Research On Torsional Vibration Fatigue Damage Analysis Method For Turbine Generator Shafting

The torsional vibration problems of turbine generator sets can easily lead to large torsional alternating stress load on journals and couplings,which brings in the fatigue damage accumulation of the shaftings,Severe low-cycle fatigue can even lead to cracks and fractures in the shaft system.Therefore,the timely and effective detection of torsional vibration and the ability to quickly and accurately feedback fatigue life damage are critical to the safe production of electricity.Based on the general analysis method for fatigue damage impact of the turbine generator shafts torsional vibration and the recorded data of the units,a comparative study was conducted about the influencing factors for the evaluation results,including torsional vibration model,stress calculation method,load spectrum analysis method and other aspects.Firstly,the torsional vibration fatigue analysis based on finite element model and rapid modal decomposition theory was carried out.The continuous mass model of the shafting system is modeled in the finite element software based on the drawings of the manufacturer,and the stress load spectrum calculation of the shafting journal and generator coupling is completed based on the inherent characteristics of the shafting which extracted from modal calculation,and finally the fatigue life loss of four units are calculated.Afterwards,torsional vibration fatigue analysis based on lumped mass model and transient analysis was performed.Corresponding angular velocity and angular acceleration constrained signals are applied at the nose,and load spectrum of each dangerous journal is extracted to calculate the fatigue life loss of the shaft system.The results show that the transient analysis results is consistent with fast modal decomposition method for the load spectrum of most dangerous journal.However,the difference of journals near the nose is relatively large.The final calculated shafting fatigue life is relatively close.Finally,aiming at the problem of data storing and processing for the TSR devices,two different sampling methods are used to apply the non-parametric kernel density estimation algorithm to load spectrum statistics in torsional vibration accidents.Different kernel functions and bandwidth are applied to achieve the accurate estimation results and a small calculation cost.The results show that the selection of different kernel functions has very little effect on the accuracy of the estimation.However,the size of the bandwidth coefficient h has a great influence.The smaller h is,the more accurate the estimation is.Under the premise of achieving a good estimation accuracy,the use of peak sample method can effectively reduce the amount of data storage.