基于置信规则库推理的电路参数成品率估计与中心值设计方法

Circuit Parameter Yield Estimation and Centering Design Based on Inference of Belief Rule Base

将电路仿真与主观设计经验相结合,建立成品率估计的置信规则库(BRB)推理模型,该模型不仅可建模输入电路参数变量与输出参数成品率之间复杂的非线性关系,且能通过证据推理(ER)过程,给出不同输入下输出成品率的估计值.利用BRB推理模型可代替电路仿真,利用较少的计算量快速地估计出满足约束条件的中心值设计区域.所提方法与传统基于蒙特·卡罗的统计性方法和基于可接受域几何逼近的确定性方法相比,其不再只给出单个满足要求的中心值,而能够给出整个中心值设计区域,且不受特定电路性能函数形式的约束,具有计算速度快、模型参数物理意义明确便于理解等优点.以高速铁路轨道电路成品率估计与中心值设计为例,验证了所提方法的有效性.

By using circuit simulation and subjective design experience,the belief rule base（BRB）reasoning model for yield estimation was established,through which the complex nonlinear relationship between circuit parameters variables and yield was modeled and identified,as well as the yield was estimated by using evidence reasoning（ER）.By using BRB model to replace circuit simulation,the center value design areas satisfying constraint conditions with less amount of calculation were estimated.Compared with the traditional statistical method based on Monte Carlo and the deterministic method based on geometry approximation of acceptable areas,the proposed method not only presented a single center value,but also provided the whole center value design area.Moreover,it was not restricted by a specific form of circuit performance function.The proposed method showed some advantages such as fast speed calculation,definite physical meaning of model parameters for better interpretation,etc.The effectivenessof the proposed method was verified by the examples about the yield estimation and the centering design of high-speed rail track circuit.