机械故障诊断的推理规律研究

机械故障诊断对于设备的安全、连续运行和预知维修至关重要。作为故障诊断成功的关键，文章阐述了诊断中的诊断知识及其结构，介绍了专家的概念知识、方法知识和面向诊断对象知识，并用一种双向、互补的有机结构来刻划这些知识在诊断过程中的综合应用。将故障熵的概念引申为广义故障熵，并结合最小互熵原理，用于阐述诊断推理过程。分别用对数模型和Sigmoid模型来刻划机械故障诊断中的认知规律。通过某炼油厂的一个实际诊断案例对这两个模型进行了分析和对比研究。结果表明，两者均能近似地刻划认知过程，但Sigmoid模型更能准确地描述机械故障诊断推理的一般规律，更符合诊断过程的实际情况。

Study of Inference Pattern in Fault Diagnosis

The fault diagnosis is essential to equipment safety, continuous operation and predictive maintenance. As a key to successful diagnosis, diagnostic knowledge and its structure are emphasized in the paper. Expert's conceptual knowledge, methodological and object-oriented knowledge are described. Then an organic, two-directional and mutually complementary structure is applied to describe their applications in fault diagnosis. Based on the principle of minimum cross-entropy, diagnostic entropy in broad sense is applied to characterize the fault diagnostic process. A logarithmic model and a sigmoid model are proposed to describe the cognitive process therein. Through a factual case study from the oil refinery, these two models are analyzed and studied contrastively. Results show that, both models can depict the cognition process approximately, but sigmoid model is better than the former. That is, sigmoid model is more exact and accords with practical diagnostic process better.