6633

6183

كارشناسي ارشد

طراحي كاربردي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده مكانيك

1390

[هشت]، 83ص.: مصور، جدول، نمودار

ص.ع. به فارسي و انگليسي

سعيد بهبهاني

سعيد ضيايي راد

9/2/91

دانش، مقيمي زند

1396/10/12

كتابنامه

مكانيك

مهندسي مكانيك

ID6183

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Condition monitoring of a turbo machine based on vibration analysis using artificial neural network theory Sajjad Salmanzadeh s salmanzadeh@me iut ac ir Department of Mechanic Engineering Isfahan University of Technology Degree M Sc Language FarsiSupervisor Saeed behbahani behbahani@cc iut ac irAbstractCondition monitoring in the turbo machinery system is controlling parameters such as vibration domain andvibration frequency for fault diagnosis in the system Condition monitoring is the main part of PredictiveMaintenance Rotors used in industrial machinery such as compressors pumps or turbines are oftensubjected to extreme loading during their operation Rotating parts in machines exposed to external forcesand temperatures may lead to a fatigue crack resulting in rotor damage Structural health monitoring of therotors is important for improving the safety of their operation and for extending their service life All theusual faults like unbalance misalignment etc encountered in the rotor systems the fatigue crack is the mostdangerous one because if left undetected it can lead to catastrophic failure This project describes the application of Wavelet Transform WT fuzzy logic and artificial Neural Network ANN for prediction of the faults effect on the frequency components of vibration signature in a shaft Thesefaults are unbalance crack and combined faults of unbalance and crack Existence of these faults in the shaftcan lead to increase the damage in the system if these faults not detected can increase the time and costs ofrepair For studding these faults finite element solution and experimental test is used and vibration signals ofshaft saved for 2 seconds Continuous wavelet transform and scaled averaged wavelet power method areused for signal processing Then by conducting principal component analysis at these coefficients they havebeen used as input of artificial neural network Also total scaled averaged wavelet power is used to identifysystem condition The developed ANN is constructed of a hidden layer with 6 neurons and an output layerwith 4 neurons The network is trained with 40 sets of data relating to faulty shaft In order to test thenetwork 30 sets of data relating to three fault states unbalancing crack and combined faults obtained fromnumerical and experimental tests are used The results show that the well developed network has been able todetect system faults with the accuracy of 96 6 Keywords Condition monitoring Wavelet Fuzzy logic Neural network

سعيد بهبهاني

سعيد ضيايي راد

دانش، مقيمي زند