Research on Fault Diagnosis Technology of Helicopter Fuel System Based on Neural Networks

This study develops a fault diagnosis model for a specific helicopter fuel system using a Feedforward Neural Network (FNN). The methodology involves the following steps: employing fluid simulation software to create a simulation model of the helicopter's fuel system and conducting fault simulations through fault injection. The FNN, supported by the Keras-Tensor Flow framework, is utilized to construct the fault diagnosis model. Data from normal operations and various fault conditions are used as samples for training and validating the model. The efficacy of the neural network tomography method for fault diagnosis is confirmed, achieving significantly improved fault recognition accuracy.