Experimental Study On Milling Force Based On Neural Network

Artificial neural network (ANN) has a strong learning and computing power which canmore easily achieve nonlinear mapping and simulate some of the human brain functions.Artificial neural network model is divided into the feed forward neural network model and thedynamic neural network model. This paper mainly studies on the application of the feedforward neural network in the prediction of the milling force. Forward neural network modelis divided into a single artificial neuron structure, single-layer neural network structure andthe multilayer neural network structure. In this paper, taking into account the timevariant andnon-linear characteristics of the milling force, we adopt multilayer feed forward neuralnetwork model which also known as the BP network model.The orthogonal experiment of the milling force is done in YCM-V65A 4-axis verticalmachining center, using YDX-Ⅲ9702 piezoelectric milling dynamometer. Firstly, thevarious factors that affect the cutting force are analysed , then the orthogonal experimentaltable is designed based on the orthogonal experiment, in the last the experiment is doneaccording to the orthogonal experiment.The neural network toolbox in MATLAB is very powerful which encapsulates a lot ofperformance function. In this paper, the correlation functions are called to achieve a networkdesign, weight initialization, network training and simulation.In this paper, the modeling and simulation of the milling force is accomplished accordingto the experimental data and artificial neural network in order that the model can moreaccurately predict the milling force according to the different milling parameters. The resultsof the experiment and simulation indicate that the milling force model established by the BPneural network can more accurately predict the milling force.