Declarative artifact-centric process models are suitable for specifying knowledge-intensive processes. Currently, such models need to be designed from scratch, even though existing model fragments could be reused to gain efficiency in designing and maintaining declarative artifact-centric process models. To address this problem, this paper proposes an approach for composing model fragments, abstracted into features, into fully specified declarative artifact-centric process models. We use Guard-Stage-Milestone (GSM) schemas as modeling language and let each feature denote a GSM schema fragment. The approach supports feature composition at different levels of granularity. Permutability of features is analyzed. Syntactic conditions that ensure permutability are defined and refactoring of non-permutable into permutable features is discussed. The approach has been evaluated by implementing it in a tool based on the Case Management Model Notation (CMMN) and applying it to three real-world processes. Using the approach, declarative artifact-centric process models can be composed from existing model fragments in an efficient, robust and correct way.

声明性的以工件为中心的过程模型适合于指定知识密集型过程。当前，即使可以重复使用现有的模型片段来提高设计和维护以声明为中心的以声明为中心的过程模型的效率，也需要从头开始设计此类模型。为了解决这个问题，本文提出了一种方法，用于将抽象为特征的模型片段组成为完全指定的以声明性工件为中心的过程模型。我们使用Guard-Stage-Milestone（GSM）模式作为建模语言，并让每个功能都表示一个GSM模式片段。该方法支持不同粒度级别的功能组合。分析特征的可置换性。定义了确保可置换性的语法条件，并讨论了不可置换成可置换特征的重构。通过在基于案例管理模型表示法（CMMN）的工具中实施该方法并将其应用于三个实际过程中，对该方法进行了评估。使用该方法，可以以有效，健壮和正确的方式从现有模型片段组成以声明为中心的以工件为中心的过程模型。