摘要
针对一般模型算法在传感器相关性识别中存在的不足,提出一种基于FuzzyART神经网络的传感器相关性量化提取与识别方法,并与免疫网组合构成诊断系统。通过对某热控系统温度传感器故障的仿真诊断,验证了方法的有效性。仿真结果表明,系统能准确识别并诊断单传感故障和多传感故障。当传感器输出偏差大于±5%时,识别与诊断的准确率均达90%以上。
Aiming at the deficiency of sensor correlation identification of model algorithm, a new sensor correlation extraction and recognition algorithm is proposed based on Fuzzy ART neural network, of which the diagnosis system is consisted with immune network. By the simulation of temperature sensor fault in certain thermal control system, the method is valid. Simulation result shows that the system can recognize and diagnose the faults accurately, regardless ofs single or multiple sensor faults. The accuracy of recognition and diagnosis is above 90 percent when the sensor output is less than ±5 percent deviation.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第1期203-205,共3页
Computer Engineering
基金
黑龙江省教育厅科学技术研究基金资助项目(11511099)
黑龙江省自然科学基金资助项目(E200615)
