Method for predicting residual life of avionics product

Abstract

According to the method for predicting the residual life of the avionics product, the prediction precision and reliability of the residual life can be improved. The method is realized through the following technical scheme: a preset sensor obtains functional index data of a real avionics product and a similar avionics product, static feature extraction and normalization processing are carried out, and the functional index data are constructed into a sample set; initializing a plurality of LSTM time sequence models to form an LSTM model group; in transfer learning, carrying out transfer training and target domain global fine tuning on each LSTM time sequence model by using a real product sample set; calculating a fusion weight according to an improved voting weighting algorithm; a standard interface of a complete prediction framework is defined, functions of the complete prediction framework are split and packaged in a sample construction module, an improved LSTM model construction module and a prediction framework integration module, the degradation evolution and the residual life of a target product are predicted, and the degradation evolution and the residual life of the target product are accurately predicted by the prediction framework.

Description

technical field [0001] The invention relates to a ground maintenance system for avionics products, in particular to a method for predicting the remaining life of avionics products based on the fusion of multiple long and short-term memory neural networks. Background technique [0002] With the passage of working time, under the combined effect of internal and external factors, the performance and health status of any equipment will inevitably show a trend of decline, and performance degradation will inevitably occur until failure. When the recession reaches a certain level, the equipment will not be able to complete normal tasks and functions, resulting in irreversible economic losses and waste of resources. Aerospace equipment often faces the problem of high failure rate, and the traditional inspection and maintenance methods with regular replacement as the mainstream can no longer meet the needs of operation and maintenance support under the trend of intelligence, modulari...

AI Technical Summary

Problems solved by technology

The core of the long short-term memory neural network LSTM model is the network unit state and three gate structures (forget gate, input gate and output gate). The network unit state is composed of two activation functions (sigmoid and tanh). Through the three gate structures It effectively solves the problems of gradient disappearance and weak long-term memory ability of the recurrent neural network, but its prediction effect depends on the degree of consistency between the distribution of training data and prediction data, and the learning ability of a single model is not enough to support the definition of real from similar product data. Product remaining life

Images