Qualitative modelling and simulation of physical systems for a diagnostic purpose
dc.contributor.author | Rozier, David | |
dc.date.accessioned | 2020-03-04T08:45:53Z | |
dc.date.available | 2020-03-04T08:45:53Z | |
dc.date.issued | 1998-09 | |
dc.description | This is a Milton Keynes De Montfort University thesis | en |
dc.description.abstract | The goal of a fault-diagnosis system is to obtain an accurate diagnosis at a low cost. In order to reach this goal, many techniques have been used, e.g. qualitative methods and multiple-models. This research investigates a novel strategy for improving the balance accuracy versus cost of consistency-based fault-diagnosis systems. This new strategy is organised around the notion of entities. These are physical entities. such as water pressure or temperature. The functioning of a physical system can involve numerous entities. Because these entities influence each other's behaviour, multiple-fault situations can occur, where several entities are affected by a fault. These situations are called complex multiple-fault situations. The existing fault-diagnosis systems do not perform satisfactorily on complex multiple-fault situations. This is because the set of entities they investigate is fixed from the start of the diagnostic process. As a consequence, depending on the entities included in this set, existing systems either perform an inaccurate diagnosis, or reach an accurate diagnosis at an unnecessarily high cost. This thesis presents a fault-diagnosis strategy called MVDS (standing for Multiple Variable Diagnosis Strategy) designed specifically for performing the efficient diagnosis of complex multiple-fault situations. The underlying principle of MVDS is that it is not possible to know from the start of the diagnostic process which entities are affected. Thus, a diagnostic process with MVDS is undertaken with the investigation of an initial set of entities, and this set of investigated entities is continuously updated along the process, as intermediate results are obtained. In order to illustrate clearly the functioning of MVDS, a fault-scenario using a small example from the air-conditioning domain is diagnosed and the process studied. The investigation of the performance of MVDS on more complex physical systems is undertaken on a larger case-study using a hot-water and heating system. In MVDS, it is possible to disable the adaptability of the set of investigated entities, so that it can be run with a fixed set. By doing so, the performance of the strategy in MVDS can be compared to the performance of traditional approaches which use a fixed set of investigated entities. The study-case shows that MVDS reaches more accurate results than traditional approaches, and that this accuracy is obtained at a low cost, since unnecessary measurements of entities are avoided. Furthermore, the strategy produces a complete trace of the process that is close to common-sense reasoning. It is also a co-operative strategy where the operator can intervene. Summary of the main research contributions: - The issue of diagnosing complex multiple-fault situations is specifically addressed for the first time. The problem caused by this diagnosis task is defined, and a strategy is constructed in order to diagnose efficiently the complex multiple-fault situations. The strategy is implemented in MVDS and tested on an example and a case-study. - Risk characteristics have been described. They allow to evaluate how prone to complex muItiple-fault situations is a physical system. - Hot-water and heating systems are offered as a new domain of research for consistency-based fault-diagnosis systems. - The inclusion of co-operation into the fault-diagnosis process is a novel approach. Its potential advantages have been identified. | en |
dc.identifier.uri | https://dora.dmu.ac.uk/handle/2086/19282 | |
dc.language.iso | en | en |
dc.publisher | De Montfort University | en |
dc.publisher.department | Department of Computer and Information Sciences | en |
dc.title | Qualitative modelling and simulation of physical systems for a diagnostic purpose | en |
dc.type | Thesis or dissertation | en |
dc.type.qualificationlevel | Doctoral | en |
dc.type.qualificationname | PhD | en |