[202] Knowledge-based artificial intelligence based on the real-time diagnostics systems of equipment health monitoring

A V Kostyukov and S N Boychenko

DYNAMICS Scientific Production Center, Omsk 644043, Russia. 
Tel: 7 381 225 42 44; Fax: 7 381 225 43 72; Email: post@dynamics.ru 

An integral part of the modern processing industry is the implementation of technologies that dramatically reduce operating costs by utilising health monitoring systems for rotating and fixed equipment. The most effective way to reduce operating costs is to implement modern resource-saving technologies based on real-time systems that monitor the health of the equipment 24/7. The main ways to improve the efficiency of monitoring systems in the processing industry are listed below:
  • To apply the real-time diagnostics systems that enable continuous health monitoring for the most critical equipment;
To apply diagnostics systems at all stages of the equipment’s lifespan (manufacturing, commissioning, operation, maintenance and others);
  • To apply comprehensive diagnostics systems that provide health monitoring of various kinds of rotation and fixed equipment, utilising a single hardware and software platform; and
  • To integrate health monitoring systems into the plant’s diagnostic network, which allows unbiased information about machinery health to be presented to all interested parties.
The effectiveness of health monitoring systems directly depends on timely and accurate detection of machinery malfunction. Undoubtably, the most effective diagnostic systems utilise artificial intelligence. Currently, the number of publications describing the use of neural networks in equipment diagnostic systems is constantly rising. But are neural networks as effective, from a diagnostics perspective, as, for example, expert systems that utilise rules of degradation, which have been successfully used for many years? This report is devoted to aspects of developing diagnostics systems utilising artificial intelligence with neural networks. The paper compares systems based on the rules of degradation with a hybrid structure system that uses both approaches: rules of degradation (a knowledge base) and a neural network.