Study On The Control Algorithm Of Electromagnetic Valve Nonlinearity Using BP Neural Networks In ABS

As the most representative active safety device of modern automobile, theantilock brake system can assure the car's control stability and direction stabilityin braking process at the greatest degree; meanwhile, it can exert the brakingsystem's braking efficacy effectively, shorten the braking length. Since the late1980s, the ABS active security technology is used widely, and become thestandard component of medium and heavy vehicles. It is composed of controller,electromagnetic valve and wheel speed sensor. When emergency braking, thebrake pressure of driver pedal is overlarge, wheel speed sensor and controller canprobe the wheel's braking lock tendency, in this condition, controller controls thebraking system to diminish braking pressure. When the wheel's speed restore andfloor friction has a tendency of decrease, the controller controls the braking systemto increase braking pressure. This makes wheels stay in the optimum, and use flooradhesion effectively, get the optimum braking length and braking stability. It canbe seen that, ABS is able to avoid and lighten traffic accident effectively.As one of the most important component of ABS, electromagnetic valve'smain effect is accept the control command from the electronic control element, byusing control valve switch, to increase or decrease the braking pressure directly orindirectly. In ideal condition, when control element gives ABS electromagnetic valve a signal to let the two valves open or close, it can do the order immediately,the pressure value can change immediately too. But in the actual conditions, forthe machinery construction of electromagnetic valve (deadtime property) , thereis a delay between the control signal's sent out and valve's open, it will open whileexceeding a valve value; in the same way, there is a defer when control signal let itclose, while exceeding a valve value. As the electromagnetic valve is the mainactuator component of antilock brake system , whose dynamic response has a veryimportant effect on the system performance, deciding if the high performancecontrol system can realization or not. We can say that, the dynamical characteristicof electromagnetic valve is the premise of ABS working, without the gooddynamic response of the electromagnetic valve, ABS can not work normally. Atpresent, the exploitation and research to antilock brake system is extensivelyconcerned by the domestic and international scholars and engineers, but mostresearch is on antilock brake control logic or algorithm, and the correlation studyon the action element characteristic to influence system dynamic performance isfew.This article binds the graduation project"active anti-skidding integratedhardware at circuit test bed and test method", research on the ABS administerdevice's nonlinear response characteristic, on the basis of deep research onelectromagnetic valve's physical construct and operating principle, build theelectromagnetic valve's mathematical model under Simulink.To the design ofelectromagnetic valve controller, the choice of control algorithm decides thecontroller behavior's property. The tradition automatic control, which includesclassics theory and modern control theory, has an common characteristic, that thesynthetic design of controller is founded on the accurate mathematical model ofcontrolled object, but in practical industrial production, the system has severalaffecting factors, which is very complex, it is very difficult to build accurate mathematical model, even impossible .Under this condition, neural networkscontrol theory is significant. Because neural networks control is not need to buildmathematical model, which only depends on actual system's input-output data, tobuild neural networks model and control the system. By model analysis, bindingthe electromagnetic valve dynamic characteristic, we can find that neural networkis better than traditional analysis control method.The ABS working process is devided into the stage of braking pressuremaintenance,braking pressure decrease and braking pressure increase . In brakingprocess, ABS by uses the wheel's braking pressure which tends to lock cyclesmaintenance- decrease- increase process , lets the wheel's slip ratio stay near thescope of peak value adhesion coefficient, until automobile speed is very low or thebrake chamber output pressure is no longer make the wheel is going to brake.Therefore, The ABS control the electromagnetic valve to pressure increase,pressure decrease and pressure maintenance. When pressure maintenance, theelectromagnetic valve's intake valve and exhaust valve is closed, air pressure isstability. So, the sample of designed controller only need to collect theelectromagnetic valve's data when pressure increase and pressure decrease, the BPneural networks controller's train and analysis is also divided into pressureincrease and pressure decrease condition. After research on BP network's trainingprocess in MATLAB, this article discusses the electromagnetic valve according tothe two working conditions. As after the control's basic structure is confirm, thechoice of two parameters is also impact the network's training effect, that thenumber of the hidden neurons and learning rate's setting. This article discusses thetwo parameter's setting separately, determines the optimum parameter byobserving training errors curve diagram.In the end, for testing the designed electromagnetic valve neural network controller, it applied the ABS test bed for graduation project to process simulationanalyses in the article. By analysis of experiments on the low adhesion roadcondition, it can be seen that, performance of ABS has been obviously enhancedafter added the neural network controller. It also validated the effectiveness ofalgorithm based on the BP neural networks. On the other hand, it confirmed thatthe hysteresis nonlinear response of the actuator had adverse effect to the behaviorof ABS system.Due to time limited, the controller only processed simulation testing on lowadhesion road condition and not on other conditions in the paper. As a result, thenext research will focus on validating the controller combined with the algorithmon complicated road condition, and planning to embed the electromagnetic valveneural network controller into the ECU of ABS system, and realize theindustrialization,utility finally.