Domain adversarial neural network-based nonlinear system identification for helicopter transmission system

Helicopter transmission system identification provides model support for fault diagnosis, health monitoring, controller design and so on. The great advantages of deep neural networks are shown in strong nonlinear identification such as helicopter transmission systems. In addition, transfer learning has advantages in the identification of helicopter transmission system with complex and changeable working conditions. In deep transfer learning, domain adversarial neural network can establish a generalization model in complex and variable working conditions, and fast and high-precision identification can be achieved by the network-based transfer method. Therefore, in order to take into account the accuracy and speediness of strong nonlinear helicopter transmission system identification with complicated and variable working conditions, this paper proposes a nonlinear system identification strategy based on domain adversarial neural network in deep transfer learning, which combines pre-trained and fine-tuning. Firstly, the domain adversarial neural network is pre-trained offline under various working conditions. Then, the pre-trained model is fine-tuned online in a single working condition to quickly obtain the high-precision model. The proposed nonlinear system identification strategy can achieve fast and accurate online identification under different working conditions. Finally, the experiment is carried out on the helicopter transmission test platform. The experimental results show that the pre-trained model with four working conditions can achieve high-precision identification under 16 working conditions. Furthermore, the model is fine-tuned based on the pre-trained model. The fine-tuning time is reduced by 41.5% compared to the training time of the same structure model.