A Study of Cutting Tool Wear Prediction Utilizing Generalized Regression Neural Network with Improved Fruit Fly Optimization

Tool wear prediction becomes increasingly important due to the growing demand for finished quality and the improvement of productivity. In this case, it is necessary to establish a well-designed monitoring system to obtain the relationship between tool wear and cutting process. Generalized regression neural network (GRNN) is able to handle non-linear problems with its memory-based character. However, it was rarely used for tool wear prediction in the past several decades. Therefore, in this paper, it was employed to tackle this problem. In addition, in order to tune the smooth parameter of the GRNN, a newly proposed evolutionary algorithm called fruit fly optimization algorithm (FOA) was adopted. Meanwhile, an improved fruit fly optimization algorithm (IFOA), in which escaping and distance control parameters were introduced to prevent FOA from falling into local optimum, was presented to enhance the search ability. Two cutting experiments showed that the IFOA-GRNN provided a comparable regression ability to the GRNN with particle swarm optimization(PSO), the least squares support vector machines(LS-SVM) and the BP neural network.