DCTT统计最优化的进一步策略和方法

The Advanced Strategy and Method of DCTT Statistical Optimization

本文基于半无穷不可微最优化方法的框架和模型,提出了求解电路成品率极大的中心值设计,容差设计,调整设计及电路制造费用极小化为一体的统计最优化(DCTT)的求解方法及其策略。该方法不需要电路函数的凸性要求和构成可按受域的电路性能函数的半光滑假设。在引入抑制约束膨胀的策略后,该方法在确定性优化框架下可解决较大规模的统计最优化问题。为统计最优化的进一步发展开辟了一条新径。

Based on the frame and method of semi-infinite and nondifferentiable optimization technique, the advanced strategy and method of solving unitized model of circuit design centering for manufacturing yield maximization, tolerancing, tuning and cost minimization (DCTT) are proposed in this paper. When the method is used to solve DCTT problem, the circuit characteristics functions can be allowed to be nonconvex and nonsmooth. After the strategy of restraining constraint expansion is introducted, the method can solve larger scale statistical optimization problem. It opens up a new way for further development in the field of statistical optimization.