Research On UAV Electrical System Fault Diagnosis Virtual Instrument Technology

UAVs play an increasingly important role in military and civilian fields since it can travelin and out of a variety of hazardous airspace for a quite long time. Virtual instrumenttechnology adhering to the "software is the instrument" design idea, so that the user can makethe same equipment to achieve different functions by only changing the software or addingsimple accessories. As a different sensor configuration assistive technology, information fusiontechnology can take full advantage of useful information in multi-sensor, make up each andverificate each other, provide more accurate, reliable, comprehensive information for faultdiagnosis, and can greatly reduce the data traffic to be processed by the classifier. The purposeof this paper is applying the "software is the instrument" design idea of virtual instrumenttechnology and sensor data fusion technology to UAV electrical system fault diagnosis,improving the system’s portability and reduce the difficulty of system upgrades while improvingthe accuracy and efficiency of the diagnosis system.Based on the study of UAV electrical system fault characteristics, this paper proposed afault characteristics extraction method which using the extended Kalman filter for multi-sensordata fusion, so it has got a solution of the UAV attitude and position. The multiple sensor datafusion eventually become UAV pose vectors. The fault classifier support vector machine (SVM)selected by this paper effectively solved the difficulty in fault identification of the high-dimensional feature vectors and nonlinear system. For multi-class classification problem in thisarticle, it was split into a combination of one two-class, one four-class and one six-classclassification problem, this method can avoid the problem that classification results is affectedbecause of too large SVM scale caused by excessive classification. As can be seen from thesimulation results, the fault diagnosis of classifier has a higher accuracy rate for most sensors,especially for the normal state, it can greatly save system resources to improve the real-timeperformance of the system, but for the diversity between different sensors and different fault recognition accuracy rate, data fusion and classifier design needs further research andimprovement.In the system design phase, this paper is in strict accordance with the "software is theinstrument" design idea to do software and hardware design. It equipped with a variousstandard interface for hardware system, for other system resources, it also gave a fullconsideration on their upgrades and transplants. It used the method of modular design, wroteeach block according to their features, and eventually formed a complete software system. Thewhole system has the characteristics of good portability, easy to upgrade and so on.