Feature enhancement method for residual life prediction

Abstract

The invention discloses a feature enhancement method for residual life prediction, and belongs to the field of aero-engine fault prediction and health management. The method comprises the following steps: firstly, normalizing engine sensor data, and calculating residual life values of training set data and test set data; secondly, dimension reduction processing is conducted on the training data set and the test data set, and training data set samples and test data set samples are extracted in a sliding window mode; thirdly, performing data feature enhancement on the training set sample, the test sample and the sample, and determining the sample after data feature enhancement; and finally, setting a deep learning neural network long-term and short-term memory structure, adding a sample for training, and predicting test data by using a neural network model. The deep neural network model is established on the basis of a data driving form, the method is irrelevant to actual engine models, the model can be migrated to engines of different models for use by training different data sets, and the method has certain universality.

Description

technical field [0001] The invention proposes a new data enhancement method for predicting the remaining life of an aeroengine system or component based on a Tensorflow deep learning framework, which belongs to the field of aeroengine failure prediction and health management (PHM). Background technique [0002] The research object of the present invention is the turbofan engine degradation simulation data set disclosed by NASA. This dataset uses C-MAPSS for engine degradation simulation. Four different sets are simulated under different combinations of conditions and failure modes. Multiple sensor channel data are recorded to characterize the failure evolution. [0003] This dataset is divided into four sub-datasets, and each dataset is divided into training dataset, testing dataset and real system or component remaining life value (RUL). For each sub-data set, the training data set and the test data set are composed of multiple parameters. The parameters include engine num...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problem that directly using standard normalized data added to the LSTM network for training cannot fully learn and mine the characteristics and information of the original data set, the present invention proposes a feature extraction method based on data augmentation

Images