The three foundations of modern vehicle electronic information technology include: information acquisition (sensor technology), information transmission (Communication Technology) and information processing (computer technology). Sensors are the frontier products of modern vehicle electronic information technology.
Tire failure is one of the main causes of inducing and leading to traffic accidents. Accurate, reliable and fault-tolerant continuous monitoring of the state of tires, vehicles and roads is a complex work that is difficult to be accurately completed by traditional sensors. Virtual sensors based on computer technology can be used to measure the friction between roads and tires, road conditions Parameters such as tire pressure and tire rotation balance state. The virtual sensor is based on the sensor hardware and computer platform and developed through software. The software can be used to complete the calibration and calibration of the sensor and realize the control of the best performance index of the vehicle. The sensor is connected to the computer through the data collector, and the computer completes the sensor inspection, sensor parameter reading, sensor setting and recording.
2. Virtual sensor composition
The hardware part of vehicle tire virtual sensor mainly includes computer system, excellent wheel speed sensor, wheel pressure sensor, connector and instrument. The virtual sensor must also include its software package, i.e. data processing program, compensation program, test processing program, etc.
Virtual sensing is a signal processing technology used to estimate the response where physical sensors cannot be placed in the system. The study of vehicle tire virtual sensor can further study the factors affecting tire parameters such as the friction between road and tire, tire expansion pressure and wheel imbalance. The multi-sensor fusion system combines the data from each sensor to describe the observed system, so as to obtain more accurate and more specific conclusions than using a single sensor.
Using computer technology, the virtual sensor can only use ordinary sensors, namely wheel speed sensor, wheel torque sensor and vehicle can bus to transmit information, and then complete the calculation of vehicle tire friction model according to a certain computer algorithm, so as to obtain effective information about wheel pressure. The measured input and output are shown in Figure 1. At the beginning, the research mainly estimates the road friction coefficient, and then estimates the wheel pressure.
3. Pavement adhesion characteristics
One of the models of vehicle tire is based on the ground machine system mechanics theory, which shows that the adhesion coefficient produced by different ground conditions is different. The relationship between vehicle driving force, adhesion coefficient and vehicle tire slip rate is shown in Figure 2. It can be seen from the figure that when the adhesion coefficient is constant and the tire slip rate s = 0, the ratio F of driving force to vehicle weight φ/ G max, slip rate s from. At the beginning of the increase, the adhesion coefficient φ With the increase, the adhesion coefficient reaches the maximum when the tire slip rate s = 0.10 ~ 0.20 φ Max, if the slip rate continues to increase, the adhesion coefficient begins to decrease. When the tire slip rate s reaches 100%, the tire will slip completely.
The slip rate is usually defined as the ratio of the relative speed of the wheel to the tangential speed. The expression is formula (1). The adhesion coefficient is φ＝ F φ/ G。 When the slip rate is very small, the wheel slip rate and adhesion coefficient satisfy the linear equation φ= KS, which is related to the characteristics of tires and the friction of pavement.
The model shows that the accurate wheel slip ratio is a function of driving force μ Add a mean white noise, which is mainly caused by the different radius of the front and rear wheels.
The speed sensor measures the working speed on each tire ω， The torque M transmitted from the engine to each tire is obtained from the CAN bus. The slip rate and adhesion coefficient of each tire can be obtained by solving the model by computer.
4. Algorithm of measuring vehicle tire pressure with virtual sensor
At present, foreign countries have carried out a large number of research and development projects on the virtual sensor algorithm for tire pressure detection, of which more than 40 patents have been applied, most of which use the standard wheel speed sensor. The main methods are:
4.1 vibration analysis algorithm
Based on the elastic characteristics of tire rubber when impacted by road surface, the frequency spectrum of wheel speed is sampled and analyzed, and the disturbance caused by vibration caused by other parts of the vehicle is removed. The vibration analysis can be completed by the detection method of fast Fourier transform FFT.
The basic idea of this method is to regard the tire as a “spring damping” system. According to the tire model and inflation pressure, the ideal vibration frequency is generally between 40-50hz, but there are other higher or lower frequencies. By studying and screening the vibration of different tires, the vibration frequency is monitored, so as to obtain the value of normal frequency and detect abnormal values.
Whether based on model or FFT method, the velocity measurement must be filtered before applying vibration analysis. Using the fitting least square method, a velocity signal with improved interference ratio is obtained. The compensated speed signal waveform clearly shows a vibration mode of about 45Hz, as shown in Fig. 3.
The average periodogram can be calculated by using FFT measuring instrument and a low pass rate filter, and the peak value shown in Fig. 4 can be automatically calculated in each test. After offline, 15% tire pressure drop can be detected, but considering the automatic threshold setting and low error warning rate, it is more reliable to detect 30% tire pressure drop. FFT is a batch processing of data with a certain lag time. A large number of 305 test data are used in Figure 4.
When a quadratic “damping spring” model is adopted, the relevant parameters for calculating tire pressure are
The model-based vibration analysis is easy to introduce the fitting method. The vibration mode obtained by the fitting least square method is shown in Fig. 5. Using this filter, we can reliably detect a 30% pressure drop within 5S. This filter can process data at the lowest frequency of 1oohz, and the amount of calculation is significantly lower than that of FFT method.
Using a Kalman filter operating only at 20Hz, it has stronger capability than RLS under the same amount of calculation. Fig. 6 under different tire pressures, the vibration estimation mode of Kalman filter has 5 times less computation than the original. The speed of 15% pressure change detected by this new method is the same as that detected by RLS algorithm.
4.2 tire pressure calculation based on tire radius
The most common tire pressure calculation is based on the difference of tire speeds with static nonlinear changes, which is close to zero when the wheel sizes are the same.
The slip offset of Kalman filter for road friction indication can well reflect the relative wheel radius. The measured data are linear, and this method is not sensitive to noise. The slip offset can detect the relative error of the left and right wheels, and the yaw rate filter can be used to calculate the relative error between the front and rear wheels, so as to calculate the wheel radius.
Most of the existing wheel pressure measurement methods are based on the static test of wheel speed ω 1／ ω 2= ω 3/ ω 4. The footmark is the code of each wheel (left front wheel 1, right front wheel 2, left rear wheel 3, right rear wheel 4). When the vehicle travels in a straight line or circle at a uniform speed, the value of the wheel speed equation of the vehicle is zero. The low transmittance form of this signal has an abnormal deviation of no less than 0, which exceeds the base point value that TPI can use. In addition, the unsatisfactory statistical state, insufficient bending strength, different friction levels, tire wear and so on caused by the nonlinear transformation of data limit the reliability of this method.
We find that the model-based method can take into account the above problems, and the performance is significantly improved. The basic filters used are road friction indication filter (RFI), high-precision yaw rate (HPY) and absolute speed indication (AVI) filter. The latter two are suitable for tire pressure measurement. See Table 1 for relevant information.
When all tire gas leaks out slowly, the deviation between the vehicle radius and the theoretical value can be completed with the diffusion function key of the detection instrument.
After field experiments, all the methods are compared. The conclusion is that the last method has the best performance in accuracy, fast response characteristics, strength and computational complexity (low-order model and low sampling frequency).
However, in certain cases, other methods can be a good supplement in some aspects, so the final algorithm system should be a combination of multiple algorithms.
The virtual sensor indirectly measures the information through the signals of other sensors to measure the adhesion coefficient and tire pressure between the road and tires, and the special sensor is very expensive. By accurately measuring the wheel speed signal of the vehicle, the friction and tire pressure can be obtained with the help of computer technology. The virtual sensor technology of adhesion coefficient is becoming more and more mature and began to be introduced into engineering application. The tire pressure virtual sensor technology is still in the research stage. The research methods focus on the virtual sensor technology based on vibration and the virtual sensor technology based on wheel radius model. However, for complex and changeable external interference factors, the correction information of CAN bus is required.
Responsible editor: GT