| Guided munitions are increasing widely applied in high-tech modern wars. Navigation and guidance system is one of the keys of Guided munitions. With the development of satellite navigation system(SNS), SNS/SINS(strapdown inertial navigatoion system) integrated navigation has been the leading development direction. But SNS is likely to be unused because of outside disturbance. Toward SNS, the proper way is to use but not depend. SINS has undisturbed and autonomous navigation capacity. Therefore, SINS is very significant for military application with certain precision.Taking medium-precision SINS suitable for inertial guidance as object, this dissertation intends to improve navigation precision and reduce the cost of SINS. Main contents are about the key technologies of SINS, including data-sampling, gyro's calibration and compensation, SINS alignment, compensation of magnetic heading system(MHS). The details of main contents are the following aspects:(1) Based on the introduced principles and developments of SINS, inertial instrument, navigation computer and data-sampling technology, we design a SINS including an inertial measurement unit(IMU) with domestic flexibility gyroscopes and quartz accelerometers, a navigation computer system with the core of digital signal processor(DSP) and complex programmable logical device(CPLD), and two high-precision and low-cost data-sampling systems with pipeline A/D chips and sigma-delta A/D chips respectively.(2) Toward SINS, the calibration methods of piecewise nonlinear interpolating(PNIP) and multilayer feedforward neural network(MLFN) are put forward to compensate the nonlinear error of gyroscope's scale factor. The PNIP can subdivide the scale factor and minish nonlinear error about 6.2 times and 2.4 times compared with least square fitting and piecewise compensation method respectively. Experiment results show that the calibration result of scale factor using MLFN is close to that of PNIP, and also indicate that MLFN is feasible. A retraining method is proposed to improve the performance of the trained MLFN. Results show that the retrained MLFN can much more approach the real scale factor and minish the nonlinear error about 1.7 times compared with the unrestrained MLFN.(3) In order to meet the requirement of SINS'batch production, an automatic calibration and test method is designed. Considering SINS'characters, we design the communication protocol and realize serial communication mechanism between monitor computer and SINS. Based on a turntable system and utilizing the serial communication mechanism, the calibration process of SINS is automatically realized without adding any hardware equipment. The automatic process includes calibration flow control, data record, calibration result computation and parameters storage.(4) Considering noisy data of inertial instruments which could depress SINS alignment precision, lifting wavelet is applied into the signal de-noising of inertial instruments and SINS coarse alignment. Results show that lifting wavelet algorithm is effective to the signal de-noising of inertial instruments. Also it can improve the alignment precision of SINS. Compared with classic wavelet, lifting wavelet can save almost half computation time and is more suitable to real-time and online application.(5) Based on the introduction of transfer alignment methods, a neural network(NN) transfer alignment method is put forward to compensate the errors of salve inertial navigation system(INS). In the NN model, NN inputs are the velocity and attitude of salve INS, and NN output references are the errors of velocity and attitude between master INS and slave INS. Simulation results show that the NN model can make satisfactory compensation to the velocity and attitude of slave INS. Also NN models with deferent neurons could have deferent compensation effect, and we ascertain the best NN architecture with 6 neurons by simulation experiments. Vehicle test results show that NN transfer alignment method has effective compensation to velocity and attitude of slave INS with average of velocity error below 0.15m/s, averages of pitch error and roll error below 0.03°, and average of yaw error below 0.03°. And errors of compensated velocity and attitude are stable and convergent. Unaided navigation position error compensated by the NN is below 8m in 60s and below 17m in 100s, which shows that the NN model can effectively restain the divergence of position error.(6) The principle of MHS is studied. A de-noising method and an anomaly detector are presented respectively to de-noise and detect the signal of a low-cost MHS. Considering the problem of MHS measurements including a lot of noises, lifting Wavelet is used in de-noising MHS measurements. MHS measurements could be anomaly due to magnetic disturbance from external interferences. Toward the anomaly measurements of MHS, lifting Wavelet is applied into detecting the anomaly signal. Experiment results show that lifting Wavelet could effectively de-noise the MHS measurements and detect the anomaly signal. After eliminating the anomaly measurements, the revised measurements could improve the usability of MHS heading.(7) According to the characters of integrated navigation, a neural network is bringed forward to compensate the error of a low-cost MHS. The error sources of MHS are studied and the compensation methods are analyzed. When the Global Positioning System(GPS) is available, a multilayer feedforward neural network is trained to compensate MHS by the learning method of kalman filter and with the reference of SINS/GPS integrated navigation result. Experiment results show that the neural network can make a significant effect and reduce the heading error of MHS from±15? to±1?.Baesd on the integration hardware system of flexible gyroscopes, this dissertation majors in researching the way to improve SINS'precision and to reduce system cost. And also the methods of this dissertation are suitable for other medium-precision SINS, such as the SINS based on fiber optical gyros.
|
PDF Full Text Request