HYBRID EARLY WARNING SYSTEMS

ABSTRACT New tools are needed to reach high goals for uptime and availability in industrial processes. Early warning of developing faults is one part of the strategy to reach these goals. A single method rarely meets all requirements, but combining methods and techniques in a hybrid system offers advantages and can overcome limitations in the individual approaches. Methods considered are physical models, artificial neural networks, and case-based reasoning. The paper discusses the pros and cons, strengths and weaknesses of the three methods and three combinations of hybrid solutions in order to assist in select a suitable combination for a specific early warning challenge ahead