IEEE Trans Cybern. 2015 Jul;45(7):1236-49. doi: 10.1109/TCYB.2014.2347801. Epub 2014 Sep 4.
Most model-based approaches to fault diagnosis of discrete-event systems require a complete and accurate model of the system to be diagnosed. However, the discrete-event model may have arisen from abstraction and simplification of a continuous time system, or through model building from input-output data. As such, it may not capture the dynamic behavior of the system completely. In a previous paper, we addressed the problem of diagnosing faults given an incomplete model of the discrete-event system. We presented the learning diagnoser which not only diagnoses faults, but also attempts to learn missing model information through parsimonious hypothesis generation. In this paper, we study the properties of learnability and diagnosability. Learnability deals with the issue of whether the missing model information can be learned, while diagnosability corresponds to the ability to detect and isolate a fault after it has occurred. We provide conditions under which the learning diagnoser can learn missing model information. We define the notions of weak and strong diagnosability and also give conditions under which they hold.
大多数基于模型的离散事件系统故障诊断方法都需要一个完整和准确的系统模型来进行诊断。然而，离散事件模型可能是通过对连续时间系统的抽象和简化，或者通过输入-输出数据的建模而产生的。因此，它可能无法完全捕捉系统的动态行为。在之前的一篇论文中，我们解决了在离散事件系统的不完整模型下诊断故障的问题。我们提出了学习诊断器，它不仅可以诊断故障，还可以通过简约的假设生成尝试学习缺失的模型信息。在本文中，我们研究了可学习性和可诊断性的性质。可学习性涉及到是否可以学习缺失的模型信息的问题，而可诊断性则对应于在故障发生后检测和隔离故障的能力。我们提供了学习诊断器可以学习缺失模型信息的条件。我们定义了弱可诊断性和强可诊断性的概念，并给出了它们成立的条件。
