CN107229801B - On-line Identification Method of Tire Rolling Resistance Coefficient - Google Patents

Abstract

The invention relates to an online identification method for tire rolling resistance coefficient. Firstly, the identification model is established by using the driving equation, and then the online identification of the rolling resistance coefficient is realized by combining with the online cluster identification algorithm. The establishment of the identification model integrates the original driving equation and the differential driving equation, which eliminates the influence of the vehicle mass and makes up for the shortcomings of the previous need to rely on the vehicle mass to calculate the rolling resistance coefficient. The measurement of tire rolling resistance coefficient by traditional technology is obtained through sliding test, which is limited by a single test environment and cannot adapt to the complex working conditions of vehicle driving. The online identification algorithm established by the invention can achieve online acquisition of important parameters of the whole vehicle, and adapt to different vehicle operation states and road environments.

Description

On-line Identification Method of Tire Rolling Resistance Coefficient

technical field

The invention relates to an on-line identification method of tire rolling resistance coefficient in automobile automatic control technology, in particular to an on-line identification method of tire rolling resistance coefficient using longitudinal acceleration sensor information and vehicle driving information without requiring the quality parameters of the whole vehicle.

Background technique

Tire rolling resistance coefficient is an important parameter for vehicle operation economy and dynamic control, and tire rolling resistance has a great influence on the longitudinal force of the whole vehicle. With the development of automatic control technology, many vehicle control parameters can be identified online. However, the rolling resistance coefficient is an abstract coefficient that simplifies and expresses the rolling resistance as a proportional relationship with the mass of the whole vehicle, and it is not easy to establish the dynamics or to express the physical model calculation of its generation mechanism.

Under the condition of the whole vehicle, the commonly used tire rolling resistance coefficient measurement method is the sliding test, which is calculated according to the driving resistance and air resistance work under the premise of cutting off the power output. The field test has many environmental limitations, such as temperature, road conditions, etc. The rolling resistance coefficient measured for a specific environment cannot express the real rolling resistance under variable working conditions. The establishment of an online tire rolling resistance coefficient identification method for vehicle control is to get rid of the harsh environmental restrictions of the test site, collect some available driving data during the driving process to identify the rolling resistance coefficient in real time, and enable the automatic control of the vehicle to achieve Adaptation of important parameters.

The traditional vehicle fixed parameter identification often uses the least squares algorithm. However, when the real sampled data has a large error or is not Gaussian white noise, the least squares algorithm will produce large errors.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an online identification method of rolling resistance coefficient in order to adapt to the change of rolling resistance coefficient with the change of vehicle use conditions and to achieve self-adaptation of important parameters of the whole vehicle, so as to realize the online acquisition of rolling resistance coefficient and adapt to Different working conditions and environments, improve the performance of the vehicle dynamic and economical control system.

The online clustering identification method of rolling resistance coefficient of the present invention is an online clustering identification model of rolling resistance coefficient established based on vehicle driving state information and vehicle longitudinal acceleration information, and includes the following steps:

S1. Model initialization

Fixed parameters required for loading the model, including vehicle parameters and algorithm parameters;

Vehicle parameters include: vehicle transmission efficiency η, tire rolling radius r, vehicle acceleration a _v , flywheel moment of inertia I _f , wheel moment of inertia I _w , air resistance coefficient C _D , vehicle windward area A, air density ρ, Gravitational acceleration g;

The algorithm parameters include: the number of clusters m, m initial cluster centers

S2. Collect vehicle driving status information

The CAN bus information that needs to be collected synchronously at each sampling moment includes: vehicle speed v, engine speed n, clutch pedal signal, brake pedal signal and the acceleration a _sen provided by the longitudinal acceleration sensor;

S3. Determine whether it is the neutral gear sliding data, if so, continue to perform the subsequent steps; if not, return to step S2 to collect the vehicle driving state information at the next moment;

S4. Calculate the vehicle mass expression

First, calculate the sliding resistance of the vehicle in the longitudinal direction, namely F _wjw = -F _w -F _jw ,

为空气阻力；

为车轮惯性力；a_v为车辆行驶加速度，是车速的导数；where:

is air resistance;

is the inertial force of the wheel; a _v is the acceleration of the vehicle, which is the derivative of the vehicle speed;

Then use the differential method to calculate the vehicle mass expression: first calculate the differential component ΔF _wjw of the sliding resistance, the differential component Δa _sen of the acceleration sensor and the differential component Δv ² of the square value of the vehicle speed, and then establish the vehicle mass expression according to the differential driving equation is: M=ΔF _wjw /Δa _sen ;

S5. Preliminary results of calculating rolling resistance coefficient

最小二乘的输入量为X＝ΔF_wjwg，输出量为Y＝F_wjwΔa_sen-a_senΔF_wjw；Bring the vehicle mass expression into the driving equation without driving force, and use the least squares algorithm with forgetting factor to estimate the initial result of the rolling resistance coefficient at the sampling time (i)

The input quantity of the least squares is X=ΔF _wjw g, and the output quantity is Y=F _wjw Δa _sen -a _sen ΔF _wjw ;

S6. Use online K-means clustering for the preliminary identification results to update the cluster center

待聚类点归类至距离最短的一类；更新输入量的聚类中心

其中F^m(i)和F^m(i-1)分别是当前采样时刻以及前一采样时刻第m类的聚类中心。Calculate the distance from the preliminary results of the rolling resistance coefficient to the centers of various clusters

The points to be clustered are classified into the class with the shortest distance; the cluster center of the input quantity is updated

where F ^m (i) and F ^m (i-1) are the cluster centers of the mth class at the current sampling time and the previous sampling time, respectively.

S7. Calculate the proportion of various types of data, and determine whether the data at the current sampling time is the category with the largest proportion of data; if so, perform subsequent steps, if not, return to step S2 to re-collect the driving state information;

S8. Use the least squares algorithm to further identify the rolling resistance coefficient;

S9. Calculate the amount of each type of data, and determine whether the termination condition is met. When the class with the largest amount of data meets the set amount, the algorithm terminates.

The invention establishes an online cluster identification model of rolling resistance coefficient based on vehicle driving information and longitudinal acceleration sensor information. Using the method of integrating the differential longitudinal dynamics formula and the original longitudinal dynamics formula, the influence of the vehicle mass on the identification is eliminated, and an online estimation algorithm of the rolling resistance coefficient is established. The established model has the advantage of adapting to complex working conditions, and the online K-means clustering algorithm can effectively eliminate the influence of bad data on the identification results. The invention designs an online identification method of rolling resistance coefficient which combines the least squares algorithm and online K-means clustering, which can effectively solve the problem that the non-Gaussian noise distribution has an adverse effect on the identification result.

The on-line identification method for rolling resistance coefficient of the invention can obtain stable and reliable rolling resistance coefficient under complex working conditions such as different vehicle qualities and changes in road environment, and is helpful for improving the performance of the vehicle dynamic and economical control system.

Description of drawings

Fig. 1 is the schematic flow chart of the rolling resistance coefficient identification method of the present invention;

Detailed ways

The content of the present invention can be further understood through further detailed description of the following embodiments, but it is not intended to specifically limit the present invention.

Example 1

1, an online identification method for rolling resistance coefficient is a rolling resistance coefficient identification model established based on driving information and vehicle longitudinal acceleration information, including the following steps:

Step S1: Model initialization. Fixed parameters required for loading the model, including vehicle parameters and algorithm parameters.

Vehicle parameters include: vehicle transmission efficiency η, tire rolling radius r, vehicle acceleration a _v , flywheel moment of inertia I _f , wheel moment of inertia I _w , air resistance coefficient C _D , vehicle windward area A, air density ρ, Gravitational acceleration g. The algorithm parameters include: the number of clusters m, m initial cluster centers

Step S2: Collect vehicle driving state information.

The CAN bus information that needs to be collected synchronously at each sampling moment includes: vehicle speed v, engine speed n, clutch pedal signal, brake pedal signal and the acceleration a _sen provided by the longitudinal acceleration sensor.

Step S3: judging whether it is neutral sliding data. If yes, continue to perform the subsequent steps, if not, return to S2 to collect vehicle driving state information at the next moment.

Step S4: Calculate the vehicle mass expression.

First calculate the sliding resistance. The longitudinal force balance equation of the whole vehicle is used in the derivation of the sliding resistance expression. The longitudinal force balance equation of the vehicle is:

F _t =F _f +F _w +F _i +F _j (1)

为空气阻力；

为汽车驱动力；

为车轮加速阻力；

为飞轮加速阻力；

为变速器传动比与主减速器传动比的乘积；a_v是车辆行驶加速度，是车速的导数。接下来将加速阻力改写为：in,

is air resistance;

for the driving force of the car;

resistance for wheel acceleration;

Acceleration resistance for the flywheel;

is the product of the transmission ratio and the final gear ratio; a _v is the driving acceleration of the vehicle, which is the derivative of the vehicle speed. Next, rewrite the acceleration resistance as:

F _j =F _ja +F _jw +F _jf (2)

F_jf为飞轮转动加速阻力

Among them, F _ja is the vehicle translational acceleration resistance (F _ja =ma _v ); F _jw is the wheel rotational acceleration resistance

F _jf is the acceleration resistance of flywheel rotation

Bring the expression of (2) into (1) and organize it as:

F _t =F _f +F _w +F _i +F _ja +F _jw +F _jf (3)

According to the longitudinal force balance equation of the whole vehicle, there is no driving force and inertial force generated by the engine flywheel during the neutral sliding process, and finally the sliding resistance expression can be obtained: F _wjw = -F _w -F _jw .

On the other hand, the measurement value of the acceleration sensor is defined as: a _sen =gi+ _av , where a _sen is the acceleration value (unit m/s ² ) collected by the acceleration sensor. According to the definition of the acceleration sensor, the rolling resistance coefficient identification model containing the information of the acceleration sensor can be obtained:

F _wjw = m(gf+a _sen ) (4)

Then use the differential driving equation to calculate the vehicle mass expression. First calculate the differential component ΔF _res of the sliding resistance, the differential component Δa _sen of the acceleration sensor and the differential component Δv ² of the square value of the vehicle speed, and then establish the vehicle mass expression according to the differential driving equation: M=ΔF _wjw /Δa _sen

Step S5: Calculate the preliminary result of the rolling resistance coefficient.

最小二乘的输入量为X＝ΔF_wjwg，输出量为Y＝F_wjwΔa_sen-a_senΔF_wjw Bring the vehicle mass expression into the driving equation without driving force, and use the least squares algorithm with forgetting factor to estimate the initial result of the rolling resistance coefficient at the sampling time (i)

The input quantity of the least squares is X=ΔF _wjw g, and the output quantity is Y=F _wjw Δa _sen -a _sen ΔF _wjw

Step S6: Use online K-means clustering on the preliminary identification result to update the cluster center.

待聚类点归类至距离最短的一类。更新输入量的聚类中心

其中F^m(i)和F^m(i-1)分别是当前采样时刻以及前一采样时刻第m类的聚类中心。Calculate the distance from the preliminary results of the rolling resistance coefficient to the centers of various clusters

The points to be clustered are classified into the class with the shortest distance. Update the cluster centers of the input

where F ^m (i) and F ^m (i-1) are the cluster centers of the mth class at the current sampling time and the previous sampling time, respectively.

Step S7: Calculate the proportions of various types of data, and determine whether the data at the current sampling time is the category with the largest proportion of data. If yes, perform the following steps, if not, return to S2 to collect the driving state information again.

Step S8: The rolling resistance coefficient is further identified. The rolling resistance coefficient is further identified by the least squares algorithm.

Step S9: Calculate the amount of various types of data, and determine whether the termination condition is satisfied. The discriminant algorithm terminates when the class with the largest amount of data satisfies a certain number. The recommended data quantity in this embodiment is 2000.

Claims (1)

1. An online identification method for a tire rolling resistance coefficient is an online clustering identification model for the rolling resistance coefficient established based on vehicle running state information and vehicle-mounted longitudinal acceleration information, and is characterized by comprising the following steps:

s1. model initialization

Loading fixed parameters required by the model, wherein the fixed parameters comprise vehicle parameters and algorithm parameters;

the vehicle-finishing parameters comprise vehicle transmission efficiency η, tire rolling radius r and vehicle running acceleration a_vFlywheel moment of inertia I_fWheel moment of inertia I_wAir resistance coefficient C_DThe windward area A of the whole vehicle, the air density rho and the gravity acceleration g;

the algorithm parameters include: number of clusters m, m initial cluster centers

S2, collecting vehicle running state information

The CAN bus information that needs to be synchronously acquired at each sampling moment includes: speed v of whole vehicle, engine speed n, clutch pedal signal, brake pedal signal and acceleration a provided by longitudinal acceleration sensor_sen；

S3, judging whether the data is neutral gear sliding data or not, and if so, continuing to execute the subsequent steps; if not, returning to the step S2 to collect the vehicle running state information at the next moment;

s4, calculating the whole vehicle mass expression

Firstly, calculating the longitudinal running sliding resistance of the whole vehicle, namely F_wjw＝-F_w-F_jw，

In the formula:

is the air resistance;

is the wheel inertia force; a is_vIs the vehicle travel acceleration, is the derivative of the vehicle speed;

and then calculating the expression of the mass of the whole vehicle by using a difference method: first, a difference amount Δ F of the sliding resistance is calculated_wjwDifferential quantity of acceleration sensor Deltaa_senDifference Δ v from the square value of vehicle speed²Then, establishing a whole vehicle mass expression according to the differential running equation as follows: m ═ Δ F_wjw/Δa_sen；

S5, calculating a rolling resistance coefficient preliminary result

Substituting the whole vehicle mass expression into a running equation without driving force, and estimating the initial result of the rolling resistance coefficient at the sampling moment i by using a least square algorithm with a forgetting factor

The input quantity of least square is X ═ Δ F_wjwg, output amount is Y ═ F_wjwΔa_sen-a_senΔF_wjw；

S6, utilizing the online K mean value clustering to the primary identification result to update the clustering center

Calculating the distance from the initial result of the rolling resistance coefficient to various cluster centers

Classifying the points to be clustered into a class with the shortest distance; updating cluster centers of input quantities

Wherein F^m(i) And F^m(i-1) clustering centers of the mth class at the current sampling moment and the previous sampling moment respectively;

s7, calculating the ratio of various types of data, and judging whether the data at the current sampling moment is the type with the largest data volume ratio; if yes, executing the subsequent steps, and if not, returning to the step S2 to collect the driving state information again;

s8, further identifying the rolling resistance coefficient by using a least square algorithm;

and S9, calculating various data volumes, judging whether a termination condition is met or not, and judging that the algorithm is terminated when the class with the largest data volume meets the set number.

Images