Support-Vector Modeling And Compensation For The Effect Of Structural Factors On Electrical Performance Of Electronic Equipments

With the trend of electronic equipments toward higher frequency, higher gain, faster response, higher pointing accuracy, higher density and smaller volume, the effect of structural factors on electrical performance becomes more and more prominent, which brings about the electro-mechanical coupling problem in electronic equipments. In order to solve the problem, the traditional methods such as improving the manufacturing accuracy of mechanical structure, multiple backup products and experience tuning have been utilized in practical engineering. However, the approaches have resulted in some disadvantages such as a long development cycle and high costs. In order to meet the need of high-performance electronic equipments, explore the effect of structural factors on the electrical performance and compensate the effect, we have focused on the research of modeling and compensation for the effect of structural factors on electrical performance of electronic equipments by using support vector regression in this dissertation. The main contributions of this dissertation can be summarized as follows:(1) In order to study the effect of structural factors on electrical performance of electronic equipments, the thesis has proposed a data-driven approach to solve the modeling problem based on support vector regression. In the approach, two kinds of support-vector models which are the modification of intermediate electric parameters and the establishment of meta-model have been presented to obtain the influencing relationship between structural factors and the electrical performance of electronic equipments. Subsequently, in order to improve the modeling accuracy of small data set from the practical engineering, a strategy of expanding the training data set has been presented by using some experiments from a prior simulator or an experimental device. Two experiments from a microstrip antenna and a panel slotted-array antenna have been carried out, and the experimental results confirm the effectiveness of the proposed data-driven modeling approach.(2) In order to obtain an accurate model from a limited amount of measuring data, a novel multiple kernel linear programming support vector regression with priori knowledge has been proposed in this thesis. In the algorithm, we firstly incorporate the data that is possible biased form a prior simulator into the existing linear programming support vector regression by modifying optimization objectives and inequality constraints. Then, multiple kernels are introduced to integrate the linear programming support vector regression with prior knowledge, in order to achieve an accurate model for complex and changeful problems. Finally, three examples from a complex function approximation, a system identification and the bandwidth calculation of a patch antenna show that the proposed algorithm is effective, and that the obtained model is sparse and accurate. The algorithm shows great potential in some practical applications, where the experimental data is very few, and some priori knowledge is available from the practical engineering.(3) In order to improve the fabrication efficiency and electrical performance of assembled cavity filters, the thesis has proposed a support-vector modeling approach that can reveal the effect of manufacturing precision on electrical performance of cavity filters by using some data sets from the manufacturing of filters. In the method, a data-based relationship which can show the effect of manufacturing precisions on the coupling matrix of filters is firstly built by a multi-kernel linear programming support vector regression proposed in the thesis. Then the traditional formulation of the filter electrical performance is modified by the developed relationship, and we can obtain a model which can reveal the effect of manufacturing precision on electrical performance of cavity filters.Finally, a method of selecting the optimal manufacturing precision is formulated by using the developed model, and the obtained results can assist the fabrication of the same filter in the future. Some experiments from an electrically tunable filter are carried out, and the results confirm the effectiveness of the model.(4) According to the modeling method of cavity filters above, a computer-aided tuning method based on support vector regression has been proposed to solve the tuning problem which is costly, time consuming, and requires skilled technicians. In the tuning method, a model that reveals the relationships between the tunable screws and filter electrical performance is firstly developed by using a multi-kernel linear programming support vector regression, according to the data from the tuning experience of skilled technicians. Then, a tuning procedure of cavity filters is constructed by using the developed model. Finally, some experiments from two cavity filters are carried out, and the experimental results confirm the effectiveness of the tuning method. The approach is particularly suited to the computer-aided tuning of volume-producing filters and can compensate the influence of design impreciseness and manufacturing tolerances on the electrical performance.(5) The thesis has presented a support-vector friction modeling and adaptive compensating method to reduce or eliminate the effect of friction on the performance of radar servo systems. In the method, we have firstly presented a support-vector friction model, in which the static friction is a meta-model established by utilizing some experimental data. Two static friction models which are adaptive Coulomb model (ACM) and fixed Coulomb model (FCM) have been developed respectively by using support vector regression. Then, based on the developed friction models, we have designed a backstepping control law and adaptively compensated the effect of friction on the performance of servo systems. The stability of the closed-loop system is proved by using Lyapunov stability analysis. Finally, experimental results show that the proposed friction method is effective, and that the compensating method using ACM has a higher performance of servo performance than the one using FCM.