膜计算是计算智能和生物计算领域的新分支。P系统一般以极大并行模式运行，具备与图灵机等价的计算能力。设计新的膜计算模型，扩展其应用领域是膜计算研究的重要问题。本项目提出了一种基于链式结构的新型膜计算模型，该模型使用链式结构来定于膜结构和对象，然后将新的膜计算模型应用于聚类分析问题。主要研究内容包括：构造链式结构膜系统，定义链式结构膜对象，设计与此相对应的规则；对新的膜系统进行理论分析，证明其图灵等价性；将部分传统P系统统一在链式膜系统框架之下；研究基于链式膜系统的直接/耦合膜算法，将聚类中的组合优化及连续优化运算采用链式膜系统描述和实现，提出相应的直接/耦合膜算法；设计基于链式膜系统的聚类算法，研究基于ROCK算法框架的链式聚类算法。本研究有望将传统膜计算模型进行扩展，通过将聚类数据及参数转化为链式膜系统可识别形式把膜计算与聚类分析进行进一步的结合，进而改进聚类算法。

Membrane computing is a new branch in computational intelligence and biology based computing. P systems normally run in maximal parallelism, with computational power equivalent to Turing machines. Two important problems in the field of membrane computing are the design of new membrane computing models together with extension of their application domains. In this project, a novel membrane computing model called the chain membrane model is proposed. In the new model, membrane structures, symbols and multisets are expressed in chain structures. The new membrane computing models are applied to cluster analysis. Research contents of this project are mainly as follows. First a well-defined chain membrane system is established. Then we use chain structures to define membrane architectures, symbols and multisets, as well as the description of corresponding rules. The new P systems are analyzed to show its Turing equivalency. Some traditional membrane systems can be unified under the frame of the chain membrane systems. Direct and coupling membrane algorithms are also investigated based on chain membrane systems. Essential computing operations of cluster analysis with combinatorial optimization as well as continuous optimization natures are described in chain P systems with implementation. Then we present some underlying direct and coupling membrane algorithms. Clustering algorithms are proposed based on chain membrane systems together with a chain clustering technique based on the ROCK algorithm. The research of this project may extend traditional membrane models. Further connections between membrane computing and cluster analysis can be built via transformation of clustering data with parameters into recognizable form by P systems. Finally this may develop clustering algorithms.