提出了解约束优化问题的一类相容SQP滤子算法.利用序列二次规划方法结合信赖域技术计算试探步,而用滤子接受准则选择接受试探步.对二次规划子问题的不相容问题,应用Powell1978年于文[9]提出的方法对其约束引进参数进行了可行化处理.在一般条件下,算法具有全局收敛性.最后,数值试验显示了较好的结果.
A compatible SQP algorithm combined with filter method is proposed for constraint optimization problem.This iterative mechanism is based on SQP method combined with trust region technique.And,filter method is used as an acceptable criterion for the trial step.For the incompatibility of QP subproblem,introduction of parameter make the solved problem always compatible by using the method proposed first by Powell[9] in 1978,so a feasibility restoration phase is not needed any more.Under mild assumptions the algorithm convergent globally.Some numerical tests show the efficiency of our method.