复杂事件处理的自适应制造情景识别方法

Adaptive context based situation identification based on complex event processing

制造过程中的任务、自然条件、电力水平等环境因素,制约物体状态及其关系的变化.智能制造单元需要自适应的对不同情境约束下的事件和复杂情形及时理解判断,提出基于复杂事件处理(Complex event processing,CEP)的情境约束情景识别方法,以实时作出合理的优化决策.针对忽视情境约束对事件判别的影响,构建基于情境约束的多层次事件模型,给出同生、情境、协同等事件新算子,提出基于事件聚合的制造情景模型与演算过程.针对情景识别知识库中模式规则生成的不足,通过整合物体数据与环境数据建立映射关联,将感知信息转化为情境事件图谱.通过综合序数、名义变量等距离计算和自适应熵权法,提出改进的混合聚类方法处理事件图谱实例属性的多样性和关联性,构建知识库以为情景实时识别提供服务支持.运用4个真实数据集和1个制造过程仿真数据集进行实验,均验证本文模型和方法的有效性,适用于大规模学习问题,并阐明情境因素能显著提升复杂制造应用中的事件判断、情景识别的准确性.

In the manufacturing process,environmental factors such as tasks,natural conditions,power levels,etc.,restrict the state change of objects and their relationships. Intelligent manufacturing units need to adaptively understand and judge events and complex situations under different context constraints,and a adaptive situation identification method based on complex event processing is proposed naturally,to make real-time optimization decision reasonable. In view of the phenomenon that the influence of context constraints on event discrimination is neglected,a context-aware hierarchical event model is built,and new operators of events such as contemporaneity,context and collaboration are given,while manufacturing situation model and aggregation process are proposed.Aiming at the shortage of generating method on situation model in knowledge base,the mapping association between an object and environment data is established firstly,and such sensed information is transformed into context based events. Integrating the distance calculation of ordinal,nominal variable and adaptive entropy weight method,a improved mixed clustering method is put forward to deal with the diversity and relevance of complex event instance attributes,providing service support for real-time situation identification. Finally,4 real data sets and 1 simulation data set of are employed for manufacturing process. Experiment results have verified the validity and adaption of the proposed model and method in large-scale problem,and expounded that context factors can significantly improve the accuracy of event judgment and situation recognition in complex manufacturing applications.