Research on Shifting of Electronically Controlled Automatic Transmission Based on Neural Network and Virtual Instrument


The automatic transmission shifting law refers to the law of the automatic shifting timing of the two shifting blocks with the control parameters, and is a nonlinear function describing the relationship between the current vehicle state parameters (such as throttle opening, vehicle speed, etc.) and the optimal gear position. There are two ways to obtain the shifting rules: First, learn the driving experience of an excellent driver and extract the best shifting rules. Secondly, according to the theory of automobile shifting, the objective function and the constraint conditions are optimized to obtain the relationship between the running state of the vehicle and the optimal gear position. M. However, no matter which method is adopted, the obtained shifting rules are some corresponding. Discrete data, if it is expressed in mathematical analytic form, it is not easy to do, and it is necessary to carry out more complicated data analysis and processing. However, for neural networks, the determination of the optimal gear position of the car is only a simple nonlinear mapping problem. . Furthermore, since the performance of the car will change during the use of the car, the shifting rules pre-stored in the ECU can not make the vehicle obtain the optimal shifting timing, and its adaptability to the change of the vehicle parameters is poor, and cannot be corrected online in time. It makes sense to use the neural network approach to solve the computational problem of the best gear. Using virtual instrument technology to collect data in real time, it is convenient to call the neural network tool to analyze and process the data, and directly obtain the result on the computer, without having to record and calculate the real-time and operability.
1 BP network-based automatic transmission shifting process modeling 1.1 Network structure and sample data At present, the automobile widely uses two-parameter shifting rules, that is, the throttle opening and the vehicle speed are used for joint control, and the throttle opening and the vehicle speed are used as the neural network. The input signal, a gear as an output signal, establishes a neural network model that reflects their relationship well. The input layer is 2 nodes, which is the throttle opening signal a and the vehicle speed signal v. The output layer is 1 node, which is the corresponding gear position 1 (the whole number) of the transmission. Because the output of the network is a real number, a simple step function is used. It can be converted to an integer. The composition of the artificial neural network whose shifting law is as shown in the figure [3. After setting the input parameters, it is necessary to formulate the corresponding training sample set of the input sample vector and the output target vector under different throttle opening and vehicle speed. The training samples of the network should be as much as possible. The data samples for training are shown in Table 1. Some training sample data are listed in the table. These sample data are used as input vectors of BP neural network. The excitation function of network hidden layer and output layer neural crest is logsig and purein, network training. The total error of the function of the trainlm system is 0. (1) 1. Training number throttle speed target gear position I expect output actual output Note: The given data is given in the table, the throttle is the voltage signal, the unit V, the tachometer 2 test sample Data No. Throttle Speed ​​Target Position I Expected Output Actual Output Note: The network with a speed of km/h is also required to be inspected. The partial inspection data is shown in Table 2. 12 Results and Analysis The actual output results from Tables 1 and 2 indicate The trained network can extract the law of shifting well and constitute the neural network model of the automatic transmission shift after the end of learning. According to the input data of different throttle opening and vehicle speed, there is always a value close to 1 in the output data of the system model, which represents the output gear position at this time, which coincides with the target gear position, thus obtaining the corresponding The best gear.
2 The implementation of automatic control based on virtual instrument technology is also very important. Under normal circumstances, from the collection of training samples to the simulation results, it is completed in several processes, and after the samples are collected, they must be input into the computer for analysis, and the efficiency and real-time performance are not high. Based on this situation, the author uses a more efficient, faster and more intelligent test program based on virtual instrument technology. See the program flow chart.
Program Flowchart Matlab has powerful signal analysis and data processing capabilities. The author also uses its neural network toolbox function to perform data calculation processing. However, Matlab cannot communicate with general instrument interface to realize data communication with signal acquisition equipment, and it is difficult to design good. Human-computer interface. As the most commonly used graphics development software for virtual instruments, Labview not only conveniently acquires signal data from data acquisition devices, but also has a simple and efficient program development process. The author combines the two, collects the throttle signal and the vehicle speed signal in real time, and sends it to the input of the established neural network model. The background processing is performed by the Mat-lab software called by Labview. The interface immediately displays the result, and the Labview can be used. Conveniently display, store and print data. The design of the instrument operation interface, such as the solenoid valve indicator light on and off is obtained by collecting signals through the data acquisition card, which can be used to determine the gear position of the vehicle. After the system inputs the throttle opening signal and the vehicle speed signal into the neural network for calculation, the output data is judged to be the calculated gear position, that is, the output position of the neural network. At the same time, the system also comes with a data logging module that records the acquired data and calculated data at any time. Experiments with a simple virtual instrument system designed show that it is very convenient in data acquisition and recording, combined with neural network to complete data calculation and processing, and the results are in line with the actual results.
3 Conclusions Using Labview virtual technology and BP neural network to test the automatic transmission shift law, the automatic transmission shifting rules are simpler in signal processing analysis and result display than other methods, the development cycle is short, and the transmission can be acquired. Multiple sensor signals, to understand its operating conditions, organically combine online detection and data analysis processing to make testing faster and more efficient. In short, the test system 1 designed by the combination of the neural network and the virtual instrument is actually recording the I instrument operation interface in the automatic blackening device I of the car for the above-mentioned research process.

Paddle & Paddle leashes

Paddle & Paddle leashes,High Quality Paddle & Paddle leashes,Paddle & Paddle leashes Details, CN

NINGBO MARINE OUTDOOR CO., LTD , http://www.marineoutdoors.com