噪声与振动控制
譟聲與振動控製
조성여진동공제
NOISE AND VIBRATION CONTROL
2015年
2期
165-169,174
,共6页
马继昌%司景萍%牛嘉骅%王二毛
馬繼昌%司景萍%牛嘉驊%王二毛
마계창%사경평%우가화%왕이모
振动与波%小波分析%模糊理论%BP神经网络%故障诊断
振動與波%小波分析%模糊理論%BP神經網絡%故障診斷
진동여파%소파분석%모호이론%BP신경망락%고장진단
vibration and wave%wavelet analysis%fuzzy theory%BP neural network%fault diagnosis
发动机是车辆的核心部件，及时有效地发现并排除故障，对降低维修费用，减少经济损失，增加发动机工作时的可靠性，避免事故发生具有重大的意义。以某型号发动机为研究对象，运用测试技术、信号处理、小波分析、神经网络和模糊控制理论，提出了自适应模糊神经网络发动机故障诊断。首先建立了发动机故障信号采集试验台，在试验台上人工模拟四种工况，通过加速度传感器采集正常工况和异常工况的振动信号。再利用小波理论对采集到的振动信号进行消噪处理，提高信噪比，并提取出故障信号的特征值，作为网络训练和测试的样本数据。用样本数据训练和检测自适应模糊神经网络，即对发动机故障进行模式识别。通过仿真分析，取得了很好的诊断效果；同时与传统的BP神经网络故障诊断方法进行对比，无论在诊断精度上还是学习速度上，模糊神经网络在故障诊断中更具有优势。
髮動機是車輛的覈心部件，及時有效地髮現併排除故障，對降低維脩費用，減少經濟損失，增加髮動機工作時的可靠性，避免事故髮生具有重大的意義。以某型號髮動機為研究對象，運用測試技術、信號處理、小波分析、神經網絡和模糊控製理論，提齣瞭自適應模糊神經網絡髮動機故障診斷。首先建立瞭髮動機故障信號採集試驗檯，在試驗檯上人工模擬四種工況，通過加速度傳感器採集正常工況和異常工況的振動信號。再利用小波理論對採集到的振動信號進行消譟處理，提高信譟比，併提取齣故障信號的特徵值，作為網絡訓練和測試的樣本數據。用樣本數據訓練和檢測自適應模糊神經網絡，即對髮動機故障進行模式識彆。通過倣真分析，取得瞭很好的診斷效果；同時與傳統的BP神經網絡故障診斷方法進行對比，無論在診斷精度上還是學習速度上，模糊神經網絡在故障診斷中更具有優勢。
발동궤시차량적핵심부건，급시유효지발현병배제고장，대강저유수비용，감소경제손실，증가발동궤공작시적가고성，피면사고발생구유중대적의의。이모형호발동궤위연구대상，운용측시기술、신호처리、소파분석、신경망락화모호공제이론，제출료자괄응모호신경망락발동궤고장진단。수선건립료발동궤고장신호채집시험태，재시험태상인공모의사충공황，통과가속도전감기채집정상공황화이상공황적진동신호。재이용소파이론대채집도적진동신호진행소조처리，제고신조비，병제취출고장신호적특정치，작위망락훈련화측시적양본수거。용양본수거훈련화검측자괄응모호신경망락，즉대발동궤고장진행모식식별。통과방진분석，취득료흔호적진단효과；동시여전통적BP신경망락고장진단방법진행대비，무론재진단정도상환시학습속도상，모호신경망락재고장진단중경구유우세。
Engine is a very important part of the vehicle. Timely recognizing and suppressing the engine faults have im-portant significance for reducing maintenance costs and economic loss, raising the reliability of engine operation and avoid-ing accidents. In this paper, using measurement technique, signal processing, wavelet analysis, neural network and fuzzy con-trol theory, an engine fault diagnosis method was proposed based on adaptive fuzzy neural network (AFNN) algorithm. A test bench was established for fault signal acquisition of the engine. Four kinds of artificial conditions were simulated on the test bench, and the vibration signals in the normal and abnormal operation conditions were collected through the acceleration sensors. Then, using wavelet theory, de-noising process was done for the collected vibration signals to raise the signal-to-noise ratio and extract the characteristic values of the fault signals as the network training sample data and testing sample da-ta. Finally, the sample data was used for training and testing the adaptive fuzzy neural network to recognize the engine fail-ure. Good diagnosis results were obtained through the simulation. Compared with the traditional BP Neural Network diagno-sis methods, the fuzzy Neural Network has more advantages in fault diagnosis no matter in learning speed or accuracy of the diagnosis.
