论文部分内容阅读
传统的动态稳健参数设计方法(田口方法)虽然在工业生产实践中展现了极大的方便,但是其本身也存在较大的改进空间.当调节变量不存在时,传统的田口方法难以实现;此外,田口方法只能根据所选取的参数水平得到最优参数组合,而这种所谓的最优结果有时并不符合实际的需要.首先构建BP神经网络模型,利用训练后的BP神经网络获得参数设计中质量特性、噪声因子以及各参数间的动态关系;然后,利用超拉丁方抽样,计算信号与特性参数间的斜率,并由此将动态稳健参数设计的寻优问题转化为相应的非线性规划问题;最后,利用次序二次规划(SQP)算法解决并优化动态稳健参数的设计。此外,我们选取了一个简单的数据案例对本文提出的方法的有效性进行了说明.
Although the traditional dynamic robust parameter design method (Taguchi method) shows great convenience in industrial production practice, there is still a lot of room for improvement, and the traditional Taguchi method is difficult to achieve when there is no adjustment variable. In addition , Taguchi method can only get the optimal parameter combination according to the selected parameter level, and this so-called optimal result sometimes does not meet the actual needs.First, the BP neural network model is constructed, and the trained BP neural network is used to obtain the parameter design Then, using the super latin square sampling to calculate the slope between the signal and the characteristic parameters, the optimization problem of dynamic robust parameter design is transformed into the corresponding nonlinear programming Problem; Finally, the use of sequential quadratic programming (SQP) algorithm to solve and optimize the design of dynamic robust parameters. In addition, we select a simple data case to illustrate the effectiveness of the proposed method.