Method for analyzing the behavior of complex systems, especially internal combustion engines

• US 7,848,910 B2
• Filed: 07/21/2005
• Issued: 12/07/2010
• Est. Priority Date: 07/22/2004
• Status: Expired due to Fees

• Current Assignee: AVL List GmbH (List Privatstiftung)
• Original Assignee: AVL List GmbH (List Privatstiftung)
• Inventors: Ebner, Thomas, Schantl, Rainer, Vogels, Marie-Sophie
• Primary Examiner(s): Shah; Kamini S
• Assistant Examiner(s): DAY, HERNG DER
• Application Number:
  US11/658,092
• Publication Number:
  US 20070288213A1
• Time in Patent Office: 1,965 Days
• Field of Search: 703/2, 703/6, 703/7, 703/8
• US Class Current: 703/6
• CPC Class Codes: G05B 17/02 electric

Abstract

A method for analyzing the behavior of complex systems, particularly internal combustion engines, wherein a model is established which shows the dependence of test variables on input variables, calibrating the model based on test values of the real system obtained at the test points subdividing into at least two partial models a first principal influential parameter is identified for the first partial model, an optimal value of the first principal influential parameter is determined at each test point, the first principal influential parameter is interpolated for all plausible constellations of input variables to calibrate the first partial model, another partial model is established to show another subset of test variables in accordance with the input variables and the previously determined first subset of test variables, an additional principal influential parameter is identified for the other partial model, and an optimal value of the additional principal influential parameter is determined at each test point.

Claims

1. A computer-implemented method for analyzing behavior of a complex system by calculating a model which represents various measured variables as a function of input variables, comprising the following basic steps:

selecting various measured points which correspond to different constellations of measured variables and performing measurements to ascertain measured variables on a real system;

preparing the model which simulates dependence of the measured variables on the input variables and calibrating the model on the basis of the measured values of the real system obtained at the measured points;

subdividing the model into at least two partial models;

preparing one first partial model which simulates a first partial set of the measured variables;

identifying at least one first main influence parameter having greatest influence on a measured variable simulated by the first partial model;

determining an optimum value of each first main influence parameter in each measured point;

interpolating each first main influence parameter for all meaningful constellations of input variables to calibrate the first partial model;

preparing a further partial model to simulate a further partial set of the measured variables as a function of input variables and the previously ascertained first partial set of the measured variables;

identifying at least one further main influence parameter having greatest influence on a measured variable simulated by the further partial model;

determining an optimum value of each further main influence parameter in each measured point;

interpolating each further main influence parameter for all meaningful constellations of input variables to calibrate the further partial model; and

analyzing the behavior of the real system based on the calibrated partial models;

wherein said real system is an internal combustion engine and at least one of the following variables are used as the measured variables;

inflowing air mass,

indexed mean pressure,

maximum cylinder pressure, and

exhaust gas temperature before catalytic converter and/or turbine.

2. The method according to claim 1, wherein multiple partial models are calibrated in sequence, the measured variables ascertained in each partial model representing input variables for subsequent partial models.

3. The method according to claim 1, wherein precisely one measured variable is ascertained in each partial model.

4. The method according to claim 1, wherein a single main influence parameter is assigned in a measured variable and the calibration of the associated partial model is performed by full factorial variation.

5. The method according to claim 1, wherein the interpolation of each main influence parameter is performed as a linear interpolation.

6. The method according to claim 1, wherein the optimum value of each main influence parameter in the measured points is determined in that a setpoint value for the particular measured variable is calculated for various discrete values of the main influence parameter, at predetermined support points, an interpolation model is prepared from the setpoint values thus calculated, which indicates setpoint values for the particular measured variable for arbitrary values of the main influence parameter, a difference between the setpoint values and the measured actual value of the measured variable is then minimized, and the optimum value of the main influence parameter is concluded therefrom.

7. The method according to claim 1, wherein a measured variable of at least one partial model of the at least two partial models is obtained by the following steps:

selecting multiple first vectors, each of which represents a specific constellation of input variables and which cover a meaningful operating range of the real system;

simulating by using the prepared model to obtain computational values of a measured variable which are assigned to the first vectors;

selecting multiple second vectors, each of which represents a further constellation of input variables;

performing measurements to obtain experimental values of the measured variable which are assigned to the second vectors;

expanding each vector by one dimension by incorporating a block variable which is fixed at a first value for the first vectors and a second value for the second vectors, said first and second values being different from one another;

preparing a multivariate regression model which represents the measured variable as a polynomial function of the expanded vectors of the input variables, on the basis of the previously determined computational values of the measured variable and the experimental values of the measured variable;

determining at least one third vector which represents a constellation of the input variables for which the system is to be analyzed;

expanding the third vector by a block variable which is fixed to the second value; and

calculating the measured variable using the regression model having the expanded third vector as input variable.

8. The method according to claim 7, wherein the regression model is nonlinear.

9. The method according to claim 8, wherein the regression model is quadratic.

10. The method according to claim 7, wherein the number of the first vectors is greater than the number of the second vectors.

11. The method according to claim 7, wherein the input variables comprise speed and load.

12. A computer-implemented method for analyzing behavior of a complex system by calculating a model which represents various measured variables as a function of input variables, comprising the following basic steps:

selecting various measured points which correspond to different constellations of measured variables and performing measurements to ascertain measured variables on a real system;

preparing the model which simulates dependence of the measured variables on the input variables and calibrating the model on the basis of the measured values of the real system obtained at the measured points;

selecting multiple first vectors, each of which represents a specific constellation of input variables and which cover a meaningful operating range of the real system;

simulating by using the prepared model to obtain computational values of a measured variable which are assigned to the first vectors;

selecting multiple second vectors, each of which represents a further constellation of input variables;

performing measurements to obtain experimental values of the measured variable which are assigned to the second vectors;

expanding each vector by one dimension by incorporating a block variable which is fixed at a first value for the first vectors and a second value for the second vectors, said first and second values being different from one another;

preparing a multivariate regression model which represents the measured variable as a polynomial function of the expanded vectors of the input variables, on the basis of the previously determined computational values of the measured variable and the experimental values of the measured variable;

determining at least one third vector which represents a constellation of the input variables for which the real system is to be analyzed;

expanding the third vector by a block variable which is fixed to the second value; and

calculating the measured variable and analyzing behavior of the real system using the regression model having the expanded third vector as the input variable;

wherein said real system is an internal combustion engine and at least one of the following variables are used as the measured variables;

inflowing air mass,

indexed mean pressure,

maximum cylinder pressure, and

exhaust gas temperature before catalytic converter and/or turbine.

13. The method according to claim 12, wherein the regression model is nonlinear.

14. The method according to claim 13, wherein the regression model is quadratic.

15. The method according to claim 12, wherein the number of the first vectors is greater than the number of the second vectors.

16. The method according to claim 12, wherein the input variables comprise speed and load.