Research on anomaly detection of turbine engine rotor unbalance based on optimized generative adversarial model
In order to investigate the anomaly detection of rotor unbalance under high-temperature conditions and to overcome challenges of the deficiencies of losing valuable data and insensitivity to the information of time sequence,an optimized generative adversarial model has been explored in this paper.This model only uses normal data as input,combining the advantages of LSTM network for time series data processing and VAE network for excellent analysis of potential features,which effectively address the problem of insufficient anomaly data,and achieve unsupervised training.Taking the example of the high-pressure rotor system of a missile turbine engine operating at 600℃,the model can make reasonably accurate judgments regarding the moments when the unbalance occurs.This approach could achieve end-to-end feature extraction and anomaly detection of unbalance.In addition,this model has been compared to another three types of anomaly detection algorithms.The results show that the evaluation indicators of proposed model are all above 99%,which can more effectively assess the healthy status of time-series information,providing a more comprehensive anomaly assessment and offering valuable guidance for the normal operational lifespan of turbine engines.
NETL
NSTL
万方数据
