三油楔固定瓦滑动轴承动特性解析模型的建立

Analytic Model Building of the Dynamic Characteristics of Three-lobe Bearing

以通过数值计算所得轴承不同结构参数下的动特性系数作为原始数据,采用多元非线性回归方法建立滑动轴承动特性系数的非线性回归解析模型.回归过程中,采用拟牛顿法与通用全局优化算法相结合的混合优化算法,以非线性最小二乘估计为准则,通过迭代计算获得模型的回归参数,建立了一种无初始模型的数据建模方法;实现了自动寻找隐含于数据间的数学模型,且无需人为提供参数初估值;以残差值为依据验证了所建模型的有效性.结果表明,所建解析模型保证了计算精度,同时克服了数值计算效率低的问题.

Taking the values of dynamic characteristic coefficients obtained under different structure parameters of bearing as the original data, the nonlinear regression models of journal bearing were obtained by using the method of multivariate nonlinear regression. By using the hybrid optimization algorithm, which combines BFGS with universal global optimization （UGO）, and taking the nonlinear least squares estimation as criterion, the regression parameters of nonlinear regression models were obtained through iterative calculation. The results show that the data modeling way without initial model is built and the connotative mathematic models which are searched automatically among data are realized without the first guess by people. According to the value of residual difference the validity of the new models was verified. The com- putational precision is ensured and the problem of inefficient numerical calculation is gotten over by new regression models. An efficient way is offered for actual applications.