A Rule-Based Expert System for Predictive Maintenance of a Hybrid Bus
Loading...
Date
2020-06-19
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Université d'Ottawa / University of Ottawa
Abstract
This thesis describes the design and implementation of constructing a predictive
maintenance system for a hybrid vehicle to meet the requirements of the STO (Société
de transport de l’Outaouais). Thousands of sensors installed on the bus allow us to
observe the real-time performance of the bus while it is running. Abnormal sensor values
represent adverse operating conditions and bring attention to the inevitable failures
of a bus’s components. Therefore, by analyzing real-time sensor streams, predictive
maintenance is accomplished based on the unnatural behaviour of a hybrid bus.
Currently, transport companies still employ traditional methods of maintenance planning, such as emergency maintenance and preventive maintenance. Traditional maintenance strategies require a great deal of technicians and time to inspect the buses
regularly and carefully. In comparison, predictive maintenance can monitor the performance of buses based on the condition of their equipment. To collect data from the
hybrid bus and share data with the Internet, IoT technology is adopted to develop predictive maintenance architecture for a fleet management system. Our team devised an
IoT architecture for the fleet management system, including the perception layer, middleware layer and application layer. My work focuses on the perception layer, which is
responsible for analyzing sensor values, reporting failures of a hybrid bus and connecting
with cloud-servers.
As one of the predictive maintenance methods, the expert system (also known as a
knowledge-based expert system) is built to store expert knowledge in a specific area.
The expert system presented in this thesis can store failures of hybrid buses, symptoms
of which were provided to us by technicians from the STO. Such breakdowns assist
the expert system in predicting the malfunctions of the bus’s components based on the
symptoms. Inspired by the IDEA methodology, failure symptoms can be represented by
active rules with three essential components: event, condition and action. These rules
can also be translated into active database features like triggers and mapped into an
active database. A gateway is installed on a bus and composed of four modules: data
acquisition module, active rules module, rules management module and user interface
module. Within the parameters of the architecture and the gateway, this thesis analyzes
the entities, relationships and operations in the dynamic system and forms a relational
database to store the information related to the bus and active rules.
Description
Keywords
Predictive maintenance, Expert systems, Active rules, IoT