Year 2000
Marko Grönbärj

Inference Methods in Industrial Fault Diagnostic

The aim of this thesis was to develop an inference engine for a fault diagnosis system.

Fault diagnostics has recently been a subject of great interest and development. Knowledge can be transferred from process engineers to the fault diagnosis system by reasoning methods. With an expert system, faults can be diagnosed more accurately and maintenance is easy.

In the literature part, the basic structure and functions of a fault diagnosis system were introduced. Also, different rule-based inference methods were studied with examples from industrial applications.

In the experimental part, the structure of the fault diagnosis system was designed. The structure consists of a link between a database and the process, databases, a Kohonen map application, a symptom generator module and an inference engine. In this thesis, the inference engine, database link and symptom generator were implemented.

The inference engine developed was tested with data collected from the steam dryer in the copper smelting process of Harjavalta Metals, and the knowledge of the process engineers was used. During testing, the system responded correctly to faults and normal conditions.

The rule-based inference used in the system succeeded in diagnosing faults and it is applicable to a structurally changing process.

This info last modified 27 Apr 2024 - contact webmaster