摘要
本文提出了一种基于前馈神经网络的焓差试验台故障预测模型,定义了查准率、查全率与F1系数作为模型的评判依据。基于空调焓差试验台的长期运行数据,建立前馈神经网络预测模型,并采取多种优化策略和方式改进模型。最终基于测试集数据进行性能测试,验证了该模型对焓差试验台故障预测的有效性。在一年多的试验工况下,该模型所预测的焓差试验台故障达到了93.33%的查准率与95.02%的查全率、以及94.14%的F1系数。对焓差试验台使用该模型,可有效预测台架故障,从而采取应对措施,减少试验台架故障导致的损失。
A fault prediction model based on feedforward neural network for enthalpy difference testbench is proposed. The precision, recall and F1 coefficient are defined as the basis for the model evaluation. Based on the long-term operation data of the air conditioning enthalpy difference testbench, the feedforward neural network model is built and improved in various optimization strategies and methods. Finally, the performance test is carried out on the test set to verify the effectiveness of the model for fault prediction of the enthalpy difference testbench. Under the experimental conditions for more than one year, the model predicts the faults of enthalpy difference testbench with the precision of 93.33%, the recall rate of 95.02% and F1 coefficient of 94.14%. Using this model for the enthalpy difference testbench can effectively predict the faults and take countermeasures to reduce the loss caused by the faults of enthalpy difference testbench.
作者
蔡博伟
陈江平
施骏业
CAI Bowei;CHEN Jiangping;SHI Junye(Institute of refrigeration and cryogenics,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai High Efficient Cooling System Research Center,Shanghai 200240,China)
出处
《制冷技术》
2018年第6期47-53,共7页
Chinese Journal of Refrigeration Technology
关键词
焓差试验台
故障预测
前馈神经网络
Enthalpy difference testbench
Fault prediction
Feedforward neural network
