Research On Turbofan Engine Remaining Life Prediction Algorithm And Component Development

With the increasing size and volume of turbofan engines,the complexity and coupling of internal components are also increasing,and the probability of failure of turbofan engines during operation is also increasing,so the turbofan is accurately predicted The method of remaining engine life has become a research hotspot.Harbin Institute of Technology has developed a basic platform(HIT-Joint Test Platform,H-JTP for short)that can be used to support various capability tests and information-based distributed testing,to provide a virtual test environment and laboratory for joint tests Related resources needed.At present,the platform lacks an algorithm for predicting the remaining life of turbofan engines.For this reason,this subject has carried out research work on a prediction algorithm for the remaining life of turbofan engines applicable to the joint test platform,and developed a training software for turbofan engine remaining life prediction models.And the remaining life prediction component based on the joint experiment platform.The main work of the paper is as follows:1.The data-driven turbofan engine remaining life prediction algorithm is studied,and the CNN-JANET model with the highest prediction accuracy is selected from them.2.Aiming at the problems that the CNN-JANET algorithm runs slowly,occupies large computing resources,and cannot be applied to the H-JTP platform,a turbofan engine remaining life prediction model combining principal component analysis(PCA)and extreme gradient boost(XGBoost)is proposed.The model uses PCA to reduce the amount of data and the characteristics of XGBoost's parallel calculation of features and pre-pruning,while taking into account the calculation accuracy and shortening the training time.3.Using an object-oriented approach,an offline training software for the prediction model of the remaining life of the turbofan engine is developed,which can save the trained model and provide a trained prediction model for the remaining life prediction component of the turbofan engine.4.Based on the requirements of the H-JTP platform,the static model,dynamic model and component interface of the turbofan engine remaining life prediction component were designed,the code was written,and the turbofan engine remaining life prediction component was developed.Experimental and application verification results show that the preferred CNNJANET network is the most accurate model among the existing prediction models.The PCA-XGBoost algorithm proposed in this paper can effectively reduce the dimensionality of the original data,and quickly train the model and predict the remaining life of the turbofan engine;the development of the turbofan engine remaining life prediction model training software can train CNN-JANET,PCAXGBoost Multiple prediction models;the developed remaining life prediction component of turbofan engine can realize the remaining life prediction function and meet the actual application requirements of H-JTP platform.