【摘要】 基于传统谱聚类算法的产品模块划分方法主要依赖局部搜索策略，缺乏全局优化的能力，特别是在面对复杂且非凸优化问题时，传统谱聚类算法无法收敛至全局最优解，不适用于零部件间耦合关系复杂的通航装备产品。鉴于此，提出了一种改进的谱聚类方法。首先，通过引入航空器产品零部件间的功能、结构和物理相关特性，构建出一种基于数字化结构矩阵的产品零部件关联关系数学模型。其次，在该模型的基础上，通过引入蝙蝠算法代替传统k-means聚类算法来进行产品的聚类划分，利用蝙蝠算法全局搜索能力强的优点克服传统谱聚类算法容易陷入局部最优解的问题。最后，以某通航飞机的环控散热器总装夹具为例，应用改进的谱聚类优化算法对产品进行聚类划分。实验结果表明，该方法在降低实验复杂性、提高产品聚类划分精确性等方面效果显著。更多还原

【Abstract】 Traditional module partition methods based on spectral clustering algorithms primarily rely on local search strategies,leading to the lack of global optimization capabilities.Especially when used to address the complex and non-convex optimization problems,the traditional spectral clustering algorithm fails to converge to the global optimum,so it is unsuitable for general aviation products with complex inter-component couplings.In view of this,this paper proposes an improved spectral clustering method.Firstly,a mathematical model of component correlation based on digital structure matrix is constructed,incorporating functional,structural,and physical correlations among aircraft product components.Secondly, based on this model,the bat algorithm is introduced to replace the k-means algorithm used in traditional spectral clustering for product clustering partition,leveraging its strong global search ability to overcome the limitation of the traditional spectral clustering algorithms which tend to fall into local optima.Finally,the proposed method is applied to cluster and partition the environmental control radiator fixture of a general aviation aircraft as a case study.Experimental results demonstrate that the proposed method achieved significant effects in reducing experimental complexity and improving the accuracy of product clustering partition.更多还原