关键词: 设备健康管理, 趋势预测, 长短期记忆网络, 检修方案优化

Abstract: The main air blower set of catalytic unit is affected by operating conditions; the operation failure rate is relatively high; the duration of unit vibration degradation is long; and the level of Prognostics and Health Management (PHM) needs to be improved. Taking the maintenance of a petrochemical main air blower set as an example, this paper establishes a training prediction model based on the long short-term memory networks which collects and trains the data in the abnormal stage of unit vibration, predicts and fits the vibration trend of the unit. Then it puts forward optimization suggestions on the fault handling action and the timing of inspection and maintenance, and makes an attempt at predictive maintenance. These help torealize the improvement of unit health management.

Key words: equipment health management, trend forecast, long short-term memory network, maintenance scheme optimization