Fault detection and diagnosis for condition based maintenance using the Logical Analysis of data

Presently, most maintenance decisions are mainly based on failure event. Nevertheless, in many cases and for many types of equipment this event can rarely be seen or can happen after many years of utilization. In these cases, maintenance decisions are based on fault diagnostics. This paper presents an artificial intelligent approach to fault detection and diagnosis. This approach is called Logical Analysis of data, and it is based on a combinatorics, Boolean, and optimization theory. Its main power stems from the fact that it detects logical patterns that can be easily interpreted. These patterns are used in the classification of observations collected by condition monitoring, and thus in the diagnosis of faults. An application is presented. Analysis of the results obtained shows high classification accuracy and useful features for detection and analysis.