织物变形舒适性的双群粒子群聚类研究

Clustering study of fabric deformation comfort using bi-swarm PSO algorithm

针对传统聚类方法存在的一些问题,提出了一种具有个体交换策略的双群粒子群聚类方法,群1根据一个基于适应度的非线性公式动态更新惯性权重,群2采用固定权重,每进化一代,2群交换部分个体。将该方法应用于织物变形舒适性聚类中,以聚类中心作为粒子位置,通过粒子群优化算法获得最优聚类中心,采用最小距离准则对样本进行聚类。最后与模糊聚类做了简单比较,结果表明该方法结论合理,便于应用,为选择服装面料和评价织物性能提供了一种新手段。

Aiming at solving the problems of some traditional cluster methods,an improved cluster method based on bi-swarm particle swarm optimization（PSO） algorithm with exchanging particles strategy was proposed.One swarm dynamically updated its inertia weight by a new nonlinear updating formula,and the other employed a constant inertia weight.The two swarms exchanged some particles after each iteration.The method was applied to the cluster of fabric deformation comfort and took the cluster center as the position of the particle.The optimal cluster center was obtained by PSO algorithm optimizing and the sample data were clustered using the minimum distance criterion.The comparison results between fuzzy cluster results and PSO-based cluster shows that the proposed method can get the proper cluster result and provides a new approach to clothing fabrics selection and evaluation.