Data-driven valve diagnosis to increase the overall equipment effectiveness in process industry

Abstract

The avoidance of plant shutdowns is one of the highest priorities for plant operators (plant owners). Shutdowns are forced by abnormal situations, e.g. unexpected equipment faults such as valve or pump faults. Each unexpected fault can lead to hazardous situations within a plant. Pumps are already well analyzed compared to valves and also frequently used in process industry. In this paper a data-driven fault detection system for valves will be introduced. To gain additional knowledge about faults of specific equipment, big data technology is applied, based on a huge number of historical data for different valves. The paper introduces an approach in which data from different competitive companies operating several process plants are filtered, selected and combined with data from equipment manufacturers. The valve diagnosis system uses historical process data obtained across company borders using physical valve models to detect faults by comparing standardized flow coefficient determined by DIN IEC 60534-2-1.