Research On ILLE And Deep Forest-Based Fault Diagnosis For Satellite Attitude Control Systems

As an important subsystem of satellite,the reliability of attitude control systems is the basic guarantee for the security and stability of satellite.In order to improve the reliability and enhance the safety and maintainability of attitude control systems,it is of great significance to carry out the research on fault diagnosis technology.However,the complex structure of satellite attitude control systems,high-dimensional and non-linear telemetry data,unknown space operation environment,multiple component faults and fault propagation in the closed-loop bring great challenges to the design of fault diagnosis methods.In this paper,data-driven methods are used to study the fault detection and identification technology of satellite attitude control systems.Firstly,For the problems of high-dimensional telemetry data,insufficient sample information and model updating in the initial operation of satellite attitude control systems,the research on fault detection technology based on incremental locally linear embedding(ILLE)is carried out.By reducing the dimension of satellite high-dimensional telemetry data and adding the useful samples in the operation process of the system to the original training database,the problem of model updating can be solved by constantly updating and improving the fault detection model.In order to reduce the influence of the satellite telemetry data not obeying the Gaussian normal distribution and the accumulation of data noise on fault detection statistics,the support vector data description(SVDD)algorithm is introduced to construct the fault detection statistics substitution for SPE and T~2 statistics.It improves the accuracy of fault detection and fault detection technology based on ILLE.In order to verify the effectiveness of the proposed method,the designed method is used to detect the fault of a high-precision satellite digital simulation platform.The test results show that the accuracy of the method is very high.Compared with the detection results of SPE and T~2 statistics,the method can effectively detect the fault of satellite attitude control systems.Secondly,for the problems of the environment disturbance torque in satellite attitude control systems and the uncertain torque of flywheel,as well as the closed-loop propagation of actuator and sensor fault,it is difficult to identify the fault.Deep forest algorithm has strong learning and generalization ability.It is not only not sensitive to super parameters and does not need a large number of fault samples,but also supports multi class classification and the training speed is fast.Compared with artificial neural network(ANN)and support vector machine(SVM),it is more suitable for fault identification of satellite attitude control systems.In this paper,according to the state parameters and characteristics of satellite telemetry data,the feature extraction method is studied.And then the deep forest algorithm is used to train the samples and establish the classification model,so as to identify the fault types.In order to verify the effectiveness and superiority of the proposed method,a semi physical simulation platform is built for simulation test.And the fault identification results of the method are compared with ANN and SVM algorithm.The test results show that the fault identification accuracy of the method is higher than ANN and SVM algorithm,and it can effectively identify the actuator and sensor faults of satellite attitude control systems.Finally,the application software of fault diagnosis on satellite attitude control systems based on ILLE and deep forest is designed and developed.The software obtains training samples and online test samples from the database through the data acquisition module and preprocesses them.It extracts the characteristics of the preprocessed training samples through the diagnostic model training module,as well as constructs the fault detection model and the fault recognition model respectively.It detects and identifies the faults of online test samples through the fault diagnosis module and outputs the diagnosis results.Therefore,the software can be directly applied to the example verification of the proposed fault detection method and fault identification method.