摘要
针对智能大厦空调系统中发生的各种传感器故障问题 ,提出了一种基于概率神经网络 (PNN)的传感器故障诊断改进方法。该方法采用贝叶斯分类决策理论建立系统的数学模型 ,以高斯函数作为激励函数 ,具有非线性处理和抗干扰能力强等特点 ,可获得对空调系统中各种传感器硬故障和软故障的有效识别和诊断。给出了该方法的理论分析 ,故障特征量的选取 ,神经网络设置和训练的具体步骤。通过仿真和空调系统模型试验证明了该方法在网络训练速度 ,抗干扰能力及各种传感器故障识别准确率等方面的有效性。
Aiming at the problems of sensor faults happening in air conditioning system of intelligent building, an improved method of sensor fault diagnosis based on probabilistic neural networks (PNN) is presented. The method uses Bayes classifying and decision-making theory to constitute the mataematic model of system, with Gauss function as activating one it possesses the characteristics of strong nonlinear processing and anti-interfering ability. It can effciently identify and diagnose the hard and soft faults of sensors in air conditioning system. Theoretical analysis, choice of fault characteristics and practical procedure of neural network setting and training are given out. Effectiveness of the method in networks training speed, anti-interfering ability and identifying veracity of sensor faults are proved by simulation and model test of conditioning system.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2004年第7期997-999,共3页
Systems Engineering and Electronics
关键词
系统工程
故障诊断
概率神经网络
传感器故障
sestem engineering
fault diagnosis
probabilistic neural networks
sensor fault
