基于模态联合仿真寻优法的机床刀具结合部参数辨识方法

为保证机床加工时的切削稳定性，获得准确的稳定性叶瓣图（切削速度-切削深度关系图），需要得到机床刀具刀尖点的频率响应函数（FRF）。刀尖在加工时处于旋转状态，无法通过粘贴式传感器直接获取其频率响应函数。针对这一问题，提出了通过准确辨识刀柄结合部间动力学参数来预测刀尖频响。为实现刀柄结合部的动态参数识别，提出了基于模态分析理论的联合仿真寻优法的辨识方法。该方法通过均布弹簧-阻尼单元模拟结合部接触特性，以模态实验测得前2阶模态固有频率及其对应的振幅为目标约束，以结合部间的刚度和阻尼为变量进行有限元分析，采用最小二乘法构建二者的多目标优化函数，并在MATLAB-ANSYS集成环境下建立非线性规划函数寻优任务实现结合部动力学参数的辨识。为了提高辨识效率减少调用有限元分析的次数，采用了样本点构造的方法；为了避免因采样点选取不合理而造成计算量增加，采用了采样区间归一化的处理方式。最后，将优化识别结果通过弹簧单元指令代入有限元模型进行谐响应分析来预测刀尖FRF，对比有限元仿真与模态试验对应测点的频响数据。结果表明，二者的频率响应函数曲线拟合度较高，且前2阶的固有频率误差分别仅为0和0.04%。验证了该方法的可行性，为进一步获得稳定叶瓣图提供支持。

Parameter Identification of the Tool Shank Based on the Method of Modal Joint Simulation Optimization

To ensure cutting stability and obtain accurate stability lobes diagram (diagram of cutting speed-cutting depth), need to get the frequency response function (FRF) of the tool nose. The tool tip is in a rotating state during processing, and its frequency response function cannot be directly obtained through the adhesive sensor.Therefore the dynamic parameter identification of the shank structure is the premise for accurate prediction the tool nose''s FRF. In order to solve this problem, based on the theory of modal analysis, the method of modal joint optimization simulation was used to identify the dynamic parameter of the joint part. Through the evenly distributed spring damping units to simulate contact characteristics, the modal frequency and amplitude of the modal experiment and finite element analysis were served as the optimized constraints and optimized goals, which were combined with least square method to construct multi-objective optimization function, The nonlinear programming function optimization algorithm was used under MATLAB-ANSYS integration environment to realize the junction parameters identification. By used the method of construct the sample points in order to reduce the number of calls the ANSYS finite element analysis.By used the method of design variable sampling interval normalization process in order to avoid the unreasonable sampling. Brought the optimize recognition results into Ansys model for forecasting tool point FRF. The result showed that the two curves fitting degree of frequency response function are high and the first two order natural frequency of harmonic response analysis error are 0 and 0.04%. The feasibility of this method is verified, which can provide support for further obtaining stable lobe pattern.