摘要
研究发动机部件性能参数变化规律,对于减少维修次数和推动视情维修具有重要意义。针对测量参数个数少于待估性能参数的情况,给出了一种通过构建代价函数和优化算法的参数估计方法。原代价函数只考虑当前点参数,缺少与前面点参数的联系,因此结合自组织神经网络,构造了包含以前与当前点参数的距离代价函数。并提出了一种快速的参数估计方法。由于准确的部件性能参数很难获取,并且参数趋势估计不同于单纯的点估计问题,以对应的测量参数为基础,利用信息熵方法评定部件性能参数估计效果。进一步得到距离代价函数对应的参数信息熵为0.6805,优于原代价函数的估计结果。最后通过实例验证了参数估计方法的有效性。
It has profound meaning for reducing maintenance times and promoting condition-based ma- intenance to study the change of component performance parameter. Aiming at the number of measured pa- rameter less than that of pre-predicting performance parameter, a parameter estimation method by establis- hing cost function and optimization algorithm was showed. Current parameter being only considered in pri- mary cost function, lacking of relationship of former parameter, therefore distance cost function of former and current parameter was established combined with self Organization Map （ SOM）. Fast parameter esti- mated method was presented. Exact component performance parameter is difficult to obtain, and estimation of parameter trend is different from that of pure point. Information entropy of measured parameter is adopted as evaluating parameter estimation. Furtherly, information entropy corresponding to distance cost function is 0. 6805, superior to estimation of former cost function. Finally, it verified effectiveness of parameter estima- tion method presented.
出处
《推进技术》
EI
CAS
CSCD
北大核心
2013年第11期1557-1566,共10页
Journal of Propulsion Technology
关键词
代价函数
信息熵
性能衰退
自组织映射
参数优化
Cost function
Information entropy
Performance deterioration
SOM （ Self OrganizationMap）
Parameter optimization
