The Research Of Control Algorithm Of The Vehicle Semi-active Suspension Based On IPSO-BP Network

The comfort and handling of vehicle driving is a measure criteria of the vehicle performance. The vehicle suspension system has great influence on the performance. So, people began to study the use of semi-active suspension system, which can change some parameters of the suspension system, to adapt to the vehicle traveling conditions better. However, in the semi-active suspension control system, the study uses the simplified model of algebraic expression of a semi-active suspension system, which can only express a number of performances of the semi-active suspension system roughly, and is difficult to express the nonlinear and the coupling, which results in failing to achieve the desired effect during the control process. This paper presents that IPSO-BP algorithm is used to have on-line identification for vehicle semi-active suspension system. With the ways of on-line control according to the on-line identification model, dSPACE test platform is built for the semi-active suspension system. This platform carries through the semi-practicality simulation on the typical road excitation, completes the debugging of the prototype development phase of the control algorithms of semi-active suspension system, which provides an important basis for further development.1,The establishment of the incentive road model and semi-active suspension model.For semi-active suspension system numerical simulation experiments and dSPACE-the-loop simulation, vehicles incentive road model is established, and give details of the relationship of space- incentive road and time-incentive road in the incentive road model, describe algebraic expressions of the semi-active suspension system model. According to these, the paper describes the working principle of variable damping shock absorber as the key components in the semi-active suspension systems and the calculation of relevant parameters. This paper detailed analyses the performance of the variable damping shock absorbers to lay the groundwork for numerical simulation experiments in the semi-active suspension system and dSPACE-the-loop simulation.2,Neural Network Pattern RecognitionFor the vehicle semi-active suspension system has nonlinear, coupling and road driving has the randomness, so the traditional algebraic expression is hard to describe the semi-active suspension system model accurately. In this paper, a neural network algorithm is proposed to have on-line identification for semi-active suspension system. Therefore, this paper discusses the feasibility of on-line identification of the semi-active suspension system with the neural network. First, the paper discusses the approximation capability of neural networks. In various types of neural network, the approximation capability of BP networks is optimal. Therefore, this paper uses the BP network as pattern recognition device of a semi-active suspension system. The convergence of BP network was discussed. The speed of convergence of neural network directly affects the control speed to the semi-active suspension system, which affects control quality of semi-active suspension system of. At the same time, this paper also demonstrated stability problems in the process of convergence of the BP network, Based on these, the model shock absorbers a numerical simulation experience is taken to validate whether the use of BP network algorithm is reasonable, at the same time, this paper provides a basis for BP network algorithm3,Improved BP Network AlgorithmIn order to make BP network work faster and more stable in semi-active suspension control system, it needs to improve BP network algorithm. This paper uses a swarm intelligence algorithm: improved particle swarm optimization algorithm to improve the BP network. Particle swarm algorithm (PSO) is an optimization algorithm that will take the solution of the object as a space, and a group of particles in the space flight, the location of each flying particle object represents a approximate solution of the particles, constantly flying objects make the solution update continually, finally find the most suitable solution, which is the working principle of PSO. But this method is easy to fall into local minima, not an optimal solution. This article improves the particle swarm algorithm, which will introduce a random value to them, later in the optimization from jumping out of local minimum. Through the test it proves that the improved particle swarm algorithm effectively improves the calculation accuracy. In this paper the improved algorithm is known as particle swarm optimization (IPSO), and IPSO algorithm is introduced to the BP network, as the network learning rules, and the traditional BP network, gradient descent learning rules, is abandoned to form a new BP network algorithm: IPSO-BP algorithm, which has a calculated speed, high precision and stable operation characteristics. Between BP network algorithm and IPSO-BP network algorithm a comparison test is taken on the same measured function optimization to achieve the same precision, BP network takes 30-step, while the PSO-BP network only takes 10 steps, which proves that IPSO-BP network algorithm is more excellent computing power.4,Control Algorithm of semi-active Suspension SystemThe control algorithm of semi-active suspension system is currently used: ceiling control algorithms, PID control algorithm, optimal control algorithms, but these algorithms is based on linear systems,in these algorithms in such a nonlinear semi-active suspension system its control performance is not satisfactory. This article describes respectively control theory of ceiling control algorithm and PID control algorithm in the semi-active suspension system, while detailed describes control structure and control theory of the neural network algorithm, on this foundation this paper expatiate semi-active suspension controller design on the IPSO-BP algorithm. The controller uses a dual neural network and indirect adaptive control structure, in the controller, a network NNI is used for on-line identification of the model of semi-active suspension, another network NNC is used to control the semi-active suspension, in the control process, NNI will continue to update themselves in order to identify the model of semi-active suspension, and recognition results are transmitted to the control network as a reference of the NNC, the NNC is according to road conditions, working conditions for suspension, the vibration of mass on the spring, the vibration of mass under the spring and the results of NNI for on-line control of semi-active suspension to adjust in order to make vehicles in the process achieve the desired ride and handling. The key problem of IPSO-BP algorithm is real-time issues, due to the control principle of controllers based on digital control, every one control instruction is completed in the interval of signal sampling. For road vehicles in general incentives on the road is 20Hz or less, according to Shannon's theorem: the sampling frequency is twice greater than the signal frequency that can be sampled more than 40Hz. Assuming a frequency of 100Hz, the sampling interval is 10ms, which is sufficient to complete the relevant calculation to meet the control real-time .According to the working principle and performance of variable damping shock absorbers, the paper proposes control strategies for IPSO-BP algorithm. When the direction of the speed of mass on the spring is the same with the direction of the relative speed of suspension, damping coefficient of the semi-active suspension is increased. When the direction of the speed of mass on the spring and the direction of the relative speed of suspension are the opposite, damping coefficient of the semi-active suspension is reduced. Accordingly, the same half of the semi-active suspension as controlled object, the ceiling control, PID control, IPSO-BP control, are taken to compared simulation test, the road use B-road, by comparing the experimental data ,it shows that IPSO - BP algorithms have greater advantages than other algorithms, to make the vehicle ride and handling great improvement.5,Hardware- in- the- Loop test platformIn order to verify the control effect of IPSO-BP algorithm, on the base of the controller structure and control model the dSPACE-the-loop simulation is established, which can simulate the actual operational status of vehicles indoors with high-precision, reduce product development cycle, saves the test cost. The dSPACE real-time simulation system was developed by dSPACE company, which is based on simulink real-time platform of rapid development and testing. It is divided into two parts, hardware and software, the hardware system is board with high-speed computing power based on DSP chip, which integrates a number of I/O interface to facilitate the development tests, the software is based on simulink and integrates a number of modules to facilitate rapid development, the system can make the developers focus on the control system algorithm, not spend too much experience for connection of the test equipment, provides a good test environment for product development. In this paper, dSPACE-the-loop simulation platform include: dSPACE real-time systems, client, ECU, expansion card, and matlab / simulink software. In the simulink, after establishing a good IPSO-BP algorithm model the semi-active suspension control module is build, and have the corresponding instruction code generated by its host download to the dSPACE real-time system for on-line control. Through the experiment it shows that IPSO-BP algorithm effectively reduces vibration of the quality of spring the semi-active suspension system, increase comfort, enhanced traffic handling, and achieve the expected results set. The test provides an important reference for further study of semi-active suspension control system.