recognition of cutting states for difficult-to-cut materials is introduced in this paper.It is the first step for realization of the adaptive control in cutting processes.A pattern dichotomizer can be adopted to judge whether BUE appears.A more general model of pattern classifier must be designed in order to classify chip shapes into three categories.The major problem in classifier design is the determination of discriminant function or weight vector.First of all,it is necessary to select the components or the features of the pattern vector properly by experience.The results of processing random signals in cutting experiments show that there exist relation-shaps between cutting phenomena and autocorrelation function or auto power spectrum.Therefore both the cutting parameter and the character of autocorrelation function or that of auto power spectrum can be selected as the components of pattern vector.It is possible to find out the discriminant function or weight vector by learning process with the aid of electronic computer.There were more than 60 different cutting states observed in cutting experiments.The discriminant functions determined by several training patterns have been used to check the remaining patterns,and the results show that the recognition of cutting states is basically successful and the rate of success is greater than 70%.The next step of our work will be to enhance the accuracy of recognition.