论文部分内容阅读
针对液体火箭发动机推进系统超高维故障样本数据的聚类问题 ,提出基于演化策略的最优统计聚类算法。为预防算法过早收敛 ,演化策略采用了父本适应值的动态调整值与共享函数 ,并针对超高维数据聚类提出了控制参数的适应性调整技术 ;为使算法能最终跳出局部最优死区 ,提出算法的局部调整策略。该算法用于液体火箭发动机典型故障仿真数据集分析 ,并取得了最优聚类结果。此外 ,还基于IRIS数据集比较了该算法与FKCN模糊自主聚类算法。仿真分析表明了算法在高维数据聚类分析中的优点。
Aiming at the clustering problem of ultra-high-dimensional fault samples in liquid propellant rocket propulsion system, an optimal statistical clustering algorithm based on evolution strategy is proposed. In order to prevent the algorithm from premature convergence, the evolution strategy adopts the dynamic adjustment value and the shared function of the paternal fitness value, and puts forward the adaptive adjustment technique of the control parameters for the ultra-high dimensional data clustering. In order to make the algorithm eventually jump out of the local optimum Dead zone, proposed local adjustment strategy algorithm. The algorithm is used to analyze the typical fault simulation data of liquid propellant rocket engine, and the optimal clustering result is obtained. In addition, this algorithm is compared with FKCN fuzzy autonomous clustering algorithm based on IRIS dataset. Simulation analysis shows the advantages of the algorithm in high-dimensional data clustering analysis.