有色观测噪声下的多变量设置调整策略

针对存在初始设置偏差的离散的多变量生产制造过程，研究了观测噪声服从Auto-Rgressive （AR）模型的情况下，考虑调整花费成本为二次型函数时的设置调整问题，在建立过程状态空间方程的基础上，利用卡尔曼滤波方法在线估计过程的状态变量，根据随机二次型最优控制理论，得到了使过程质量损失最小的最优调整策略. 通过算例解释了最优调整策略的实现方法，并进行了仿真验证，结果表明，得到的调整策略与观测噪声为白噪声时的质量调整策略相比，能更好地减少过程总体质量损失.

Setup adjustment scheme for multivariate process considering colored observation noises

For the finite horizon multivariate process with setup error, the optimal adjustment scheme is developed to minimize the total process quality loss with quadratic cost and AutoRegressive observation noises. Based on the state-space process control model, the optimal adjustment scheme is derived by Kalman filter on line estimation and linear quadratic Gaussian (LQG) theory. A simulation case is presented to illustrate the implementation method of the optimal adjustment policy. The optimal adjustment scheme is compared with quality adjustment policy with white noise observation noises by simulations. The results show that the proposed adjustment solution is more effective than other to reduce the total quality loss of the process.