Integrated Navigation Of Fault Diagnosis System Design Based On Neural Network

The high-precision integrated navigation system is the integrated core of positioning,navigation and timing in military and civil fields.If the navigation system fails to be diagnosed in time,it will pose a great threat to the safety and economic benefits.Therefore,the fault diagnosis technology is very important to ensure the stable and healthy operation of the integrated navigation.For integrated navigation system with nonlinear,high noise,high coupling inherent features,and the carrier operation is accompanied by a large number of real-time feature information output.All above bring about traditional fault diagnosis methods generally have the defects of single fault diagnosis type and low diagnosis rate,and the neural network has good capability of nonlinear characteristics of data processing.Therefore,this paper develops a set of integrated navigation fault diagnosis software system based on neural network fault diagnosis method.The software system can realize the fault diagnosis of integrated navigation system effectively and has the good reliability.In the process of solving the above problems,this paper mainly consists of the construction of integrated navigation simulation model,integrated navigation data generation,integrated navigation fault diagnosis these three aspects to achieve the design of the software system.The main work is as follows:Firstly,the composition structure,combination method,filtering method and other contents of the integrated navigation are studied and analyzed.INS navigation(inertial navigation)and GPS navigation are used as subsystems,the subsystem used loose coupling,Kalman indirect filtering and inertial navigation tool to build the simulation model of INS/GPS integrated navigation.Secondly,based on the INS/GPS integrated navigation simulation model,the generation of integrated navigation data was realized.Before data generation,the causes and classification of faults were analyzed,determining the types of failures that have a greater impact on the integrated navigation work and frequently occur,such as sudden faults,slow-changing faults.Then according to the characteristics of type fault establishes a mathematical model,determines the relevant parameter values of the mathematical model,such as the fault amplitude,and finally superimposes the fault mathematical model into the integrated navigation simulation model for fault injection to realize data generation.Thirdly,aiming at the low diagnosis rate of the existing integrated navigation fault diagnosis method,this paper proposes a neural network fault diagnosis method.At the same time,in order to improve the accuracy of fault diagnosis,this paper adopts two fault diagnosis methods according to the different characteristics of sudden faults and slow-changing faults and the unique advantages of different neural networks.Among them,the classical BP neural network is used for sudden faults and the Deep Belief Networks(DBN)is used for slow-changing faults.Finally,integrated navigation injection fault data and normal state data constructed a data set,completed the training and verification of the corresponding network mode.Finally,the developed integrated navigation fault diagnosis software system has undergone various functional tests.The test experiments and comparison experiments show that the system can diagnose six types of failures,including gyro mutation failure,meter addition mutation failure,GPS mutation failure,gyro ramp failure,meter addition failure,and GPS ramp failure.The fault diagnosis rate is above 90%,and stable at 98.429%.Compared with other methods,it has a higher fault diagnosis rate,also has better market application value in the field of integrated navigation fault diagnosis.