利用径向基函数神经元网络估测信号概率密度识别镗削加工中的颤振

Estimation of Signal Probability Density Using Radial Basis Function Neural Networks and Rccognition of Chatter in Boring Process

提出一种检削加工中预报颤振的新方法。通过理论分析得出颤振和未颤振时力信号动态分量分别服从不同的概率分布，且它们的差异十分明显。同时利用径向基函数神经元网络在线的估测力信号的概率密度，得出和理论分析相近的结果，基于此提出一新的颤振预报准则。该方法在实际应用中具有较高的准确性和快速性。

In this article one new approach to forecastthe chatter in boring process is presented. It can be concluded by analyzing theoretically that there are distinctlydifferent probability distributions of the dynamic cuttingforce under the stable cutting condition and the chatterrespectively. At the same time, the probability density ofdynamic cutting force can be estimated on line by radialbasis function neural networks. Therefore, a new judging standard can be applied in the recognition of chatter.and this proposed approach can be implrmented veryclutchly and accurately.