CAJ | 학술논문

为避免故障诊断中单一信息的固有缺陷,提高诊断精度,将改进的BP神经网络和D-S证据理论相结合,针对齿轮箱的3个测点,采用改进BP神经网络进行独立的局部故障诊断,以及采用D-S证据理论规则,将3个测点的神经网络输出结果进行融合,得到整个齿轮箱的故障诊断结果。通过分析可以看出,融合后的诊断结果更加精确,信任度更高。
위피면고장진단중단일신식적고유결함,제고진단정도,장개진적BP신경망락화D-S증거이론상결합,침대치륜상적3개측점,채용개진BP신경망락진행독립적국부고장진단,이급채용D-S증거이론규칙,장3개측점적신경망락수출결과진행융합,득도정개치륜상적고장진단결과。통과분석가이간출,융합후적진단결과경가정학,신임도경고。
In order to avoid inherent defect of single information in the fault diagnosis, and to improve the diagnostic accuracy, the improved BP neural network is combined with the D-S evidence theory. For the three gauge points of gear-box, the improved BP neural network is used for independent and local fault diagnosis, and then the neural network output results of them is fused by D-S evidence theory rules, and the whole gear-box fault diagnosis is obtained. Through the analysis, one can see that the fused diagnosis result is more accurate, and the trust is higher.