| As the core equipment of the power supply vehicle,diesel generator is carried out reasonable health management,which not only can improve the reliability and safety of the operation of the power supply vehicle,but also facilitate the scientific and rational development of the operation and maintenance of the power supply vehicle.Health management of diesel generator can reduce economic losses to users that is brought by blind maintenance,and make up for the lack of health management strategies for power vehicles.In view of this,based on the data-driven method,with the help of the 120 KW military diesel power generator system simulation platform developed by the team.Starting from the establishment of the health recognition model of the power vehicle diesel generator,this thesis identifies the three states of diesel generator that include health,sub-health and fault,and assesses its health status according to the concept of distance measurement quantitatively;Based on the in-depth analysis of the degraded state of the diesel engine generator(sub-health-fault),the stochastic measurement error affecting the validity of the degradation data is introduced,predicting the useful life of the diesel generator of the power supply vehicle and confirming the feasible domain of the measurement error parameter under the life prediction performance constrain.The specific work includes:1)Aiming at the comprehensive evaluation of the health status of diesel generators for power vehicles,combined with the Denoising Auto-Encoder(DAE)depth network and the Mahalanobis Distance(MD)algorithm,a health status of diesel generators based on deep learning and distance fusion is proposed.Firstly,the DAE depth network is used to extract the characteristic sequences of diesel generators under different health conditions,and the diesel generator health state recognition model is established.Next the MD algorithm is introduced to measure the distance between the DAE model output state and the normal state feature sequence.Finally quantitative assessment of the health status of the generator is given.The simulation shows that the quantitative evaluation of the health status of the diesel generators of the power supply vehicle is realized in the classification by DAE and the calculation by MD,which lays a foundation for the research of remaining life prediction.2)Aiming at the problem of useful life prediction of diesel generators for power vehicles,a method based on nonlinear wiener process for predicting the remaining life of diesel generators for power vehicles is studied.Considering thatthe degradation data is affected by the measurement error,the degradation model of the diesel generator is established by the nonlinear wiener process,and the measurement error parameters are introduced in the modeling process.In the sense of the first time,the distribution results of the remaining life of the diesel generator considering the measurement error are derived.The simulation results show that the life prediction model based on the Wiener process can realize the trend of the remaining life of diesel generators,and provide a basis for the prevention and maintenance of the situation.3)Aiming at the feasible measurement domain of the allowable measurement error under the life prediction performance constraint of the diesel engine,the corresponding calculation method is given based on the similarity measure(SM).First,the remaining life distribution of a diesel engine generator with and without measurement errors is determined.Then,according to the performance requirements of the life prediction of the power supply vehicle under different working environments,the measurement error parameters under different life prediction performance constraints are calculated.The results show that the feasible domain for calculating the measurement error parameters can provide guidance for the backward prediction of the remaining life of diesel generators,showing its potential engineering application value. |
PDF Full Text Request