| Along with the development of high way construction in china, road transportation will play a more important role in Chinese national economy. Chinese Automobile industry is faced with the serious problem for its development: the first is to increase the vehicle's handling stability and ride comfort with high speed driving; the second is to decrease the damage to the high way surface. The stiffness of Air spring suspension can be controlled through adjusting air pressure in air spring and the expected characteristic curve can be obtained. This paper use intelligent control method to control semi-active air spring suspension. The result is that the vehicle's ride comfort and ride safety can be improved greatly. Fuzzy neural network control is the focus of the research on the intelligent control and automation field in recent years. It is a new technology with strong express capability of fuzzy logic's inference and neural network's self-learning capability, which combines fuzzy control theory and neural network. The thesis is the key part of the research for the project of JiLin Science and technology's development"The research of Automobile electronic control system for air spring suspension".The main work is to research into control strategy and control system design for air spring suspension. At the same time a test bed for air spring suspension has been designed. It can provide test method to verify the control strategy. 1.The air spring suspension system consists of air spring,guided framework,electro-magnetic valve ,damper,sensor,gas storage and so on. The air spring is composed of air spring bellows,base and top plate. Air springs have strong nonlinear stiffness characteristic due to different bases and air spring bellows, which is a complicated vulcanized material of rubber and lining cloth. Air springs can have perfect stiffness characteristic by using it's nonlinear stiffness characteristic, which will improve the vehicle's performance. In addition, the stiffness characteristic of air spring has a long time-delay in response to electro-magnetic valve. All these factors have influence on the control effect. The nonlinear stiffness characteristic of the air spring is due to rubber material nonlinear,composite of rubber and lining cloth nonlinear, geometry nonlinear and contact nonlinear. So air spring suspension is considered a nonlinear suspension with adjustable stiffness. When the load of vehicle changes, the stiffness values of air spring can been changed by electro-magnetic valve. But the frequency of automobile body and the height of the car body will not been changed basically. This will improve the vehicle's ride comfort greatly. When the vehicle is in the conditions of sharp turn,acceleration and deceleration,the state of the car body can been controlled with the adjusting stiffness of air spring. This will also improve the vehicle's handling stability. In addition, the weight of air suspension is light and the stiffness value of air spring is adjustable. When the vehicle is in the condition of high speed, the friction between the tire and the ground will be increased and the brake distance will be shortened. This will also improve the vehicle's handling stability. Because the pressure of air spring can varies with the inflation/deflation time of the air spring on the designed test bed, the relation curve between stiffness of air spring and inflation/deflation time of the air spring under different initial pressure can be obtained. This will provide the gist for designing of controller. 2.The key issue for the elect-control air spring suspension is the design of control system and optimization of control algorithm. The paper presents a fuzzyneural network controller for the quarter-vehicle suspension model and a fuzzy controller for the half-vehicle suspension model because of the nonlinear characteristic,time-delay and some uncertainty of air suspension system. The fuzzy neural network controller that the paper designed has six layers fuzzy neural network.The first layer is the input layer, and the input variables are body's perpendicular vibration acceleration RMS(root mean square) error and error change rate. Each one has 5 fuzzy elements. The sixth layer is output layer with one variable, which is the on/off time of electro-magnetic valve, and it also has 5 fuzzy elements. Layer five is the layer of fuzzy control rules. In this paper, 25 fuzzy control rules are employed. The other layers are the membership-function layers. The membership function used in this paper for fuzzification is of a Gaussian type. GA(genetic algorithm) off-line optimization can give initial weigh value of fuzzy neural network controller. In order to improve adaptability of controller, the weigh value of fuzzy neural network controller is on-line modified by BP algorithm. The simulation on the air suspension is carried out by used on the basis of model reference adaptive control, the input of the controller is from reference model(initial pressure of air spring is the same). It represents the RMS of perpendicular vibration acceleration of the sprung mass multiplied by a coefficient of attenuation. The coefficient of attenuation equals 0.7. The control output of fuzzy neural network controller is constrained by maximum stroke. The random road input is used in this paper(the road grade is B and C, the speed is 50km/h and120km/h). The simulation on the quarter-vehicle air suspension model has been carried out.The good simulation result has been obtained. The fuzzy controll system in the paper consists of two fuzzy controllers (one is for perpendicular vibration, and the other is for pitching motion) and one logical controller. The input variables of the fuzzy controller for perpendicular vibration are the body's perpendicular velocityv and accelerationαand each of them hasthree fuzzy sets. The output is M 1(it is the input of logical controller) and it has five fuzzy sets. The input variables of the fuzzy controller for pitching motion are the body's pitch angular velocity w and accelerationεand each of them has three fuzzy sets. The output is M 2(it is the input of logical controller) and it has five fuzzy sets. Through some mathematic and logical operations, the logical controller using the variables M 1 and M 2 as its input variables can export the forces F1 and F2 . And then the simulation on the half-vehicle air suspension model (parameters are from DD6115H Bus) has been carried out .The simulation result indicates that the controller in the paper can increase the ride comfort and handling stability of the vehicle. 3.To verify the effect of fuzzy neural network control algorithm used in this paper, a test system for the quarter-vehicle air suspension model has been designed according to the performance requirements to air spring suspension and practical conditions. The electro-hydraulic servo actuator system of Test Center in Jilin University is applied to the design of the test system. The dynamic simulation and stiffness test of air suspension can be carried out in this test system. Many different control algorithms can be applied to the stiffness control of air suspension in this test system. The test results indicate:The maximum and RMS of the perpendicular vibration acceleration of the sprung mass decline 35.5% and 22.2% respectively; The maximum and RMS of dynamic tyre load decline 6.01% and 13.09% respectively;The amplitude of transfer function for the perpendicular vibration acceleration of sprung mass relative to road input decline 13.9% at the point of the first order frequency;And the maximum and RMS of the perpendicular vibration acceleration of the unsprung mass have little increase when using fuzzy neural network controller . The results show the fuzzy neural network control system used in this paper...
|
PDF Full Text Request