基于CART算法的多联机压缩机回液故障检测

Fault detection for compressor liquid return of multi-connected air-condition(heat pump) unit based on CART algorithm

近年来,多联机在各类建筑上得到广泛应用。一旦压缩机发生回液故障,将会导致多联机偏离正常工况,能效降低,同时无法保障室内环境的舒适性。长久的回液故障还会导致压缩机部件的机械故障,造成高额的维修费用。因此,本文提出决策树模型检测多联机压缩机回液故障。首先,通过数据集成和剔除缺失值得到齐整数据;其次,通过专家知识和线性相关分析分别选取变量和剔除冗余变量;最后,建立决策树模型检测回液故障。结果表明,决策树模型能够有效地检测出回液故障,而且与专业知识吻合得较好。

Nowadays,multi-connected air-condition（heat pump）unit have been widely used in various buildings.Liquid return makes compressors work under abnormal operation,which results in low energy efficiency and poor comfort for indoor environment.And liquid return for a long term can cause the mechanical failures,also contributes to a high maintenance cost.Thus,the CART（classification and regression tree）model is proposed to detect compressor liquid return fault.Firstly,data integration and missing value remove are employed to get tidy data.Secondly,domain knowledge and linear correlation analyses are adopted to select input variables for CART model and remove redundant variables respectively.Finally,CART model is set up to detect liquid return fault.The results show that the CART models can achieve desirable performance in the liquid return fault detection,and it coincides with domain knowledge better.