基于不完整描述样本的制冷系统故障诊断研究

Research and fault diagnosis of refrigeration system based on incompletely described samples

制冷系统故障信息样本中的部分特征缺失时,已有的故障诊断模型无法使用这些样本。为解决该问题,提出了一种新颖的故障诊断策略。该诊断策略可通过以下步骤实现:(1)在部分问题描述与历史数据库的相互作用及领域知识的基础上,产生了用于特征转换的相似转换矩阵;(2)将不完整描述样本中的未知特征转变为与其相关的已知特征,形成检索目标;(3)通过计算和比较检索目标与数据库中样本间的相似度,得到最佳案例,将该案例中的相应值赋予未知特征;(4)运用已有的故障诊断模型及再生的完整描述样本,对制冷系统进行故障诊断。利用实验数据对该诊断策略进行验证,取得了满意的结果。

The offered samples of refrigeration system can hardly be directly utilized by existing fault diagnosis （FD） methods when some features in the recorded samples are omitted. Hence, a novel strategy for FD was presented for the previous trained FD medel to effectively utilize the incompletely described samples. It was actualized in the following steps ： （ 1 ） with the help of domain knowledge, the similarity transformation matrix of partial problem description （PPD） -problems with incomplete feature description - was generated based on the historical database; （2） the unknown features of the samples were transformed to related known features, generating a new retrieval feature vector; （3） the values of unknown features were assigned by the optimal cases which can be retrieved by measuring and comparing similarities between the retrieval feature vector and the completely described samples in the historical database; （4） the regenerated completely described samples were utilized to diagnose based on existing FD models. Finally, the presented FD strategy was verified by experimental data and achieved satisfying results.