Application of Data Fusion Technology in Vehicle Dynamic Weighing System

    With the development of social economy and the expansion of economic regional division of labor, the traffic volume on highways has increased rapidly, especially the number of freight trucks has increased year by year. Some drivers have disregarded the carrying capacity of vehicles and highways and driving safety, and have arbitrarily modified vehicles to overload, causing serious damage to the highways. Traffic accidents caused by this are increasing day by day. Therefore, in order to protect the integrity and smooth flow of highways, it is imperative to strictly limit overloaded vehicles.
    The dynamic weighing system uses sensors to measure the force signals of dynamic tires during driving, and then analyzes these force signals to calculate the corresponding static vehicle weight. Unlike traditional static weighing of automobiles, dynamic weighing does not require the car to stop for weighing. As long as the car passes the weighing sensor, it can be weighed. Therefore, it is timely, concealed and continuous. It not only realizes weighing, but also does not affect normal traffic. It is a very valuable research topic.

1 Circuit Design
    The hardware structure block diagram of the vehicle weighing system is shown in Figure 1, pressure sensor, speed sensor, analog signal processing circuit, A/D conversion circuit, button, LCD display module, printer, PC. The main chips of the circuit include CPU T2368BI, A/D conversion chip AD7799, RS232 level converter chip SP3223EEY, LCD interface circuit and keyboard interface circuit.

    When the car passes through the pressure plate, the pressure sensor converts the pressure signal into an analog electrical signal, which is then sent to the analog input of the A/D converter through the instrument amplifier. The converted digital quantity is sampled and processed, and the processing result is sent to the LCD as the dynamic weighing weight for display. At the same time, the data is sent to the printer for printing, or the data can be sent to the computer through the RS232 level converter for storage or further processing.

2 Common data fusion theory
    Sensor data fusion refers to the multi-level, multi-faceted, and multi-level processing and integration of data from multiple sensors to obtain more accurate and reliable useful information. Compared with a system that only uses a single sensor, the information from multiple sensors is redundant, complementary, and related. Multi-sensor data fusion is to make full use of the resources of multiple sensors, and through the reasonable control and use of various sensors and their observation information, the complementary and redundant information of various sensors in space and time is combined according to a certain optimization criterion. At present, in foreign countries, multi-sensor fusion systems have been widely used in battlefield analysis and monitoring, ballistic missile defense, target tracking, robots, automatic cars, assisted driving, complex intelligent manufacturing systems, and nuclear power plant monitoring. According to the diversity of information collected by sensors, the multi-sensor information fusion of smart meters adopts three methods: related information fusion, complementary information fusion and collaborative information fusion.
2.1 Related Information Fusion
    Related information refers to the information about the same environmental features obtained by a group of sensors. For example, when detecting an object, multiple sensors can be placed in the same area, and the output information of these sensors is the related information about the detected object.
2.2 Complementary Information Fusion
    Complementary information is the description of the same object or environment from two or more independent sensors from different perspectives, and multiple information that does not repeat each other. The fusion of complementary information can give a more comprehensive and complete description of the object and environment: sometimes it can enable the multi-sensor system to perceive the object and environmental features that each single sensor cannot obtain.
2.3 Collaborative Information Fusion
    Collaborative information refers to the multi-source data information obtained by sensors in a multi-sensor system that is interdependent or coordinated with each other.
    In short, the goal of multi-sensor data fusion is to obtain more accurate information than any single sensor through combination, so as to achieve the best coordination between sensors, that is, to improve the performance of the entire sensor system through the advantages of coordination and complementary performance between multiple sensors.

3 Application of data fusion in vehicle dynamic weighing technology
    In the vehicle dynamic weighing system, two data fusion methods are mainly used, namely, multi-sensor data fusion of similar sensors on the weighing plate and heterogeneous sensor data fusion of pressure sensors and accelerometers. (Here, the optimal weighted average is used for data fusion).
3.1 Multi-data fusion of similar sensors
    Assume that the s-th proposition As (s = 1, 2, ..., K) obtained in the i-th sensor (i = 1, 2, ..., P) in Q measurement cycles, its single sensor multi-measurement cycle fusion posterior basic credibility distribution function is
   
3.2 Heterogeneous sensor data fusion
    Assuming that the measurement errors of the pressure sensor and the accelerometer are independent, zero-mean and constant-variance Gaussian distribution random variables, then
    x(t+1) = φ(t)x(t) + B(t)u(t) + Γ(t)w(t) (1)
    yi(t) = Hi(t)x(t) + Vi(t) (2)
    In the above formula, x(t)∈Rn is the pressure measurement value, i=1, 2…, l is the measurement result, u(t)∈Rp is the control input, w(t)∈Rr is the acceleration measurement value, Vi(t)∈Rmi, i=1, 2…, l and φ(t), B(t), Γ(t), Hi(t) are time-varying matrices. [page]
    The key to vehicle dynamic weighing is intelligent pressure detection. The intelligent pressure detection system using optimal weighted average for data fusion consists of four parts: sensor module, signal conditioning module, data fusion center module and display circuit. Its working process is: the sensor and signal conditioning module complete the detection and processing of the input signal, the fusion center integrates the information of each sensor, and performs corresponding data processing, and the final result is displayed by the display circuit. As shown in Figure 2.

    1) The sensor part outputs two voltage signals, of which U1 is the voltage output signal of the measured pressure P, and U2 is the detection voltage signal of a non-target parameter. For
an ideal pressure sensor, its output U should be a univariate function value of the input pressure, that is, U=f(P). Its inverse function is P=f(U).
    2) Fusion center. The fusion center adopts a data processing technology based on weighted average. Using the weighted average method for data fusion is actually to find the weighted average of the output data of each sensor. If the output of sensor i (i=1, 2…n) is xi, the mean square error of the measurement result is σi, the weight is Wi, and the data fusion result is y=WX=[w1, w2,…, wn][x1, x2,…, xn]T. If the weights are properly distributed, the fusion effect is better; if the distribution is unreasonable,
    the accuracy and reliability of the system will not be greatly improved. The criterion for optimal weight allocation is as follows:
   

4 Effect of data fusion processing
    In order to verify the effect of the multi-sensor data fusion method, this paper conducted a series of weighing measurement experiments: the initial value of the sensor weight was 4 tons, and then increased by 2 tons each time until it reached 22 tons. The comparison of the weight value error before and after data fusion is shown in the following table.

    The above experimental data show that: under the same temperature change and power supply fluctuation, the data error after data fusion is significantly reduced, and its error value is less than 1.39% of the static weighing value, and the nonlinear error is 1.16%. It can be seen that the fused value is closer to the theoretical value. Therefore, the multi-sensor data fusion technology based on the optimal weighted allocation principle can effectively reduce or eliminate the cross-interference of sensors during operation.

5 Conclusion
    According to the structure and characteristics of the vehicle dynamic weighing system, the author adopts several data fusion calculation methods to effectively improve the anti-interference ability of the vehicle dynamic weighing system and ensure the reliability and accuracy of the measurement. Through many experiments, the technical parameters of the dynamic weighing system designed in this paper are as follows: in static mode, the accuracy is higher than 20 kg; in dynamic mode, when the vehicle passes at a speed lower than 20 km/h, the error is less than 1.39% of the static weighing value, and the nonlinear error is 1.16% (axle load and total weight). The main features of this dynamic weighing system include light weight and easy to carry; it can be connected to external computer-aided equipment; it has two working modes, dynamic and static; it accurately measures the weight of dynamic vehicles; it automatically compares the measured vehicle weight with the stored data to determine the vehicle's net weight or whether the vehicle is overweight and other data; it has a fully automated weighing process; and the data is automatically stored for retrieval and statistics.