基于模式识别的合成氨统计过程控制

Statistical Process Control for Ammonia Synthesis Based on Pattern Recognition

合成氨生产过程因影响因素多且关系复杂,容易引发各类异常。利用模式识别方法,依据操作类和目标类之间的对应关系,提出“越类量”概念,并给出计算方法。对“越类量”利用统计过程控制方法进行跟踪,在线检测过程异常,在一套年产8×104t的氨合成系统实际投运中取得了满意的效果。

The concept of ＂slop-over value＂ is proposed based on pattern recognition clustering method. A method to calculate slop-over value is presented according to the relationship between operation class and objective class. The exception condition is detected on line by tracking slop-over value with statistical process control method. The proposed method has been applied in the exception detection of an ammonia synthesizer （80,000 tons per year） with satisfactory results.