CN113570057B - Vehicle wheel center vertical displacement measuring method and device based on model training - Google Patents

Vehicle wheel center vertical displacement measuring method and device based on model training Download PDF

Info

Publication number
CN113570057B
CN113570057B CN202111132180.5A CN202111132180A CN113570057B CN 113570057 B CN113570057 B CN 113570057B CN 202111132180 A CN202111132180 A CN 202111132180A CN 113570057 B CN113570057 B CN 113570057B
Authority
CN
China
Prior art keywords
signal
wheel center
simulation
model
center vertical
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111132180.5A
Other languages
Chinese (zh)
Other versions
CN113570057A (en
Inventor
丁鼎
韩广宇
张永仁
卢放
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Lantu Automobile Technology Co Ltd
Original Assignee
Lantu Automobile Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lantu Automobile Technology Co Ltd filed Critical Lantu Automobile Technology Co Ltd
Priority to CN202111132180.5A priority Critical patent/CN113570057B/en
Publication of CN113570057A publication Critical patent/CN113570057A/en
Application granted granted Critical
Publication of CN113570057B publication Critical patent/CN113570057B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model

Abstract

The invention relates to the technical field of wheel center parameter measurement, in particular to a method and a device for measuring the vertical displacement of a wheel center of a vehicle based on model training. The method comprises the following steps: obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the simulation device comprises a wheel center vertical displacement simulation signal, and the output of the simulation device comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal which correspond to the wheel center vertical displacement simulation signal; constructing an initial neural network model; and (3) iteratively training the initial neural network model to obtain a target neural network model for measuring the vertical displacement of the wheel center of the vehicle. The invention utilizes the neural network model to measure the vertical displacement of the wheel center of the vehicle, reduces the measurement cost, takes the vertical acceleration signal of the wheel center and the spring signal into consideration as calculation input in the measurement, reduces the calculation error of the vertical acceleration of the wheel center in a low-frequency interval, improves the measurement precision, and thus measures the vertical displacement of the wheel center of the vehicle with low cost and high precision.

Description

Vehicle wheel center vertical displacement measuring method and device based on model training
Technical Field
The invention relates to the technical field of wheel center parameter measurement, in particular to a method and a device for measuring the vertical displacement of a wheel center of a vehicle based on model training.
Background
The wheel center vertical displacement signal of the vehicle in the driving process is an important signal of the vehicle, but the wheel center vertical displacement signal of the vehicle cannot be directly measured and obtained through a single sensor at present, combined measurement of a potentiometer sensor, a laser sensor and the like is needed, the cost is high, and the accuracy of a measuring result is not high.
Therefore, how to measure the vertical displacement of the wheel center of the vehicle with low cost and high precision is a technical problem which needs to be solved at present.
Disclosure of Invention
The invention aims to provide a method and a device for measuring the vertical displacement of a wheel center of a vehicle based on model training, which can measure the vertical displacement of the wheel center of the vehicle with low cost and high precision.
In order to achieve the above object, the embodiments of the present invention provide the following solutions:
in a first aspect, an embodiment of the present invention provides a model training method for measuring a vertical displacement of a wheel center of a vehicle, where the method includes:
obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal;
constructing an initial neural network model;
and inputting the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the wheel center vertical displacement of the vehicle by taking the minimum error between the initial wheel center vertical displacement measurement signal output by the initial neural network model and the wheel center vertical displacement simulation signal as a target.
In a possible embodiment, the obtaining a full vehicle multi-body simulation model of the target type vehicle includes:
obtaining an initial whole vehicle multi-body simulation model of the target type vehicle;
inputting the wheel center vertical displacement simulation signal into the initial whole vehicle multi-body simulation model for simulation to obtain a simulation output signal of the initial whole vehicle multi-body simulation model;
driving the target type vehicle based on the wheel center vertical displacement actual signal, and obtaining a test sensing signal through a whole vehicle sensor group on the target type vehicle; wherein, whole car sensor group includes: one or more of a wheel center acceleration sensor, a shock absorber mounting point acceleration sensor, a spring displacement sensor and a transverse stabilizer bar strain sensor;
and optimizing the initial finished automobile multi-body simulation model into the finished automobile multi-body simulation model based on the test sensing signal and the simulation output signal.
In a possible embodiment, after obtaining the initial full-vehicle multi-body simulation model of the target type vehicle, the method further includes: controlling the initial whole vehicle multi-body simulation model and the target type vehicle to be in the same whole vehicle test simulation working condition;
based on the test sensing signal and the simulation output signal, the initial finished automobile multi-body simulation model is optimized into the finished automobile multi-body simulation model, and the method specifically comprises the following steps:
taking one or more expressions that the test sensing signals and the simulation output signals meet optimization calibration criteria as targets, adjusting modeling parameters of the initial finished automobile multi-body simulation model, and obtaining the finished automobile multi-body simulation model; wherein the expression of the optimized calibration criterion is:
Figure 770931DEST_PATH_IMAGE001
；
RMS is a root mean square value regularization solving function;
Figure 829017DEST_PATH_IMAGE002
simulating signals for vertical acceleration of the wheel center;
Figure 671071DEST_PATH_IMAGE003
is a wheel center vertical acceleration test signal;
Figure 800701DEST_PATH_IMAGE004
is a first error threshold;
Figure 162544DEST_PATH_IMAGE005
simulating a signal for the displacement of the spring;
Figure 47323DEST_PATH_IMAGE006
is a spring displacement test signal;
Figure 924143DEST_PATH_IMAGE007
is a second error threshold;
Figure 919781DEST_PATH_IMAGE008
simulating signals for vertical acceleration of upper points of shock absorbers；
Figure 401709DEST_PATH_IMAGE009
Vertical acceleration test signals of points on the shock absorber are obtained;
Figure 129494DEST_PATH_IMAGE010
is a third error threshold;
Figure 946140DEST_PATH_IMAGE011
strain simulation signals of the transverse stabilizer bar are obtained;
Figure 230622DEST_PATH_IMAGE012
a transverse stabilizer bar strain test signal;
Figure 629373DEST_PATH_IMAGE013
is a fourth error threshold.
In a possible embodiment, the constructing the initial neural network model includes:
constructing an input layer; wherein the output signal matrix of the input layer
Figure 121535DEST_PATH_IMAGE014
The wheel center vertical acceleration simulation signal and the spring displacement simulation signal are included;
constructing a first hidden layer; wherein the propagation expression of the first hidden layer is:
Figure 972947DEST_PATH_IMAGE015
；
wherein the content of the first and second substances,
Figure 310388DEST_PATH_IMAGE016
a matrix of output signals for the first hidden layer;
Figure 501329DEST_PATH_IMAGE017
a first weight matrix for the input layer to the first hidden layer;
Figure 570916DEST_PATH_IMAGE018
a first bias matrix for the input layer to the first hidden layer;
constructing a second hidden layer; the propagation expression of the second hidden layer is as follows:
Figure 96575DEST_PATH_IMAGE019
；
wherein the content of the first and second substances,
Figure 785177DEST_PATH_IMAGE020
a matrix of output signals for the second hidden layer;
Figure 79892DEST_PATH_IMAGE021
a second weight matrix for the first hidden layer to the second hidden layer;
Figure 585959DEST_PATH_IMAGE022
a second bias matrix for the first hidden layer to the second hidden layer;
constructing an output layer; the propagation expression of the output layer is as follows:
Figure 84068DEST_PATH_IMAGE023
；
wherein the content of the first and second substances,
Figure 825628DEST_PATH_IMAGE024
outputting a wheel center vertical displacement measurement signal for the output layer;
Figure 725582DEST_PATH_IMAGE025
a third weight matrix for the second hidden layer to the output layer;
Figure 402551DEST_PATH_IMAGE026
a third bias matrix for the second hidden layer to the output layer.
In one possible embodiment, the obtaining a target neural network model for measuring vertical displacement of the wheel center of the vehicle includes:
iterative training with the goal of satisfying the minimum of the error function Emin
Figure 902802DEST_PATH_IMAGE027
、
Figure 870889DEST_PATH_IMAGE028
、
Figure 936934DEST_PATH_IMAGE029
、
Figure 332275DEST_PATH_IMAGE030
、
Figure 116560DEST_PATH_IMAGE031
And
Figure 888338DEST_PATH_IMAGE032
(ii) a The specific iterative training formula is as follows:
Figure 684256DEST_PATH_IMAGE033
；
wherein the content of the first and second substances,
Figure 171869DEST_PATH_IMAGE034
simulating a signal for the vertical displacement of the wheel center;
Figure 397445DEST_PATH_IMAGE035
the number of times of iterative training is accumulated;
Figure 159865DEST_PATH_IMAGE036
is a learning efficiency parameter.
In a second aspect, an embodiment of the present invention provides a model training apparatus, where the apparatus includes:
the first establishing module is used for obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal;
the second establishing module is used for establishing an initial neural network model;
and the modeling module is used for inputting the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the vehicle wheel center vertical displacement by taking the minimum error between the initial wheel center vertical displacement measurement signal and the wheel center vertical displacement simulation signal output by the initial neural network model as a target.
In a possible embodiment, the first establishing module includes:
the first acquisition module is used for acquiring an initial whole vehicle multi-body simulation model of the target type vehicle;
the second acquisition module is used for inputting the wheel center vertical displacement simulation signal into the initial finished automobile multi-body simulation model for simulation to obtain a simulation output signal of the initial finished automobile multi-body simulation model;
the third acquisition module is used for driving the target type vehicle based on the wheel center vertical displacement actual signal and acquiring a test sensing signal through a whole vehicle sensor group on the target type vehicle; wherein, whole car sensor group includes: one or more of a wheel center acceleration sensor, a shock absorber mounting point acceleration sensor, a spring displacement sensor and a transverse stabilizer bar strain sensor;
and the fourth acquisition module is used for optimizing the initial finished automobile multi-body simulation model into the finished automobile multi-body simulation model based on the test sensing signal and the simulation output signal.
In a possible embodiment, the apparatus further comprises:
the first control module is used for controlling the initial whole vehicle multi-body simulation model and the target type vehicle to be in the same whole vehicle test simulation working condition after the initial whole vehicle multi-body simulation model of the target type vehicle is obtained;
the fourth obtaining module specifically includes:
the optimization adjustment module is used for adjusting modeling parameters of the initial finished automobile multi-body simulation model by taking one or more expressions of the test sensing signal and the simulation output signal meeting optimization calibration criteria as targets to obtain the finished automobile multi-body simulation model; wherein the expression of the optimized calibration criterion is:
Figure 403764DEST_PATH_IMAGE037
；
RMS is a root mean square value regularization solving function;
Figure 406486DEST_PATH_IMAGE038
simulating signals for vertical acceleration of the wheel center;
Figure 103047DEST_PATH_IMAGE039
is a wheel center vertical acceleration test signal;
Figure 810103DEST_PATH_IMAGE040
is a first error threshold;
Figure 659241DEST_PATH_IMAGE041
simulating a signal for the displacement of the spring;
Figure 426340DEST_PATH_IMAGE042
is a spring displacement test signal;
Figure 875776DEST_PATH_IMAGE043
is a second error threshold;
Figure 979998DEST_PATH_IMAGE044
simulating signals for the vertical acceleration of the upper point of the shock absorber;
Figure 683643DEST_PATH_IMAGE045
vertical acceleration test signals of points on the shock absorber are obtained;
Figure 543015DEST_PATH_IMAGE046
is a third error threshold;
Figure 105846DEST_PATH_IMAGE047
strain simulation signals of the transverse stabilizer bar are obtained;
Figure 623546DEST_PATH_IMAGE048
a transverse stabilizer bar strain test signal;
Figure 493282DEST_PATH_IMAGE049
is a fourth error threshold.
In a possible embodiment, the second establishing module includes:
a first construction module for constructing an input layer; wherein the output signal matrix of the input layer
Figure 250849DEST_PATH_IMAGE014
The wheel center vertical acceleration simulation signal and the spring displacement simulation signal are included;
the second construction module is used for constructing the first hidden layer; wherein the propagation expression of the first hidden layer is:
Figure 612561DEST_PATH_IMAGE050
；
wherein the content of the first and second substances,
Figure 137214DEST_PATH_IMAGE051
a matrix of output signals for the first hidden layer;
Figure 64719DEST_PATH_IMAGE052
a first weight matrix for the input layer to the first hidden layer;
Figure 282205DEST_PATH_IMAGE053
a first bias matrix for the input layer to the first hidden layer;
the third construction module is used for constructing a second hidden layer; the propagation expression of the second hidden layer is as follows:
Figure 6578DEST_PATH_IMAGE054
；
wherein the content of the first and second substances,
Figure 318611DEST_PATH_IMAGE055
a matrix of output signals for the second hidden layer;
Figure 116934DEST_PATH_IMAGE056
a second weight matrix for the first hidden layer to the second hidden layer;
Figure 489009DEST_PATH_IMAGE057
a second bias matrix for the first hidden layer to the second hidden layer;
the fourth construction module is used for constructing an output layer; the propagation expression of the output layer is as follows:
Figure 372783DEST_PATH_IMAGE058
；
wherein the content of the first and second substances,
Figure 488506DEST_PATH_IMAGE059
outputting a wheel center vertical displacement measurement signal for the output layer;
Figure 62707DEST_PATH_IMAGE060
a third weight matrix for the second hidden layer to the output layer;
Figure 90837DEST_PATH_IMAGE061
a third bias matrix for the second hidden layer to the output layer.
In one possible embodiment, the modeling module includes:
a training module for iterative training with the objective of satisfying the minimum error function Emin
Figure 242333DEST_PATH_IMAGE062
、
Figure 912479DEST_PATH_IMAGE063
、
Figure 403504DEST_PATH_IMAGE064
、
Figure 789486DEST_PATH_IMAGE065
、
Figure 241327DEST_PATH_IMAGE066
And
Figure 980744DEST_PATH_IMAGE067
(ii) a The specific iterative training formula is as follows:
Figure 77007DEST_PATH_IMAGE068
；
wherein the content of the first and second substances,
Figure 368311DEST_PATH_IMAGE069
simulating a signal for the vertical displacement of the wheel center;
Figure 776290DEST_PATH_IMAGE070
the number of times of iterative training is accumulated;
Figure 709610DEST_PATH_IMAGE071
is a learning efficiency parameter.
In a third aspect, an embodiment of the present invention provides a method for measuring a vertical displacement of a wheel center of a vehicle, where the method includes:
acquiring a wheel center vertical acceleration actual measurement signal and a spring displacement actual measurement signal of the whole vehicle in the running process of the whole vehicle;
inputting the measured wheel center vertical acceleration signal and the measured spring displacement signal into the target neural network model in the model training method according to any one of the first aspect, and obtaining the measured wheel center vertical displacement signal of the whole vehicle.
In a fourth aspect, an embodiment of the present invention provides a vehicle wheel center vertical displacement measuring apparatus, where the method includes:
the acquiring unit is used for acquiring a wheel center vertical acceleration actual measurement signal and a spring displacement actual measurement signal of the whole vehicle in the running process of the whole vehicle;
and the processing unit is used for inputting the measured wheel center vertical acceleration signal and the measured spring displacement signal into the target neural network model in the model training method in the first aspect, and acquiring the measured wheel center vertical displacement signal of the whole vehicle.
In a fifth aspect, an embodiment of the present invention provides an electronic device, including:
a memory for storing a computer program;
a processor for executing the computer program to implement the steps of the method of any one of the first aspect; or for executing the computer program to perform the steps of the method of any of the third aspects.
In a sixth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, the computer program being executed by a processor to implement the steps of the method according to any one of the first aspect; or for executing the computer program to perform the steps of the method of any of the third aspects.
Compared with the prior art, the invention has the following advantages and beneficial effects:
firstly, establishing a whole vehicle multi-body simulation model; the input of the simulation device comprises a wheel center vertical displacement simulation signal, and the output of the simulation device comprises a wheel center vertical acceleration simulation signal and a damper spring displacement simulation signal; then establishing an initial neural network model; and finally, training by utilizing multiple groups of input quantities and output quantities corresponding to the whole vehicle multi-body simulation model to obtain a target neural network model for measuring the vertical displacement of the wheel center of the vehicle. The invention utilizes the neural network model to measure the vertical displacement of the wheel center of the vehicle, reduces the measurement cost, takes the vertical acceleration signal of the wheel center and the spring signal into consideration as calculation input in the measurement, reduces the calculation error of the vertical acceleration of the wheel center in a low-frequency interval, improves the measurement precision, and thus measures the vertical displacement of the wheel center of the vehicle with low cost and high precision.
Furthermore, the common sensors are adopted to build the whole vehicle sensor group, and the whole vehicle multi-body simulation model is obtained through optimized calibration, so that the use of laser sensors with high original cost and the like is reduced, and the whole measurement cost is further reduced. Meanwhile, the corrected and optimized whole vehicle multi-body simulation model can provide a high-precision training sample for the training of the initial neural network model, so that the model training quality is further improved, and the measurement precision is improved.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present specification, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a model training method provided by an embodiment of the present invention;
fig. 2 is a schematic layout diagram of a sensor group of a whole vehicle according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an initial neural network model according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a model training apparatus according to an embodiment of the present invention;
FIG. 5 is a flow chart of a method for measuring vertical displacement of a wheel center of a vehicle according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a vehicle wheel center vertical displacement measurement device according to an embodiment of the present invention.
Description of reference numerals: 1 is wheel center acceleration sensor, 2 is bumper shock absorber mounting point acceleration sensor, 3 is spring displacement sensor, and 4 is stabilizer bar strain transducer.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by those skilled in the art based on the embodiments of the present invention belong to the scope of protection of the embodiments of the present invention.
Referring to fig. 1, fig. 1 is a flowchart of a model training method according to an embodiment of the present invention, which specifically includes steps 11 to 13.
And 11, obtaining a whole vehicle multi-body simulation model of the target type vehicle.
The input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal.
Specifically, the target type vehicle may be a vehicle obtained by setting a type classification rule based on the displacement, model, chassis type, and the like of the vehicle.
Specifically, the spring displacement simulation signal is used for representing the displacement of a suspension spring in a finished automobile multi-body simulation model.
Specifically, the wheel center vertical displacement simulation signal may be a section of wheel center vertical displacement simulation signal y set at randomtestWith a signal amplitude of
Figure 660380DEST_PATH_IMAGE072
The signal frequency bandwidth is 0 Hz-F (F is a set frequency upper limit), and the signal time length is t. The signal is used as an input signal of a multi-body simulation model of the whole vehicle, so that a simulation output signal can be obtained: a. thewheel1Wheel center vertical acceleration simulation signal, Dspring1Spring displacement simulation signal, Atop1Vertical acceleration simulation signal and sigma of shock absorber mounting point1A stabilizer bar strain simulation signal, etc.
Specifically, the whole vehicle multi-body simulation model can comprise a plurality of simulation components of the whole vehicle, such as a front suspension, a rear suspension, a power assembly, steering, braking, rigid wheels, a vehicle body and the like, and can simulate and feed back physical quantities such as stress, offset and the like among all components of the whole vehicle under various working conditions.
And step 12, establishing an initial neural network model.
Specifically, the type of the initial neural network model may be a feedback neural network model, a deep learning neural network model, a convolutional neural network model, and the like, which is not limited herein, and the operation of step 13 may be completed according to the type of the initial neural network model.
Specifically, the initial neural network model may be understood as an untrained target neural network model, and may output an initial wheel center vertical displacement measurement signal by calculating the input wheel center vertical acceleration simulation signal and the spring displacement simulation signal through a neural network.
In general, the initial neural network model may include an input layer, a hidden layer, and an output layer, where the hidden layer is responsible for related calculation of the neural network, and through iterative training, it may gradually adjust weight parameters and other related transfer function parameters in the hidden layer, so that an initial wheel center vertical displacement measurement signal output by the initial neural network model conforms to a set training target, and at this time, the initial neural network model may be regarded as a target neural network model, and an initial wheel center vertical displacement measurement signal output by the initial neural network model may be regarded as a wheel center vertical displacement measurement signal.
And step 13, inputting the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the vehicle wheel center vertical displacement by taking the minimum error between the initial wheel center vertical displacement measurement signal and the wheel center vertical displacement simulation signal output by the initial neural network model as a target.
Specifically, by using the finished automobile multi-body simulation model obtained in step 11, multiple sets of wheel center vertical displacement simulation signals, wheel center vertical acceleration simulation signals and spring displacement simulation signals which meet the test precision can be obtained, so that a training set of the initial neural network model is constructed, iterative training of the initial neural network model is completed, and a target neural network model for measuring the wheel center vertical displacement of the automobile is obtained.
The target neural network model finally obtained in this embodiment can accurately acquire the vertical displacement signal of the wheel center (the vertical displacement measuring signal of the wheel center, or the vertical displacement measuring signal of the wheel center) according to the vertical acceleration signal of the wheel center (the vertical acceleration simulating signal of the wheel center, or the vertical acceleration measuring signal of the wheel center), and the spring displacement signal (the spring displacement simulating signal, or the spring displacement measuring signal), and the overall cost is low.
According to the wheel center vertical displacement signal processing method and device, the wheel center vertical acceleration is not only used as input to be solved to obtain the wheel center vertical displacement signal, the input signal comparison is single, and the situation that the accuracy of the obtained wheel center vertical signal is not easy to evaluate is reduced.
Because the wheel center vertical acceleration signal has some trend terms and noises, and the core of some schemes for obtaining the wheel center vertical acceleration signal by using the frequency domain integration method is to convert the ratio relation between the acceleration frequency domain amplitude and the frequency square to obtain the wheel center vertical displacement signal, the error is easily amplified by solving the wheel center vertical displacement signal by using the wheel center vertical acceleration in the low-frequency interval through the quadratic frequency domain integration method, and the accuracy of the result is reduced. This embodiment obtains the vertical displacement signal of wheel center with the vertical acceleration of wheel center and the solution of spring signal, can avoid the calculation error of the vertical acceleration of wheel center on the low frequency interval for it is higher to obtain the vertical displacement signal precision of wheel center to solve.
Here, the present embodiment provides a specific implementation manner of step 11, which specifically includes step 21 to step 24.
And step 21, obtaining an initial whole vehicle multi-body simulation model of the target type vehicle.
Specifically, the initial full-vehicle multi-body simulation model can be understood as a full-vehicle multi-body simulation model with incomplete optimization calibration.
Specifically, a front suspension model, a rear suspension model, a power assembly model, a steering model, a braking model, a rigid wheel model and a vehicle body model are respectively built so as to build an initial whole vehicle multi-body simulation model.
Specifically, the initial whole vehicle multi-body simulation model comprises necessary whole vehicle components for simulating a wheel center vertical acceleration simulation signal and a spring displacement simulation signal through the vertical displacement of a vehicle wheel center.
And step 22, inputting the wheel center vertical displacement simulation signal into the initial whole vehicle multi-body simulation model for simulation to obtain a simulation output signal of the initial whole vehicle multi-body simulation model.
Specifically, the initial whole vehicle multi-body simulation model simulates the relative movement condition and the relative load condition of each part of a target type vehicle under the influence of the wheel center vertical displacement simulation signal.
And step 23, driving the target type vehicle based on the wheel center vertical displacement actual signal, and obtaining a test sensing signal through a whole vehicle sensor group on the target type vehicle.
Wherein, whole car sensor group includes: one or more of a wheel center acceleration sensor, a shock absorber mounting point acceleration sensor, a spring displacement sensor and a stabilizer bar strain sensor.
Specifically, the wheel center vertical displacement actual signal can be obtained through the whole axle coupling test bed. The whole axle coupling test bed applies a road simulation test technology, adopts an axle coupling mode, simultaneously provides 6 degrees of freedom (vertical, horizontal, lateral, camber, steering and braking) input for each wheel of a passenger car, realizes the force and displacement simulation drive of the road load of a test yard in a laboratory, and performs relatively quick and comprehensive examination on a whole chassis and a car body structural member.
Specifically, when a section of wheel center vertical displacement simulation signal y is set randomlytestThe wheel center vertical displacement actual signal is output to drive a target type vehicle to move to obtain Awheel2Wheel center vertical acceleration test signal, Dspring2Spring displacement test signal, Atop2Vertical acceleration test signal and sigma of shock absorber mounting point2The method comprises the steps of testing sensing signals such as transverse stabilizer bar strain test signals and the like, and calibrating and adjusting one or more parameters such as finished automobile hard point coordinates, the mass of a chassis piece and an automobile body, the mass center and the rotational inertia of the chassis piece and the automobile body, the spring stiffness, the damper damping, the stiffness of a limiting block and a gap and the stiffness of a lining in an initial finished automobile multi-body simulation model according to the testing sensing signals, so that the simulation precision of the initial finished automobile multi-body simulation model is improved, the optimal calibration of the initial finished automobile multi-body simulation model is completed, and the finished automobile multi-body simulation model is obtained.
Specifically, the wheel center vertical displacement actual signal is a wheel center vertical displacement simulation signal ytestCorresponding to a true signal also having a signal amplitude of
Figure 981640DEST_PATH_IMAGE072
The signal frequency bandwidth is also 0 Hz-F (F is a set frequency upper limit), and the signal time length is also t.
Specifically, as shown in fig. 2, the layout of the whole vehicle sensor group provided in the embodiment of the present invention is schematically illustrated, and the whole vehicle sensor group in fig. 2 includes a wheel center acceleration sensor 1, a shock absorber mounting point acceleration sensor 2, a spring displacement sensor 3, and a stabilizer bar strain sensor 4, and of course, the types and the number of sensors in the whole vehicle sensor group may be increased or decreased according to actual needs, which is not limited herein.
The whole vehicle sensor group utilized by the embodiment is a common sensor, the installation is simple, the measurement requirement is not high, and the overall cost is further reduced.
And 24, optimizing the initial finished automobile multi-body simulation model into the finished automobile multi-body simulation model based on the test sensing signals and the simulation output signals.
Specifically, the error between the simulation output signal output by the optimized finished automobile multi-body simulation model and the test sensing signal should be smaller than a set threshold.
After the step 21, the initial finished automobile multi-body simulation model and the target type vehicle are controlled to be in the same finished automobile test simulation working condition, so that the initial finished automobile multi-body simulation model is optimized and adjusted, and the adjustment precision is improved.
Specifically, firstly, relevant parameters of a finished automobile test simulation working condition are applied to an initial finished automobile multi-body simulation model, then a target type vehicle is installed on a finished automobile shaft coupling test bed, and the finished automobile shaft coupling test bed is controlled to simulate the finished automobile test simulation working condition, so that the initial finished automobile multi-body simulation model and the target type vehicle are in the same finished automobile test simulation working condition.
Specifically, when a section of wheel center vertical displacement simulation signal y is set randomlytestAnd when the input is input into the whole vehicle shaft coupling test bed, the whole vehicle shaft coupling test bed can be controlled to simulate a whole vehicle test simulation working condition, and each wheel of the passenger vehicle is controlled to run with the limited degree of freedom under the working condition. Because the whole vehicle sensor group is arranged on the whole vehicle assembly, A can be obtainedwheel2Wheel center vertical acceleration test signal, Dspring2Spring displacement test signal, Atop2Vertical acceleration test signal and sigma of shock absorber mounting point2A transverse stabilizer bar strain test signal and other test sensing signals.
Here, the present embodiment provides a specific implementation manner of step 24, which specifically includes step 31.
Step 31, taking one or more expressions that the test sensing signal and the simulation output signal meet optimization calibration criteria as targets, adjusting modeling parameters of the initial finished automobile multi-body simulation model, and obtaining the finished automobile multi-body simulation model; wherein the expression of the optimized calibration criterion is:
Figure 345756DEST_PATH_IMAGE073
；
RMS is a root mean square value regularization solving function;
Figure 145085DEST_PATH_IMAGE074
simulating signals for vertical acceleration of the wheel center;
Figure 215940DEST_PATH_IMAGE075
is a wheel center vertical acceleration test signal;
Figure 708102DEST_PATH_IMAGE076
is a first error threshold;
Figure 418569DEST_PATH_IMAGE077
simulating a signal for the displacement of the spring;
Figure 772321DEST_PATH_IMAGE078
is a spring displacement test signal;
Figure 212529DEST_PATH_IMAGE079
is a second error threshold;
Figure 360745DEST_PATH_IMAGE080
simulating signals for the vertical acceleration of the upper point of the shock absorber;
Figure 948721DEST_PATH_IMAGE081
vertical acceleration test signals of points on the shock absorber are obtained;
Figure 840585DEST_PATH_IMAGE082
is a third error threshold;
Figure 604142DEST_PATH_IMAGE083
strain simulation signals of the transverse stabilizer bar are obtained;
Figure 923259DEST_PATH_IMAGE084
a transverse stabilizer bar strain test signal;
Figure 998531DEST_PATH_IMAGE085
is a fourth error threshold.
Specifically, the simulated output signal includes
Figure 428506DEST_PATH_IMAGE086
For simulating vertical acceleration of wheel centerA signal,
Figure 843307DEST_PATH_IMAGE087
Is a spring displacement simulation signal,
Figure 598905DEST_PATH_IMAGE088
Simulating the sum of signals for the vertical acceleration of the upper point of the shock absorber
Figure 99156DEST_PATH_IMAGE089
One or more of strain simulation signals of the transverse stabilizer bar; the test sensing signal comprises
Figure 254194DEST_PATH_IMAGE090
Is a wheel center vertical acceleration test signal,
Figure 539813DEST_PATH_IMAGE091
Is a spring displacement test signal,
Figure 450000DEST_PATH_IMAGE092
For the vertical acceleration test signal of the upper point of the shock absorber and
Figure 516176DEST_PATH_IMAGE093
one or more of the stabilizer bar strain test signals.
Specifically, one or more parameters such as the whole vehicle hard point coordinates, the mass of a chassis part and a vehicle body, the mass center and the rotational inertia of the chassis part and the vehicle body, the spring stiffness, the damper damping, the stiffness of a limiting block and the gap and the stiffness of a lining in the initial whole vehicle multi-body simulation model are corrected and adjusted according to the test sensing signals, the simulation precision of the initial whole vehicle multi-body simulation model is improved, the optimal calibration of the initial whole vehicle multi-body simulation model is completed, and the whole vehicle multi-body simulation model is obtained.
Therefore, the present embodiment provides a determination scheme of the current calibration precision of the initial full-vehicle multi-body simulation model, and the current calibration precision of the initial full-vehicle multi-body simulation model can be rapidly and accurately obtained through the above optimization calibration criterion, so as to complete the optimization calibration of the initial full-vehicle multi-body simulation model, thereby obtaining the full-vehicle multi-body simulation model.
Here, the present embodiment provides a specific implementation manner of step 13, and as shown in fig. 3, is a schematic structural diagram of an initial neural network model provided in the embodiment of the present invention, where the present embodiment adopts a feedback neural network model to implement the operation of step 13, and specifically includes steps 41 to 44.
Step 41, an input layer is constructed.
Wherein the output signal matrix of the input layer
Figure 553534DEST_PATH_IMAGE094
The simulation device comprises the wheel center vertical acceleration simulation signal and the spring displacement simulation signal.
Specifically, in this embodiment, the wheel center vertical acceleration simulation signal and the spring displacement simulation signal are used as the output signal matrix of the input layer, and the whole vehicle multi-body simulation model obtained in step 12 can be used to provide a large number of high-precision training samples for the neural network model, so as to complete the iterative training of the initial neural network model.
Step 42, a first hidden layer is constructed.
Wherein the propagation expression of the first hidden layer is:
Figure 411768DEST_PATH_IMAGE095
；
wherein the content of the first and second substances,
Figure 305906DEST_PATH_IMAGE096
a matrix of output signals for the first hidden layer;
Figure 515170DEST_PATH_IMAGE097
a first weight matrix for the input layer to the first hidden layer;
Figure 12011DEST_PATH_IMAGE098
a first bias matrix for the input layer to the first hidden layer.
In particular, the method comprises the following steps of,
Figure 741064DEST_PATH_IMAGE099
，
Figure 258633DEST_PATH_IMAGE100
，
Figure 768242DEST_PATH_IMAGE101
，
Figure 131091DEST_PATH_IMAGE102
；
wherein, a1~aiIs the output signal 1-i, x of the first hidden layer1Is a wheel center vertical acceleration signal, x2Is a spring displacement signal; kaA 2 × i matrix; b is a 1 × i matrix.
Step 43, constructing a second hidden layer; the propagation expression of the second hidden layer is as follows:
Figure 980229DEST_PATH_IMAGE103
；
wherein the content of the first and second substances,
Figure 340803DEST_PATH_IMAGE104
a matrix of output signals for the second hidden layer;
Figure 524660DEST_PATH_IMAGE105
a second weight matrix for the first hidden layer to the second hidden layer;
Figure 238669DEST_PATH_IMAGE106
a second bias matrix for the first hidden layer to the second hidden layer.
In particular, the method comprises the following steps of,
Figure 191582DEST_PATH_IMAGE107
，
Figure 723057DEST_PATH_IMAGE108
，
Figure 879363DEST_PATH_IMAGE109
；
wherein, b1~bjOutput signals 1-j of the first hidden layer;
Figure 911910DEST_PATH_IMAGE110
is a matrix of i x j; c is a matrix of 1 xj.
Step 44, constructing an output layer; the propagation expression of the output layer is as follows:
Figure 470061DEST_PATH_IMAGE111
；
wherein the content of the first and second substances,
Figure 172438DEST_PATH_IMAGE112
outputting a wheel center vertical displacement measurement signal for the output layer;
Figure 65308DEST_PATH_IMAGE113
a third weight matrix for the second hidden layer to the output layer;
Figure 160320DEST_PATH_IMAGE114
a third bias matrix for the second hidden layer to the output layer.
In particular, the method comprises the following steps of,
Figure 150141DEST_PATH_IMAGE115
，
Figure 836469DEST_PATH_IMAGE116
。
wherein, KcA matrix of j × 1; d is a matrix of 1 xj.
Here, the present embodiment provides a specific implementation manner of step 13, which specifically includes step 51.
Step 51, iterative training with the objective of satisfying the minimum error function E
Figure 154317DEST_PATH_IMAGE117
、
Figure 466350DEST_PATH_IMAGE118
、
Figure 140039DEST_PATH_IMAGE119
、
Figure 512115DEST_PATH_IMAGE120
、
Figure 723784DEST_PATH_IMAGE121
And
Figure 511612DEST_PATH_IMAGE122
(ii) a The specific iterative training formula is as follows:
Figure 164441DEST_PATH_IMAGE123
；
wherein the content of the first and second substances,
Figure 707418DEST_PATH_IMAGE124
simulating a signal for the vertical displacement of the wheel center;
Figure 812908DEST_PATH_IMAGE125
the number of times of iterative training is accumulated;
Figure 466743DEST_PATH_IMAGE126
is a learning efficiency parameter.
In particular, the method comprises the following steps of,
Figure 239659DEST_PATH_IMAGE127
for p-th iteration training
Figure 297744DEST_PATH_IMAGE128
，
Figure 405378DEST_PATH_IMAGE129
For p-th iteration training
Figure 348057DEST_PATH_IMAGE130
，
Figure 631271DEST_PATH_IMAGE131
For p-th iteration training
Figure 781629DEST_PATH_IMAGE132
，
Figure 596133DEST_PATH_IMAGE133
For p-th iteration training
Figure 591771DEST_PATH_IMAGE134
，
Figure 870436DEST_PATH_IMAGE135
For p-th iteration training
Figure 926117DEST_PATH_IMAGE136
，
Figure 149288DEST_PATH_IMAGE137
For p-th iteration training
Figure 699349DEST_PATH_IMAGE138
；
Figure 81789DEST_PATH_IMAGE139
For p-1 th iteration training
Figure 793524DEST_PATH_IMAGE140
，
Figure 97467DEST_PATH_IMAGE141
For p-1 th iteration training
Figure 451219DEST_PATH_IMAGE142
，
Figure 563531DEST_PATH_IMAGE143
For p-1 th iteration training
Figure 961014DEST_PATH_IMAGE144
，
Figure 34144DEST_PATH_IMAGE145
For p-1 th iteration training
Figure 175275DEST_PATH_IMAGE146
，
Figure 407673DEST_PATH_IMAGE147
For p-1 th iteration training
Figure 461211DEST_PATH_IMAGE148
，
Figure 739746DEST_PATH_IMAGE149
For p-1 th iteration training
Figure 232038DEST_PATH_IMAGE150
。
Specifically, as can be seen from the initial neural network model established in the foregoing, solving the initial neural network system model is to solve the coefficient matrix: ka、B、Kb、C、Kc、D。
Randomly setting a section of wheel center vertical displacement simulation signal ytestSignal amplitude of yAThe frequency bandwidth of the signal is 0 Hz-F, and the time length of the signal is t.
Simulating the vertical displacement of the wheel center by using the signal ytestInputting the signals into a multi-body simulation model of the whole vehicle, and calculating to obtain corresponding wheel center vertical acceleration simulation signals atestSpring displacement simulation signal dtest。
Randomly set wheel center vertical displacement simulation signal ytestAnd calculating to obtain a wheel center vertical acceleration simulation signal atestAnd spring displacement simulation signal dtestInputting the data into a trained neural network system model, and solving by combining a minimum gradient learning algorithm to minimize an error function ESo as to obtain a coefficient matrix: ka、B、Kb、C、Kc、D。
Based on the same inventive concept as the method, an embodiment of the present invention further provides a model training apparatus for measuring vertical displacement of a wheel center of a vehicle, as shown in fig. 4, which is a schematic structural diagram of the apparatus embodiment, and the apparatus includes:
the first establishing module 61 is used for obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal;
a second building module 62, configured to build an initial neural network model;
and the modeling module 63 is configured to input the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the vehicle wheel center vertical displacement, with a minimum error between the initial wheel center vertical displacement measurement signal output by the initial neural network model and the wheel center vertical displacement simulation signal as a target.
In a possible embodiment, the first establishing module includes:
the first acquisition module is used for acquiring an initial whole vehicle multi-body simulation model of the target type vehicle;
the second acquisition module is used for inputting the wheel center vertical displacement simulation signal into the initial finished automobile multi-body simulation model for simulation to obtain a simulation output signal of the initial finished automobile multi-body simulation model;
the third acquisition module is used for driving the target type vehicle based on the wheel center vertical displacement actual signal and acquiring a test sensing signal through a whole vehicle sensor group on the target type vehicle; wherein, whole car sensor group includes: one or more of a wheel center acceleration sensor 1, a shock absorber mounting point acceleration sensor 2, a spring displacement sensor 3 and a stabilizer bar strain sensor 4;
and the fourth acquisition module is used for optimizing the initial finished automobile multi-body simulation model into the finished automobile multi-body simulation model based on the test sensing signal and the simulation output signal.
In a possible embodiment, the apparatus further comprises:
the first control module is used for controlling the initial whole vehicle multi-body simulation model and the target type vehicle to be in the same whole vehicle test simulation working condition after the initial whole vehicle multi-body simulation model of the target type vehicle is obtained;
the fourth obtaining module specifically includes:
the optimization adjustment module is used for adjusting modeling parameters of the initial finished automobile multi-body simulation model by taking one or more expressions of the test sensing signal and the simulation output signal meeting optimization calibration criteria as targets to obtain the finished automobile multi-body simulation model; wherein the expression of the optimized calibration criterion is:
Figure 381260DEST_PATH_IMAGE151
；
RMS is a root mean square value regularization solving function;
Figure 792649DEST_PATH_IMAGE152
simulating signals for vertical acceleration of the wheel center;
Figure 778054DEST_PATH_IMAGE153
is a wheel center vertical acceleration test signal;
Figure 526567DEST_PATH_IMAGE154
is a first error threshold;
Figure 343345DEST_PATH_IMAGE155
simulating a signal for the displacement of the spring;
Figure 253532DEST_PATH_IMAGE156
is a spring displacement test signal;
Figure 991812DEST_PATH_IMAGE157
is a second error threshold;
Figure 950541DEST_PATH_IMAGE158
simulating signals for the vertical acceleration of the upper point of the shock absorber;
Figure 339934DEST_PATH_IMAGE159
vertical acceleration test signals of points on the shock absorber are obtained;
Figure 171754DEST_PATH_IMAGE160
is a third error threshold;
Figure 381019DEST_PATH_IMAGE161
strain simulation signals of the transverse stabilizer bar are obtained;
Figure 222067DEST_PATH_IMAGE162
a transverse stabilizer bar strain test signal;
Figure 262704DEST_PATH_IMAGE163
is a fourth error threshold.
In a possible embodiment, the second establishing module includes:
a first construction module for constructing an input layer; wherein the output signal matrix of the input layer
Figure 531006DEST_PATH_IMAGE164
The wheel center vertical acceleration simulation signal and the spring displacement simulation signal are included;
the second construction module is used for constructing the first hidden layer; wherein the propagation expression of the first hidden layer is:
Figure 227566DEST_PATH_IMAGE165
；
wherein the content of the first and second substances,
Figure 606726DEST_PATH_IMAGE166
a matrix of output signals for the first hidden layer;
Figure 377236DEST_PATH_IMAGE167
a first weight matrix for the input layer to the first hidden layer;
Figure 65706DEST_PATH_IMAGE168
a first bias matrix for the input layer to the first hidden layer;
the third construction module is used for constructing a second hidden layer; the propagation expression of the second hidden layer is as follows:
Figure 593771DEST_PATH_IMAGE169
；
wherein the content of the first and second substances,
Figure 511042DEST_PATH_IMAGE170
a matrix of output signals for the second hidden layer;
Figure 401638DEST_PATH_IMAGE171
a second weight matrix for the first hidden layer to the second hidden layer;
Figure 261010DEST_PATH_IMAGE172
a second bias matrix for the first hidden layer to the second hidden layer;
the fourth construction module is used for constructing an output layer; the propagation expression of the output layer is as follows:
Figure 417316DEST_PATH_IMAGE173
；
wherein the content of the first and second substances,
Figure 184283DEST_PATH_IMAGE174
outputting a wheel center vertical displacement measurement signal for the output layer;
Figure 742435DEST_PATH_IMAGE175
a third weight matrix for the second hidden layer to the output layer;
Figure 444811DEST_PATH_IMAGE176
a third bias matrix for the second hidden layer to the output layer.
In one possible embodiment, the modeling module includes:
a training module for iterative training with the objective of satisfying the minimum error function Emin
Figure 337681DEST_PATH_IMAGE177
、
Figure 659072DEST_PATH_IMAGE178
、
Figure 852156DEST_PATH_IMAGE179
、
Figure 538484DEST_PATH_IMAGE180
、
Figure 590753DEST_PATH_IMAGE181
And
Figure 902786DEST_PATH_IMAGE182
(ii) a The specific iterative training formula is as follows:
Figure 497846DEST_PATH_IMAGE183
；
wherein the content of the first and second substances,
Figure 604343DEST_PATH_IMAGE184
simulating a signal for the vertical displacement of the wheel center;
Figure 143908DEST_PATH_IMAGE185
the number of times of iterative training is accumulated;
Figure 255436DEST_PATH_IMAGE186
is a learning efficiency parameter.
Based on the same inventive concept as the method, the embodiment of the present invention further provides a method for measuring the vertical displacement of the wheel center of the vehicle, as shown in fig. 5, which is a flowchart of the embodiment of the method, and specifically includes steps 71 to 72.
And step 71, acquiring a wheel center vertical acceleration actual measurement signal and a spring displacement actual measurement signal of the whole vehicle in the running process of the whole vehicle.
Specifically, the type of the entire vehicle is the same as the type of the target type vehicle.
And 72, inputting the measured wheel center vertical acceleration signal and the measured spring displacement signal into a target neural network model in any one of the training methods to obtain the measured wheel center vertical displacement signal of the whole vehicle.
Specifically, a wheel center three-direction acceleration signal and a spring displacement signal which are acquired in the running process of the vehicle are input into the neural network model system, and finally a wheel center vertical displacement signal in the running process of the vehicle is obtained.
Specifically, the target neural network model in any one of the above training methods is a trained target neural network model.
Based on the same inventive concept as the method, an embodiment of the present invention further provides a device for measuring a vertical displacement of a wheel center of a vehicle, as shown in fig. 6, which is a schematic structural diagram of the embodiment of the device, and the device includes:
the acquiring unit 81 is used for acquiring an actual measurement signal of the vertical acceleration of the wheel center of the whole vehicle and an actual measurement signal of the spring displacement during the running process of the whole vehicle;
and the processing unit 82 is configured to input the measured wheel center vertical acceleration signal and the measured spring displacement signal into the target neural network model in the model training method, so as to obtain the measured wheel center vertical displacement signal of the entire vehicle.
Based on the same inventive concept as in the previous embodiments, an embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of any one of the methods when executing the program.
Based on the same inventive concept as in the previous embodiments, embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of any of the methods described above.
The technical scheme provided by the embodiment of the invention at least has the following technical effects or advantages:
firstly, establishing an initial whole vehicle multi-body simulation model; the input quantity comprises a wheel center vertical displacement simulation signal, and the output quantity comprises a wheel center vertical acceleration simulation signal and a damper spring displacement simulation signal; then, simulating a finished automobile test simulation working condition by using the finished automobile axle coupling test bed, acquiring and obtaining sensing signals output by a finished automobile sensor group, and obtaining a finished automobile multi-body simulation model through optimized calibration; then establishing an initial neural network model; and finally, training by utilizing multiple groups of input quantities and output quantities corresponding to the whole vehicle multi-body simulation model to obtain a target neural network model for measuring the vertical displacement of the wheel center of the vehicle. According to the embodiment of the invention, the common sensor and the neural network model are used for measuring the vertical displacement of the wheel center of the vehicle, so that the measurement cost is reduced, the vertical acceleration signal of the wheel center and the spring signal are considered as calculation input in the measurement, the calculation error of the vertical acceleration of the wheel center in a low-frequency interval is reduced, the measurement precision is improved, and the vertical displacement of the wheel center of the vehicle is measured at low cost and high precision.
Furthermore, the embodiment of the invention adopts common sensors to build the whole vehicle sensor group, obtains the whole vehicle multi-body simulation model through optimized calibration, reduces the use of laser sensors with higher cost and the like, and further reduces the whole measurement cost. Meanwhile, the corrected and optimized whole vehicle multi-body simulation model can provide a high-precision training sample for the training of the initial neural network model, so that the model training quality is further improved, and the measurement precision is improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (modules, systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (10)

1. A method of model training, the method comprising:
obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal;
constructing an initial neural network model;
and inputting the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the wheel center vertical displacement of the vehicle by taking the minimum error between the initial wheel center vertical displacement measurement signal output by the initial neural network model and the wheel center vertical displacement simulation signal as a target.
2. The model training method of claim 1, wherein said obtaining a full-vehicle multi-body simulation model of a target type vehicle comprises:
obtaining an initial whole vehicle multi-body simulation model of the target type vehicle;
inputting the wheel center vertical displacement simulation signal into the initial whole vehicle multi-body simulation model for simulation to obtain a simulation output signal of the initial whole vehicle multi-body simulation model;
driving the target type vehicle based on the wheel center vertical displacement actual signal, and obtaining a test sensing signal through a whole vehicle sensor group on the target type vehicle; wherein, whole car sensor group includes: one or more of a wheel center acceleration sensor, a shock absorber mounting point acceleration sensor, a spring displacement sensor and a transverse stabilizer bar strain sensor;
and optimizing the initial finished automobile multi-body simulation model into the finished automobile multi-body simulation model based on the test sensing signal and the simulation output signal.
3. The model training method of claim 2, wherein after obtaining an initial full-vehicle multi-body simulation model of the target type vehicle, the method further comprises: controlling the initial whole vehicle multi-body simulation model and the target type vehicle to be in the same whole vehicle test simulation working condition;
based on the test sensing signal and the simulation output signal, the initial finished automobile multi-body simulation model is optimized into the finished automobile multi-body simulation model, and the method specifically comprises the following steps:
taking one or more expressions that the test sensing signals and the simulation output signals meet optimization calibration criteria as targets, adjusting modeling parameters of the initial finished automobile multi-body simulation model, and obtaining the finished automobile multi-body simulation model; wherein the expression of the optimized calibration criterion is:
Figure DEST_PATH_IMAGE002
；
RMS is a root mean square value regularization solving function;
Figure DEST_PATH_IMAGE004
simulating signals for vertical acceleration of the wheel center;
Figure DEST_PATH_IMAGE006
is a wheel center vertical acceleration test signal;
Figure DEST_PATH_IMAGE008
is a first error threshold;
Figure DEST_PATH_IMAGE010
simulating a signal for the displacement of the spring;
Figure DEST_PATH_IMAGE012
is a spring displacement test signal;
Figure DEST_PATH_IMAGE014
is a second error threshold;
Figure DEST_PATH_IMAGE016
simulating signals for the vertical acceleration of the upper point of the shock absorber;
Figure DEST_PATH_IMAGE018
vertical acceleration test signals of points on the shock absorber are obtained;
Figure DEST_PATH_IMAGE020
is a third error threshold;
Figure DEST_PATH_IMAGE022
strain simulation signals of the transverse stabilizer bar are obtained;
Figure DEST_PATH_IMAGE024
a transverse stabilizer bar strain test signal;
Figure DEST_PATH_IMAGE026
is a fourth error threshold.
4. The model training method of claim 1, wherein the constructing an initial neural network model comprises:
constructing an input layer; wherein the output signal matrix of the input layer
Figure DEST_PATH_IMAGE028
The wheel center vertical acceleration simulation signal and the spring displacement simulation signal are included;
constructing a first hidden layer; wherein the propagation expression of the first hidden layer is:
Figure DEST_PATH_IMAGE030
；
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE032
a matrix of output signals for the first hidden layer;
Figure DEST_PATH_IMAGE034
a first weight matrix for the input layer to the first hidden layer;
Figure DEST_PATH_IMAGE036
a first bias matrix for the input layer to the first hidden layer;
constructing a second hidden layer; the propagation expression of the second hidden layer is as follows:
Figure DEST_PATH_IMAGE038
；
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE040
a matrix of output signals for the second hidden layer;
Figure DEST_PATH_IMAGE042
a second weight matrix for the first hidden layer to the second hidden layer;
Figure DEST_PATH_IMAGE044
a second bias matrix for the first hidden layer to the second hidden layer;
constructing an output layer; the propagation expression of the output layer is as follows:
Figure DEST_PATH_IMAGE046
；
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE048
outputting a wheel center vertical displacement measurement signal for the output layer;
Figure DEST_PATH_IMAGE050
a third weight matrix for the second hidden layer to the output layer;
Figure DEST_PATH_IMAGE052
a third bias matrix for the second hidden layer to the output layer.
5. The model training method of claim 4, wherein the obtaining of the target neural network model for measuring the vertical displacement of the wheel center of the vehicle comprises:
iterative training with the goal of satisfying the minimum of the error function Emin
Figure DEST_PATH_IMAGE054
、
Figure DEST_PATH_IMAGE056
、
Figure DEST_PATH_IMAGE058
、
Figure DEST_PATH_IMAGE060
、
Figure DEST_PATH_IMAGE062
And
Figure DEST_PATH_IMAGE064
(ii) a The specific iterative training formula is as follows:
Figure DEST_PATH_IMAGE066
；
RMS is a root mean square value regularization solving function;
Figure DEST_PATH_IMAGE068
simulating a signal for the vertical displacement of the wheel center;
Figure DEST_PATH_IMAGE070
the number of times of iterative training is accumulated;
Figure DEST_PATH_IMAGE072
is a learning efficiency parameter.
6. A model training apparatus, the apparatus comprising:
the first establishing module is used for obtaining a whole vehicle multi-body simulation model of a target type vehicle; the input of the whole vehicle multi-body simulation model comprises a wheel center vertical displacement simulation signal, and the output of the whole vehicle multi-body simulation model comprises a wheel center vertical acceleration simulation signal and a spring displacement simulation signal;
the second establishing module is used for establishing an initial neural network model;
and the modeling module is used for inputting the wheel center vertical acceleration simulation signal and the spring displacement simulation signal into the initial neural network model for iterative training to obtain a target neural network model for measuring the vehicle wheel center vertical displacement by taking the minimum error between the initial wheel center vertical displacement measurement signal and the wheel center vertical displacement simulation signal output by the initial neural network model as a target.
7. A method for measuring vertical displacement of a wheel center of a vehicle, the method comprising:
acquiring a wheel center vertical acceleration actual measurement signal and a spring displacement actual measurement signal of the whole vehicle in the running process of the whole vehicle;
inputting the measured wheel center vertical acceleration signal and the measured spring displacement signal into a target neural network model in the model training method according to any one of claims 1 to 5, and acquiring the measured wheel center vertical displacement signal of the whole vehicle.
8. A vehicle wheel center vertical displacement measuring device, the method comprising:
the acquiring unit is used for acquiring a wheel center vertical acceleration actual measurement signal and a spring displacement actual measurement signal of the whole vehicle in the running process of the whole vehicle;
the processing unit is used for inputting the measured wheel center vertical acceleration signal and the measured spring displacement signal into the target neural network model in the model training method according to any one of claims 1 to 5, and acquiring the measured wheel center vertical displacement signal of the whole vehicle.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program to carry out the steps of the method of any one of claims 1 to 5; or for executing said computer program for carrying out the steps of the method of claim 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, is adapted to carry out the steps of the method of any one of claims 1 to 5; or for executing said computer program for carrying out the steps of the method of claim 7.
CN202111132180.5A 2021-09-27 2021-09-27 Vehicle wheel center vertical displacement measuring method and device based on model training Active CN113570057B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111132180.5A CN113570057B (en) 2021-09-27 2021-09-27 Vehicle wheel center vertical displacement measuring method and device based on model training

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111132180.5A CN113570057B (en) 2021-09-27 2021-09-27 Vehicle wheel center vertical displacement measuring method and device based on model training

Publications (2)

Publication Number Publication Date
CN113570057A CN113570057A (en) 2021-10-29
CN113570057B true CN113570057B (en) 2021-12-14

Family

ID=78174676

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111132180.5A Active CN113570057B (en) 2021-09-27 2021-09-27 Vehicle wheel center vertical displacement measuring method and device based on model training

Country Status (1)

Country Link
CN (1) CN113570057B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114894138B (en) * 2022-06-14 2023-06-30 广汽本田汽车有限公司 Method, system, equipment and storage medium for detecting vertical displacement difference of automobile wheel
CN116702628A (en) * 2023-07-31 2023-09-05 江铃汽车股份有限公司 Automobile wheel center vertical displacement solving method based on convolution depth neural network
CN116663434B (en) * 2023-07-31 2023-12-05 江铃汽车股份有限公司 Whole vehicle load decomposition method based on LSTM deep neural network

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103852268B (en) * 2012-12-03 2016-05-18 上海汽车集团股份有限公司 Learn the system of characteristic for testing suspension dynamic motion
CN103196681B (en) * 2013-03-13 2015-11-11 北京交通大学 Based on the train operation comfort degree predication method of bogie acceleration
CN104239734A (en) * 2014-09-24 2014-12-24 重庆长安汽车股份有限公司 Load analysis method for four-wheel six-component road spectrum of finished automobile
CN106644519B (en) * 2017-01-17 2020-03-06 北京汽车研究总院有限公司 Method and device for identifying parameters of vehicle vertical dynamics model
JP6842112B2 (en) * 2017-04-27 2021-03-17 国立大学法人 東京大学 Road surface profile estimation device, road surface profile estimation system, road surface profile estimation method and road surface profile estimation program
US11029693B2 (en) * 2017-08-08 2021-06-08 Tusimple, Inc. Neural network based vehicle dynamics model
CN107480395B (en) * 2017-08-29 2019-12-13 燕山大学 Method and system for constructing vehicle steering knuckle load spectrum prediction model
CN108859648B (en) * 2018-08-16 2020-02-18 辽宁工业大学 Suspension shock absorber damping control switching weighting coefficient determination method
CN109927501A (en) * 2019-03-12 2019-06-25 辽宁科技大学 A kind of intelligent control method of Vehicle Semi-active Suspension System
CN111241706B (en) * 2020-01-23 2022-10-21 上海理工大学 Automobile suspension dynamic load performance evaluation method based on dynamic K & C test system
CN111634291A (en) * 2020-05-07 2020-09-08 中国第一汽车股份有限公司 Damper state detection method and device, vehicle and storage medium

Also Published As

Publication number Publication date
CN113570057A (en) 2021-10-29

Similar Documents

Publication Publication Date Title
CN113570057B (en) Vehicle wheel center vertical displacement measuring method and device based on model training
US5610330A (en) Effective road profile control method for a spindle-coupled road simulator
US6134957A (en) Multiple degree-of-freedom tire modeling method and system for use with a vehicle spindle-coupled simulator
Dodds et al. Laboratory road simulation for full vehicle testing: a review
CN104239734A (en) Load analysis method for four-wheel six-component road spectrum of finished automobile
JP2010529420A (en) Method and system for axle evaluation and adjustment using load system and vehicle model
CN103076148B (en) Drop test dual-four-degree-of-freedom half vehicle model-based vehicle parameter identification method
CN106644519B (en) Method and device for identifying parameters of vehicle vertical dynamics model
You et al. Advances of virtual testing and hybrid simulation in automotive performance and durability evaluation
CN112595528B (en) Method for identifying driving side exciting force of power assembly suspension under vehicle running condition
JP2019023667A (en) Method and system for testing connection hybrid dynamic system
US5900542A (en) Method and apparatus for modelling a tire for use with a vehicle spindle-coupled simulator
GB2494712A (en) Test rig and method for testing a subsystem of a vehicle
US5750890A (en) Method and apparatus for modelling a tire for use with a vehicle spindle-coupled simulator
Elkafafy et al. Machine learning and system identification for the estimation of data-driven models: An experimental case study illustrated on a tire-suspension system
Yang et al. A study on a vehicle semi-active suspension control system based on road elevation identification
CN111090959B (en) Vehicle load spectrum acquisition method and system
Schudt et al. Virtual road load data acquisition in practice at general motors
CN111380649A (en) Vehicle multi-dimensional force sensor decoupling method based on PSO (particle swarm optimization) optimization LSSVM (least squares support vector machine)
KR20160053983A (en) Test system having a compliant actuator assembly and iteratively obtained drive
Wenkui et al. Dynamic simulation analysis of pothole event based on vRLDA
KR102218146B1 (en) Method and system for off-line control for simulation of coupled hybrid dynamic systems
US20160363508A1 (en) Method and system for inspecting a motor vehicle subsystem
CN113607392A (en) Spring arm durability test method and device
CN114861335B (en) Calibration method of automobile dynamics calculation model and related equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant