Use of Abstractions within Fault Diagnosis of Shipssystems

Abstract

This paper describes a fault diagnosis system developed for ships. The system is designed to support the operator in dealing with malfunctions in a ships machinery installation. The scope of the diagnosis system is to diagnose malfunctions in the ships auxiliary and ancillary systems. Processes internal to the main engine are not within the scope. The main engine is thus handled as a single component. The purpose of the diagnosis system is not only to detect malfunctions in terms of possibly faulty components, but also to assist in determining consequences of the detected malfunctions and to suggest appropriate actions according to the situation. The diagnosis system is based on an intentional model of the ships machinery installation. The model is described in a language called Multilevel Flow Modelling (MFM) which support derivation of statements about failure to achieve system intentions. MFM describes causal and continuity relationships in the physical process in terms of flow of mass and energy. Reasoning within the model is based on qualitative values given in terms of normality. These qualitative values are derived by means of a filter which continuously calculates RMS values of measurements. The system is implemented in C++, executes under the Unix operating system, connects to a conventional alarm system and has an X-Windows based user interface.