CN105975442A - Method and device computing reversed function value of function model based on data - Google Patents
Method and device computing reversed function value of function model based on data Download PDFInfo
- Publication number
- CN105975442A CN105975442A CN201610110437.XA CN201610110437A CN105975442A CN 105975442 A CN105975442 A CN 105975442A CN 201610110437 A CN201610110437 A CN 201610110437A CN 105975442 A CN105975442 A CN 105975442A
- Authority
- CN
- China
- Prior art keywords
- input data
- data point
- data points
- point
- output valve
- 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.)
- Pending
Links
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/24—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means
- F02D41/26—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means using computer, e.g. microprocessor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/17—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D2041/1433—Introducing closed-loop corrections characterised by the control or regulation method using a model or simulation of the system
- F02D2041/1434—Inverse model
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
Abstract
The invention relates to a method used for resolving searching input data points applied to a function model; the input data points are attributed to output values of output parameters; the method comprises the following steps: using output values, assigned by the function model, of the output parameters to provide (S1) a plurality of input data points applied to the function mode; using the first solving method to determine (S2, S3) two input data points of the input data points, wherein the output values of the input data points comprise preset output values; executing (S4) the second reliable convergent resolving method on the base of the two input data points, thus obtaining the searching input data points.
Description
Technical field
The present invention relates to function model, particularly the backwards calculation of nonlinear function model.In addition the present invention relates to one
For trying to achieve the possible defeated of reverse, function models based on data in the case of the output valve of previously given output parameter
The method entering data point.
Background technology
In the control equipment of motor vehicles, generally realize function by function model.In order in control equipment, especially
It is in the device for controlling engine of explosive motor, to perform function model, go back in addition to traditional function model for this
Function model based on data can be set.Function models based on data are also known as non-parametric model and can not have
Special previously given in the case of by refering to data (Trainingsdaten), the most substantial amounts of set up refering to data point.
By the control module with main computation unit and single model computing unit known in the art, described model meter
Calculate unit for calculating the function models based on data in control equipment.The most such as document DE 10 2,010 028 259
A1 discloses a kind of control equipment with additional logic circuit and is set as model computing unit, described model computing unit
Meter is for gauge index function, in order to auxiliary performs Bayes-Return Law, and it is required for calculating Gaussian process mould especially
Type.
The input data point that function model the is one or more dimensions distribution output used in this device for controlling engine
The output valve of parameter.Multiple function models are the most nonlinear and can have the section with positive and negative gradient.
For multiple applicable cases, especially for explosive motor, by the functions based on data in control equipment
Model be enough to calculate functional value.But known these applicable cases, wherein need the inverse function value of Computation function model,
I.e. based on output parameter previously given output valve calculates input data point.
Such as by using Newton method can realize, on the basis of the previously given output valve of function models based on data
On calculate and input data point accordingly, achieve described output for described input data point when forward Computation function model
Value.But cannot ensure, Newton method restrains for each reliable output valve, i.e. Newton method is not for each output valve
Furnish an answer.
Summary of the invention
Propose a kind of according to claim 1, for calculating reverse functions based on data according to the present invention
The method of the functional value of model, and according to the corresponding device described in claim arranged side by side.
Other design is given in the dependent claims.
Propose a kind of for trying to achieve the side of input data point that found, for function model according to first aspect
Method, for calculating particularly for controlling the letter of the explosive motor of motor vehicles in control equipment the most in a motor vehicle
Number.The input data point found is associated with the output valve of output parameter.Described method has a following step:
-utilize that the output valve arranged by described function model of output parameter provides for described function model multiple
Input data point;
-determine two input data points in multiple input data point according to the first solution, it is associated with these inputs
The output valve of data point includes previously given output valve;
-on the basis of two input data points determined, perform the solution that the second is restrained reliably, in order to try to achieve institute
The input data point found.
Said method is capable of reliably determining the input data point being associated with previously given output valve.
Traditional method, such as Newton method for backwards calculation function model not must go out answer.Just from so-called
Input data point for the input data point that newton is fractal never to restrain.And can not identify in previously given output valve
Go out, whether convergence occurs when determining the input data point of arranging.
By the combination of both solutions it is ensured that phase can be tried to achieve for each previously given output valve
One of the input data point answered or multiple possible input data point.For this select: the first solution, this solution with
The output valve selected the most always can find the best answer;And second workaround, this solution always to
Real answer restrains, as long as it exists.
The design of said method is the combination of both solutions, thus can hold in previously given function model
Row backwards calculation, described backwards calculation reliably provides the input data point being associated with previously given output valve.Relative to
Newton method advantage is, even if also can calculate the function model on basis along forward in the case of backwards calculation, thus relatively
Output valve is can determine that in each input data point.
By also providing input data point for each output valve when backwards calculation function model, can be in control equipment
Use this " reverse model " for function is calculated, described function always be the need for effect, be associated with output parameter
The input data point of output valve.
Additionally, the first solution described can be corresponding to the backstepping method (Inversi for piecewise linear approximation
Onsverfahren).Alternatively or additionally, described second workaround can be corresponding to Brunt method.By this way
Can extend described function model, thus calculate reverse model, described reverse model is for each output of function model
Value the most reliably provides and inputs data point accordingly.
According to a kind of embodiment, the plurality of input data point being provided can be with uniformly in input data space
The mode of distribution is arranged.
It can be stated that provide two input data points in the plurality of input data point in the way of a kind of sequence,
Described sequence is constituted in the way of monotone variation at least with one of its input parameter.
Especially, two inputs the first and second input data points can being defined as in the plurality of input data point
Data point, wherein said first input data point is by selecting described first input data from the sequence of described input data point
One or more this mode in point determines, and described second input data point is by from described input data point
The one or more this mode selected in sequence in last input data point determines.
In addition two input data in the plurality of input data point can be selected from the plurality of input data point
Point so that said two input data point has the interval of maximum possible about the input parameter of monotone variation.
Accompanying drawing explanation
Embodiment is elaborated below by accompanying drawing.Wherein:
Fig. 1 shows the schematic diagram of function model based on data, and described function model is based on coming by means of input data point
Calculate the output valve of output parameter;
Fig. 2 shows the flow chart for the method for function models based on data carry out backwards calculation is described;With
Fig. 3 shows the schematic diagram of the curve trend of function model based on data, and wherein output valve is associated with multiple input number
Strong point.
Detailed description of the invention
Fig. 1 shows that the schematic diagram of function model is as functional blocks 1.The function model f(x realized in functional blocks 1) energy
Realize in any way, such as function models based on data, such as Gaussian process model, and can be defeated
Go out to draw in output valve y of parameter the combination of (dimension n) input data point x(input value of one or more dimensions).
The function model such as control equipment in motor vehicles can be arranged for calculating particularly for controlling machine
The function of the explosive motor of motor-car.Described control equipment includes data processing equipment for this, and this data processing equipment can be held
Row method as described below.
May need anti-in order to the function in device for controlling engine is designed for controlling the explosive motor of motor vehicles
To the function model performed by calculating, it is the output valve of the determination of described output parameterArrange input data point x, its conduct
Input parameter vector draws output valve y of output parameter in forward calculates (Vorw rtsberechnung).Using reversely
Especially needed during model guarantee, under any circumstance can provide the most defeated for previously given output valve y of output parameter
Enter data point x.
The trend of function can be almost the trend of arbitrary continuation, is wherein likely to occur the section with negative or positive gradient.
Particularly when the section with negative or positive gradient occurs, function f(x) can have a following trend, wherein, one determine defeated
Go out to be worth y be associated with multiple input data point x=x_1, x_2 ..., x_n}.This trend such as figure 2 illustrates.Can see
Go out, it is possible to be an exemplary output valveArrange multiple input data point x_1, x_2 and x_3.
For at function model f(x) output parameter output valve y on the basis of reliably determine input data point x
Method propose the first solution, such as the backstepping method of piecewise linear approximation and the solution of the second convergence,
The particularly combination of Brunt method.These solutions are used for, and determine function g(x)=f(x)-Zero-bit.
In the backstepping method of piecewise linear approximation, along one or more input parameters xiWherein i=1 ... n(n=
The dimension of input data point) (wherein xi < xi+ 1) try to achieve or select sample point x1,x2,…,xkQuantity k, wherein sample point
x1_ 1 and xn_ k defines the outer limit of input parameter space.This is calculated affiliated functional value yj(j=1,2 ... k).
Described calculating such as can realize by hardware or by software.In the case of calculating in hardware cell
Multiple functional value y can also be performed concurrently or partly in parallel according to the quantity calculating core (Rechenkern)j(j=1,
2 ..., calculating k).
Determine the output valve that the backwards calculation to be passed through of output parameter is tried to achieve in the following manner: find yjAnd yj+1, thus
MakeAt yjAnd yj+1Between.By interpolation method interpolation in other words, by xjAnd xj+1Approximate found input data point。
When the input data point foundDuring corresponding to output valve y, this output valve is in by described sample point x1,x2,…,xk
Output valve y at placej(j=1,2 ..., in region k) defined, described input data point can be for the output valve found
Independently determine with the quantity calculating sample point.If about previously given output valveThere is an input data point, then lead to
Cross the first solution and reliably find out this input data point.
The another kind of solution that can be designed to Brunt method is the solution of convergence.Solution excellent of convergence
Point is, they make the output valve providedAlways join possible output data point。
Brunt method is that it makes two way classification, secant method (linear interpolation in other words for the method determining zero-bit iteratively
Method) and reverse quadratic interpolattion be bonded to each other.In Brunt method, each for calculate iterative step only need one defeated
Go out to be worth y.The derivative of gradient or higher can not be calculated at this.As generally in the solution of convergence, Brunt method is given in advance
Fixed suitable limit number strong point a, b ∈ { x1,x2,…,xkFor previously given output valve in the case of }Reliably restrain to
The input data point found.Therefore specify, find out applicable ultimate value, thus condition is g(a) * g(b) < 0, in other words
The input data point foundBe in input data point g(a) and g(b) between interval in.
The combination of both solutions gives the following flow process illustrated in the flowchart of fig. 3.
In step sl, in application, the first is associated with previously given output valve for reliably determiningInput number
Strong pointSolution in, it is provided that substantial amounts of sample point x1And xk, these sample points are in and are described by function model
In input point space.
The most in step s 2 for the sample point x of two outsides1And xkTest, if meet initial condition g(x1) * g
(xk) < 0.(select: yes) if this is the case, call the solution of convergence, such as Brunt method the most in step s3, and
And correspondingly try to achieve input data pointIn this input data point or one of them.If not this situation (select:
No), limit the interval between end points the most further, and use corresponding back to back sample point x2And xk-1。
If present four input data points x1, x2, xk-1, xkIn meet condition g(a a pair) * g(b) < 0(wherein a ∈
{x1,x2, b ∈ { xk-1,xk) (selecting: be "Yes" in step s 2), then by means of convergence method (such as Brunt method) in step
S3 tries to achieve described input data point, inputs data point plus other the most in step s 4, and
And test in step s 2, whether a pair a, b have broken away from condition g(a) * g(b) < 0.As long as this point performs so, until meeting
For performing the above-mentioned condition of convergence method.
It can be stated that the quantity of the reduction (Schleifen) carried out along with S4 is defined as the quantity determined, in order under
The interruption calculated it is capable of: can not be for previously given output valve in the case of statingArrange input data point。
The advantage of said method is, calculating merely with forward just can be relative to the previously given output of output parameter
Value determines input data point.Thus, this method can apply to every kind of arbitrary function model and can be used for base especially
Function model in data.
Claims (10)
1. it is used for trying to achieve the method for input data point that found, for function model, the most in a motor vehicle
For calculating especially for controlling the function of the explosive motor of motor vehicles, the input data wherein found in control equipment
Point is associated with the output valve of output parameter, and described method has a following step:
-utilize the output valve arranged by described function model of output parameter to provide (S1) for described function model
Multiple input data points;
-determine two input data points in (S2, S3) multiple input data point according to the first solution, it is associated with
The output valve of these input data points includes previously given output valve;
-perform, on the basis of said two input data point, the solution that (S4) the second is restrained reliably, in order to try to achieve
The input data point found.
Method the most according to claim 1, the first solution wherein said is corresponding to for piecewise linear approximation
Backstepping method.
Method the most according to claim 1 and 2, wherein said second workaround is corresponding to for Brunt method.
The most according to the method in any one of claims 1 to 3, the plurality of described input data point being provided is in input
Data space is arranged in a uniformly distributed manner.
Method the most according to any one of claim 1 to 4, wherein provides the plurality of defeated in the way of a kind of sequence
Entering two input data points in data point, described sequence is at least with one of its input parameter structure in the way of monotone variation
Become.
First and second input data points are wherein defined as the plurality of input number by method the most according to claim 5
Two input data points in strong point, wherein said first input data point is by selecting from the sequence of described input data point
Described first input data point in one or more this modes determine, and described second input data point by from
The one or more this mode selected in the sequence of described input data point in last input data point determines.
7., according to the method according to any one of claim 5 to 6, wherein select described many from the plurality of input data point
Two input data points in individual input data point so that said two input data point has about the input parameter of monotone variation
There is the interval of maximum possible.
8. a computer program, it is arranged for performing all steps according to method according to any one of claim 1 to 7
Suddenly.
9. a machine-readable storage medium, stores the computer program described in good grounds claim 8 thereon.
10. the electricity being particularly used for for calculating the most in a motor vehicle controlling the function of the explosive motor of motor vehicles
Sub-control equipment, its be arranged for perform according to method according to any one of claim 1 to 7 institute in steps.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102015204218.9 | 2015-03-10 | ||
DE102015204218.9A DE102015204218A1 (en) | 2015-03-10 | 2015-03-10 | Method and device for calculating a function value of an inverted data-based function model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105975442A true CN105975442A (en) | 2016-09-28 |
Family
ID=56800680
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610110437.XA Pending CN105975442A (en) | 2015-03-10 | 2016-02-29 | Method and device computing reversed function value of function model based on data |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN105975442A (en) |
DE (1) | DE102015204218A1 (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102020000327A1 (en) | 2020-01-21 | 2021-07-22 | Mtu Friedrichshafen Gmbh | Method for model-based control and regulation of an internal combustion engine |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1474046A (en) * | 2002-08-06 | 2004-02-11 | �����Զ�����ҵ��ʽ���� | Control device and its contro method for output power of internal combustion engine |
US20050038576A1 (en) * | 2003-08-12 | 2005-02-17 | Honda Motor Co., Ltd. | Control apparatus for hybrid vehicle |
CN101641511A (en) * | 2007-03-05 | 2010-02-03 | 丰田自动车株式会社 | Internal combustion engine controller |
US20130338975A1 (en) * | 2012-06-13 | 2013-12-19 | Hitachi, Ltd | Method for co-simulation of two or more mathematical models |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102010028259A1 (en) | 2010-04-27 | 2011-10-27 | Robert Bosch Gmbh | A microcontroller having a computing unit and a logic circuit and method for performing calculations by a microcontroller for control or in-vehicle control |
-
2015
- 2015-03-10 DE DE102015204218.9A patent/DE102015204218A1/en active Pending
-
2016
- 2016-02-29 CN CN201610110437.XA patent/CN105975442A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1474046A (en) * | 2002-08-06 | 2004-02-11 | �����Զ�����ҵ��ʽ���� | Control device and its contro method for output power of internal combustion engine |
US20050038576A1 (en) * | 2003-08-12 | 2005-02-17 | Honda Motor Co., Ltd. | Control apparatus for hybrid vehicle |
CN101641511A (en) * | 2007-03-05 | 2010-02-03 | 丰田自动车株式会社 | Internal combustion engine controller |
US20130338975A1 (en) * | 2012-06-13 | 2013-12-19 | Hitachi, Ltd | Method for co-simulation of two or more mathematical models |
Also Published As
Publication number | Publication date |
---|---|
DE102015204218A1 (en) | 2016-09-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3349153B1 (en) | Convolutional neural network (cnn) processing method and apparatus | |
JP6507375B2 (en) | Battery state estimation device and battery state estimation method | |
KR102177721B1 (en) | Apparatus and Method for estimating deterioration of battery pack | |
JP6309690B2 (en) | Accelerator response method, apparatus, program, and recording medium | |
CN110290972B (en) | Method for estimating state of charge of battery cell | |
US10325036B2 (en) | Method and system for determing welding process parameters | |
JP6171897B2 (en) | Approximation function creation program, approximation function creation method, approximation function creation device, and charging rate estimation program | |
US20160355101A1 (en) | Method for calculating hydrogen consumption amount of fuel cell vehicle | |
US20140379284A1 (en) | System and method for calculating total available energy from vehicle battery | |
JP2017533458A5 (en) | ||
CN105975442A (en) | Method and device computing reversed function value of function model based on data | |
JP6215949B2 (en) | System and method for state-based test case generation for software verification | |
CN112861239A (en) | Method, system, device and storage medium for initially balancing ground stress of numerical model | |
Moslehian | Approximately vanishing of topological cohomology groups | |
Parriaux et al. | Nilpotent semigroups for the characterization of flat outputs of switched linear and LPV discrete-time systems | |
CN112528388A (en) | Method and device for analyzing strength of suspension bracket, storage medium and terminal equipment | |
CN105489954A (en) | Method for compensating charging state of battery unit and battery system for executing the method | |
CN105894481B (en) | Histogram equalization method and device for video monitoring equipment | |
Füredi et al. | A discrete isodiametric result: the Erdős–Ko–Rado theorem for multisets | |
Könnö et al. | Mixed finite element methods for problems with robin boundary conditions | |
CN113705795A (en) | Convolution processing method and device, convolution neural network accelerator and storage medium | |
JP2018084548A (en) | State estimating device of secondary battery and vehicle | |
US10871519B2 (en) | Fuel cell stack prediction utilizing IHOS | |
CN111179175A (en) | Image processing method and device based on convolutional neural network and storage medium | |
Gómez-Expósito et al. | Factored solution of infeasible load flow cases |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160928 |
|
RJ01 | Rejection of invention patent application after publication |