AUV state monitoring method based on dynamic model and complex network theory

Abstract

The invention relates to an AUV (Autonomous Underwater Vehicle) state monitoring method based on a dynamic model and a complex network theory, which is characterized in that a simplified dynamic model with a constraint relationship is taken as a basis, a neural network is utilized to carry out parameter identification to obtain an AUV dynamic model with wide applicability, and a complex network theory is taken as a basis to analyze a residual sequence generated by the constructed model and the measured value, feature matrixes of different states are extracted, the feature matrixes are classified by using a support vector machine, and a normal state and an abnormal state are accurately identified and classified by monitoring the running state of the AUV in real time during the navigation period of the AUV; according to the scheme, the dynamic model is converted into the simplified model with the constraint term, the method has high generalization, does not need high sampling frequency, does not depend on a large amount of training data, analyzes the fluctuation characteristics of the abnormal state, can accurately recognize the normal state and the abnormal state, and achieves the accurate recognition of the abnormal state.

Description

technical field [0001] The invention belongs to the field of abnormal state monitoring of underwater robots, is mainly used for autonomous underwater vehicles (AUV), and specifically relates to a method for monitoring the state of an AUV. Background technique [0002] Today, autonomous underwater vehicles are increasingly recognized in fields such as national defense, marine geology, and industry. However, with the expansion of the application field, AUVs are vulnerable to strong external disturbances and sudden failures when performing tasks, which will significantly affect the stability and safety of AUVs. Therefore, in order to ensure that the AUV can complete the expected tasks safely, efficiently and reliably, it is necessary to monitor the abnormal state of the AUV during the mission. [0003] At present, the relevant technologies for condition monitoring mainly include: signal analysis method, model method and artificial intelligence method; According to the differe...

AI Technical Summary

Problems solved by technology

[0005] 1. For the signal analysis method, since time domain and frequency domain analysis need to use high-frequency data collection, this will occupy the data interaction capability inside the AUV; and low-frequency data collection cannot fully demonstrate the characteristics of the abnormal state of the AUV; and the The method is limited by the method itself, and it is easy to produce wrong analysis results, resulting in lower recognition accuracy

[0006] 2. For the model method, since the AUV is a complex model with high nonlinearity, strong coupling, and multiple degrees of freedom, it has a large number of parameters and is difficult to solve, and the commonly used parameter identification methods are easily affected by the noise of the measurement data, resulting in model parameters Does not converge, affecting model output accuracy

[0007] 3. When using artificial intelligence methods to identify the state of AUVs, the meaning of its features is not clear, and the recognition accuracy obtained under limited data is limited

Images