The Design And Realization Of Multiple Mems Gyros

MEMS devices are smaller, lighter cheaper, and can effectively increase function density, and shorten the development cycle of satellite attitude measurement unit. But be-cause of the restrictions of technological level, MEMS gyro is still weaknesses in accuracy and reliability. Therefore, aiming at the multiple gyros system, this thesis has studied the hardware design and manufacture, the data fusion and the fault detection and isolation. The specific contents are as follows:The component model of MEMS gyro error is stated. Aiming at the system error, with the input some couple of positive and negative angular rate with the same absolute value, calculating the expectation and deference of each couple of output, the influence of the measurement noise, zero bias, and the earth rotation rate is eliminated. Using the least square method, the calibration factor of gyro output is obtained. For the random error, the ARMA(p,q) model has been set up, and the method of model parameters determination has been described.The multiple MEMS gyros system is studied. The system is divided into three parts, that is the measurement unit, the controlling unit, and the data processing unit.The general design scheme that two level, multi-group data fusion and FDI has been proposed. Then the thesis studies the reliability of the system, the fusion accuracy and time complexity of the algorithm, and on this basis, the performance index of selecting gyro number has been put forward. Using the commercial MEMS gyros and related components, the engineering sample has been designed and made, and related C program has been done. So, it realizes the function of measuring and controlling unit. At last, the function of the engineering sample has been verified on a single axis air bearing table.The data fusion method of multiple MEMS gyros is researched. First, the thesis has studied the observability of ARMA(p,q) state-space equations, and designed a Kalman filter based on this model. Considering the small angular acceleration, this thesis presents a support vector regression (SVR) based trend prediction and compensation algorithm for the designed filter. In order to overcome the weakness in generalization ability using the sample point number as the independent variable, referring the idea of ARMA(p,q) model, with the help of phase space reconstruction, the historical sample point values has been treated as the independent variables, and the generalization ability of SVR model has been effectively improved. When with a large angular acceleration, the prediction compensation algorithm is invalidation. Then, this thesis has proposed an improved data fusion algorithm based on KMOD kernel function, to integrate the processed data of each gyro. Finally, the fusion algorithm has been verified by the engineering sample.The FDI algorithm of multiple MEMS gyros is proposed. Assuming that the system only exists a single gyro fault, the thesis has designed a parity vector and support vector classification (SVC) based FDI algorithm, and analyzed the influence pattern of measure-ment noise on the FDI algorithm. Then, through the union of some single fault FDI, a modified FDI algorithm has been presented, which can deal with two simultaneous fault. In order to overcome the influence of measurement noise on the FDI, a DWT based mod-ified method has been put forward. At last, it is verified in some conditions with different fault magnitudes and combinations of fault gyros.