In the process of knowledge reasoning applied to the intelligent method,in order to solve the problem of concurrency matching of knowledge and usability,a method of uncertainty reasoning under multi-level framework is designed. With introduction of rule-knowl-edge derivation,establishment of the production model and design of multi-level frame structure,its precision and accuracy of reasoning is improved effectively;introducing slot value coincidence rate and membership degree as confirmation factor as well as knowledge con-currency matching,the forward-based reasoning is adopted as knowledge reasoning and the Bayes method applied to eliminate conflict in knowledge matching. Experiments on its accuracy and concurrent reasoning have been conducted with some concurrent abnormal sam-ples,which show that its accuracy and concurrent reasoning is promoted greatly,achieving the purpose for reasonable design of knowledge structure and significant improvement in knowledge abstraction,meeting the refinement requirements of complex industrial production in the field of multi concurrent reasoning. The analysis on results of practical testing early dynamic warning for oil well production show that the proposed UnMF method has improved accuracy of early warning,which can reduce the occurrence of abnormal production and there-fore is of great significance for stable and efficient industrial production.%在知识推理应用于智能化方法的过程中,为解决知识并发性匹配及可用性等问题,设计了一种多级框架下的不确定性推理方法.该方法通过采用规则-知识衍生方式、构建产生式模型和设计多级框架结构以有效提高推理精度和准确率,通过引入槽值符合率及隶属度作为确信因子以及实现知识的并发性匹配,选定正向推理作为知识推理方式,应用论据累计的Bayes方法消解知识匹配冲突.选用并发异常样本对所提出方法的准确率和并发推理能力进行验证实验.实验结果表明,该算法在准确率和并发推理能力方面具有较大提升,达到了知识结构设计合理、知识抽象化程度明显提高的目的,满足了复杂工业生产等领域对多并发推理的精细化要求.油井生产动态预警应用的实际测试分析表明,所提出的UnMF方法较大程度上提高了预警的准确性,降低了生产异常情况的发生,对稳定高效工业生产具有重要的意义.
展开▼
