| Suspension is one of the important assemblies of automobile. Its function is to improve the automotive ride performance through decreasing excitation of road unevenness, and improve the handling performance through maintaining adhesion between tire and road. Spring stiffness and damping coefficient are the main parameters of suspension vibration system. They are chosen according to the optimal comprehension of ride performance and handling performance under the specific working condition. Once chosen, they can't be changed. Because the characteristic parameters of passive suspension aren't usually adjusted to the working condition and road profile, it is difficult for passive suspension to meet higher requirements of automotive ride performance and handling performance at the same time. The more improvement of its performance is limited. With the development of electronic control technology, profound researches on active suspension system and its control technology have been carried out to overcome the limitation of passive suspension and improve its performance.At present, the researches on control algorithm of the active suspension are mostly based on the simplified model of the suspension system. Although this simplified model can express some fundamental characteristics of real suspension, the nonlinear factors of the suspension structure are not taken into consideration. The simplified model described by nominal parameter (namely, the characteristic parameter of the parts of the real system) can't equal the effectiveness of the real suspension system very well. Meanwhile, it is very difficult to obtain the exact suspension mathematic model because suspension system is a complicated, time-changing and uncertain nonlinear system, and road excitation is also at random. The automotive active suspension can't be easily controlled if the traditional control model-based, both classical and modern control, whether linear control or nonlinear control, is used. Therefore, the intelligent control displays the obvious advantage because it doesn't depend on an accurate mathematic model and has the flexible decision mode and strong robustness. It can be applied to control uncertain and complicated system, and it is proper for automotive active suspension system with random disturbance and complicated mathematic model. Now the intelligent technology includes fuzzy logic, specialist system, neural network, genetic algorithm and their combination etc among which the development of neural network and genetic algorithm is from brain nervous system and genetic system of mankind's information process system. To a certain extent, some effects have been achieved in their researches and applications. However, these methods still have some disadvantages, for example, the neural network method is difficult to determine weight coefficients and may easily converge to local maximum. Genetic algorithm can search in the global area, but its local search ability is limited. Besides, it is not flexible and converges slowly when dealing with complicated, confusable and multi-task problem, which will affect its control performance.The immune system is one of mankind's four information process systems. It has the abilities of identifying antigen's diversity, memorizing, self-adjusting function, which can prevent itself being captured by local maximum according to affinity between antibodies etc. The immune system is a very robust adaptive system because it can maintain self-stability in the dynamic condition and deal with various robustness and uncertainties. These characteristics of immune system lay foundation for application in the field of automotive active suspension's control system.In this dissertation, the immune system and control system are combined based on these characteristics of immune system to further improve the performance of automotive active suspension. Firstly, immune algorithm is used as auxiliary functions. It is applied to automotive suspension system to build the model and learn the fuzzy logic controller(FLC) in order to get simplified model which can relatively reflect the real system and resolve design problem because fuzzy controller's design relies on specialist's experiential knowledge excessively. Secondly, it is applied to simulate controller based on training and learning of control. So the immune control strategy of active suspension is put forward. Finally, based on ADAMS/View and Hydraulic module, the virtual experimental simulation bench of hydraulic servo is built for automotive active suspension. The simulation associated with ADAMS and Matlab is implemented through ADAMS/Control interface. And virtual experimental validation of immune control strategy is realized.The main research work and innovative points in this dissertation are as follows:1. Considering that the research model of automotive active suspension is based on the simplified model, the parameters are nominal parameters, and the nonlinear factor of the suspension structure is not considered, the immune identifying method is applied to identify the automotive suspension parameters. The research results show that the identification precision of the method is better than the recursive least squares method. When there are some intrusions among the input and output signals, the recursive least squares method fails to identify, while IA method maintains to have a good result of identification, which shows that it has a good anti-jamming ability.2. Immune control strategy is proposed for automotive active suspension. Immune control strategy based on three different encoding (binary, decimal and DNA encoding) is also comparatively studied. The simulation researches show that decimal-based immune control strategy is optimal, but its convergent velocity is slow. Convergent velocity of binary-based immune control strategy is close to DNA encoding-based, but the control effect of the latter is obviously better than the former. Therefore, considering the control effect and convergent velocity, the performance of DNA encoding-based is better than decimal-based and binary-based, and it is optimal.3. Immune control strategy based on chaos immune algorithm is proposed for automotive active suspension. The results show that its control stability excels the immune evolutionary control (namely, immune control decimal-based).4. Focusing on the problem of FLC design which depend on specialist's knowledge excessively, a FLC design method IA-based is proposed for automotive active suspension. Simulation researches show that the effect of FLC which is optimally designed by IA is equivalent to that of manual design, the effect of FLC with partial rules which are extracted is better than FLC with total rules. Meanwhile, The FLC design method not only can avoid the trifle caused by manual design but also greatly enhance the efficiency, especially for high dimension's FLC design. To a certain extent, it has the practical value.5. Considering the efficiency of immune control and the singleness of FLC input, an immune-fuzzy control strategy, which combines the immune control and FLC, is proposed. The researches show that the control method has the advantage of immune control and FLC. And its control effect is superior to both of them while they control active suspension respectively.6. Because the experimental studies for automotive active suspension, being the simplified model or complicated real automobile, are complex and costly. Moreover, the experimental study is usually difficult to be preceded if academies and companies etc. don't work together. Under this circumstance, the virtual experimental simulation bench based on ADAMS/View and Matlab/Simulink is designed and researches have been carried out in the following two aspects. (1) In order to verify the validity of the virtual experimental simulation bench, bsed on ADAMS/Control module, PID control method is applied to automotive active suspension which control force is respectively generated from ideal force and hydraulic servo. The results show that the control effect of two situations is consistent and prove that the virtual experimental simulation bench, including hydraulic servo, is feasible. The virtual experimental method may shorten experimental periods, and reduce experimental expenses. It is an efficient method for research on control strategy of the automotive active suspension, and also plays a theoretic role in designing the real active suspension experimental bench.(2) Based on virtual experimental bench above-mentioned and Matlab interface, the virtual experimental research on immune evolutionary control strategy is completed. The virtual experimental results show that the control effect is better when active suspension is controlled by immune evolutionary control.
|
PDF Full Text Request