一种新型可靠性模型参数优化求解方法

A New Type of Solving Method for Reliability Model Parameter Optimization

对可靠性增长模型参数进行求解多采用构造极大似然函数,并对似然函数求极值的方法.用极大似然法进行参数优化估计时,有容易受迭代初值的影响不易收敛到全局最优解的缺点,文中采用进化规划(EP)算法,建立以适应函数为目标,求其极大值点即可确定参数最优解的优化模型,不再需要求极值和估计优化变量的初始值即可获得全局近似最优解.为了更好地确保获得全局最优解,进一步保证方程解的精度,进化规划算法采用了并行操作、保留最优个体等方法.新的优化参数求解方法可以在求解效率和收敛性能上达到较好的平衡,能更好地将优化方法与最大似然估计法相结合.最后利用某固体火箭发动机的可靠性增长实验数据验证了该优化方法的有效性和正确性.

Most of solving method for reliability growth model parameters is to structnre maximum likelihood function and extremum on the likelihood function. When using the maximum likelihood method for parameter optimization, it is not easy to converge to the global optimal solution by the effect of iterative initial value, evolutionary programming （EP） algorithm is used in this paper, establish the fitness function, find great value point to determine optimization model of the parameters optimal solution, and no longer need to require extreme value and estimate initial value of the optimization variable, global approximate optimal solution can be obtained. In order to better ensure global optimal so- lution, to further ensure the accuracy of the equations, a parallel operation and retain the best individual and other methods are used in evolutionary programming algorithm. The new method for solving the optimization parameters can achieve a better balance in the solving efficiency and convergence performance and better combined optimization meth- ods with maximum likelihood estimation method. The validity and correctness of the optimization method is verified in the last of the paper through the experimental data in the reliability growth process of a solid rocket motor.