Structural Integrity And Damage Analysis Of Steam Turbine Rotor Under Flexible Operating Condition

Flexible operation is an inevitable operating mode for coal-fire power plant in an extended period.Frequent startups and shutdowns and varying degrees of deratings under the flexible operation,which cause creep-fatigue interaction,will bring great challenges to the integrity evaluation of its high temperature components such as steam turbine rotor.For the purpose of understanding the influence of flexible operation on rotor,and ensuring a safe operation,the mechanical responses and damage behavior of steam turbine rotors under flexible operating condition are investigated through numerical and experimental methods.A novel framework of physics-based neural networks is proposed,based on which the creep behavior of key locations under in-service condition is successfully predicted online.The main research work and conclusions are as follows:(1)The mechanical behavior of X12 Cr Mo WVNb N10-1-1 steel was simulated through a complex unified viscoplastic model.The accuracy of the simulation results are validated by multi-level tests including uniaxial tests,multiaxial tests and TMF tests.(2)The mechanical responses and damage behavior of steam turbine rotors under designed flexible operating conditions are studied based on a unified viscoplastic constitutive model.The results show that the accumulated viscoplastic strain changes the stress state under stable operation,and amplifies the cycling range in the transient operation.For the steam inlet region,the viscoplastic strain is mainly caused by thermal stress during startup and shutdown processes.While for the blade groove regions,the viscoplastic strain mainly accumulates during steady running.As the blade stage increases,the fatigue behavior becomes significant,but the total damage decrease.Due to more creep-fatigue interaction,the damaging rate is much higher in the high-temperature section than in the low-temperature section.(3)The mechanical properties and damage condition of an ex-service rotor are tested,and full-history flexible operation are simulated.The results show that the derated condition mainly affectes the stress concerntration notches in the high-temperature section,but only large derating causes fatigue damage.Startup and shutdown are the main causes of the plastic strain.The life consumption of the 10-year-serviced subcritical rotor is about 20% according to experiments.Compared with this value,the ASME and DIN EN codes both over-conservatively estimated the damage.Calculated damage values of EPRI code and CDM method are within ±2.5times of the experimental result.(4)A Model-guided Neural Network(MGNN)is established by embedding a phenomenological model into the feedforward neural network.The proposed MGNN model has been successfully applied to the on-line calculation of the stress relaxation in high-temperature blade root under inservice condition.In this process,it is found that LRe LU function is more suitable for the extrapolation of creep behavior than Re LU and Sigmoidtype activation functions.By using the structural characteristics of neural network,a distributed initial weight assignment method is proposed.The MGNN model outperforms FFNN and NARX network in the extrapolation of creep behavior,and it can track the stress change caused by the fluctuation of steam parameters under in-service conditions,which provides a new idea for the on-line strength evaluation of rotor under flexible operating conditions.