Adaptive Single Neuron Intelligent Control And Its Application In Vehicle Active Suspensions

As one of the most important parts of a vehicle, suspension system plays a vital role in determining the dynamic behaviors of the vehicle. Active vehicle suspensions could improve ride comfort and provide good road handling, which makes better vehicle performance than traditional passive suspensions. Recently researches on active suspensions have been gaining increased attentions at home and abroad, and it has become the advanced subject in the field of vehicle dynamics and control.A key task in designing an active suspension is to look for appropriate control laws which can provide satisfactory vehicle performance. Controller design for an active suspension is by its very nature an uncertain and multivariable control problem. Unfortunately, however, most of the active suspension control methods are usually too complex for engineering practice due to the complicated control principles or too many additional conditions. Therefore, designs of active suspension control methods which are conceptually simpler, easier to implement and effective are presented. It is based on single neuron intelligent control to deal with multivariable and uncertain suspension characteristics. It reflects basic ideas of intelligent control, i.e., controlling complex systems in a simple way. The main researches are as follows:1. The single neuron PID control and the adaptive neuron control are presented respectively,then the two similar but different traditional neuron strategies are applied to the control of vehicle active suspensions. In order to search for the optimal control parameters and avoid blindness in the tuning of control parameters, a parameter optimization design minimizing a quadratic performance index is carried out for a satisfactory control performance. The adaptability and robustness of the controllers are taken into account when subject to parameter uncertainties of the vehicle model. The comparison between the single neuron PID method and the adaptive neuron method is also conducted. The analysis results show that the single neuron PID control is able to obtain satisfactory performance and thus can be regarded as another simple, effective, robust control strategy for vehicle active suspensions, while the adaptive neuron control is inferior to the single neuron PID control and needs to improve further.2. From the viewpoint of multivariable control of complex systems, the limitations of the adaptive neuron control in active suspension are analyzed, then a new neuron algorithm called the integrated error neuron control is given. The novelty of the new method lies in the use of an integrated error approach, which is then combined with the adaptive neuron control. The integrated error neuron control is applied to active suspension of a quarter-car model. The research results demonstrate that the integrated error neuron control is superior to the adaptive neuron control in terms of almost all evaluation criteria, also has a certain level of adaptability and robustness in various driving conditions. Moreover, the integrated error neuron method provides another possible new way for multivariable control of similar systems such as inverted pendulum system.3. By combining neuron control with linear quadratic optimal control (LQR), another novel neuron control named the state feedback neuron control is presented. It utilizes the basic single neuron model, introduces state feedback and control law formulation of LQR, is thus able to deal with multivariable control problem in the same way as LQR. The state feedback neuron controller is studied using a quarter-car model, including its robustness against various vehicle parameters. The research results indicate that the state feedback neuron control not only improves the vehicle performance significantly but also performs better than LQR. Furthermore, it provides a concise and possibly more effective new way to the control of similar LQR problems.4. A new composite performance index characterizing vibration control effect of the entire suspension system is proposed. It reflects the interaction and compromise of the existing performance criteria of vehicle suspension systems, and enriches the existing performance criteria of vehicle suspensions.5. In order to validate the proposed neuron approaches in this dissertation, virtual prototyping technology is introduced into active suspension studies. A multi-body dynamics mechanical model of a quarter-car suspension system is constructed using ADAMS/View software, single neuron controllers are designed for active suspension system in MATLAB/Simulink, and then the co-simulations with ADAMS/View and MATLAB/Simulink are conducted. The simulation results show that the proposed neuron approaches are reasonable and feasible.In summary, neuron control strategies are applied to the control of vehicle active suspensions with complex,uncertain and multivariable characteristics in this dissertation. It makes good use of the advantages of single neuron controllers, such as simplicity, practicality, self-learning and self-tuning ability. Efforts are made for multivariable control problem by merely using single neuron. The research results not only provide simple, effective, mode-free, robust intelligent control strategies to adaptive control and multivariable control of vehicle active suspensions, but also offer possible new way for neuron control and its applications.