Modeling method

Abstract

A method for determining feature signal filters for preparing signal measurement data sequences of a plurality of measurement variables for experimentally determining a mathematical model which maps model measurement data for at least one target signal sensor (7) on the basis of detected measurement data of a plurality of feature signal sensors (5) is disclosed. Further disclosed is a method for determining a mathematical model (16) which maps model measurement data for at least one target signal sensor (7) on the basis of detected measurement data of a plurality of feature signal sensors (5), wherein training input measurement data sequences (19) ascertained by the feature signal sensors (5) are mapped onto at least one training output measurement data sequence (18) ascertained by at least one target signal sensor (7).

Claims

1.-10. (canceled)

11. A method for determining feature signal filters for preparing signal measurement data series of a plurality of measurement variables for experimentally determining a mathematical model (16) that maps model measurement data for at least one target signal sensor (7) based on detected measurement data of a plurality of feature signal sensors (5), comprising: recording feature signal measurement raw data series (1) with the feature signal sensors (5) using a data processing system; ascertaining feature signal measurement data series (2) from the feature signal measurement raw data series (1) using the data processing system; recording at least one target signal measurement raw data series (6) with the at least one target signal sensor (7) using the data processing system; ascertaining at least one target signal measurement data series (8) from the at least one target signal measurement raw data series (6) using the data processing system; ascertaining, in a frequency analysis step (3), a feature amplitude spectrum (4) by the data processing system for each feature signal measurement data series (2) by a frequency analysis method and ascertaining a target amplitude spectrum (9) for the at least one target signal measurement data series (8) by the frequency analysis method; dividing each feature amplitude spectrum (4) into a plurality of mutually adjacent or partially overlapping feature amplitude spectrum sections (10), wherein the feature amplitude spectrum sections (10) each comprise a manually or automatically predetermined feature frequency range; dividing the target amplitude spectrum (9) into target amplitude spectrum sections (11), wherein target frequency ranges of the target amplitude spectrum sections (11) correspond to the feature frequency ranges; ascertaining, in a match checking process (12) in a plurality of repetitive match checking steps, in each case a match measure for each feature amplitude spectrum section (10) by the data processing system, wherein the match measure is a measure for the matching of the amplitude spectrum of the respective feature amplitude spectrum section (10) and an associated target amplitude spectrum section (11); selecting the feature frequency ranges whose match measure exceeds a predetermined match measure number as selection signal frequency ranges (13) by the data processing system; and designing, in a subsequent determination step (14) for the selection signal frequency ranges (13), in each case a respective selection band pass filter (15) by the data processing system, such that signal measurement data series filtered with the respective selection band pass filter (15) have signal components lying within the respective selection signal frequency range (13) and signal components lying outside the respective selection signal frequency range (13) are filtered out of the filtered signal measurement data series, wherein each selection band pass filter (13) forms a feature signal filter for each feature signal sensor (5), with which the match measure between the feature amplitude spectrum section (10) of the feature signal measurement data series (2) recorded by the respective feature signal sensor (5) and the target amplitude spectrum section (11) in the feature frequency range associated with the respective selection band pass filter (13) exceeds the match measure number.

12. The method according to claim 11, further comprising: determining the match measure by a correlation analysis by the data processing system.

13. The method according to claim 11, wherein the band pass filter has a filter order of at least eight.

14. The method according to claim 11, wherein the feature amplitude spectrum sections (10) have a predetermined amplitude spectrum width.

15. The method according to claim 11, further comprising: ascertaining the feature amplitude spectrum sections (10) by the data processing system by dividing the feature amplitude spectrum (4) into two feature amplitude spectrum sections (10) in a first sub-step and determining a first match measure for each feature amplitude spectrum section (10); subsequently further dividing the feature amplitude spectrum sections (10) in each case into smaller feature amplitude spectrum sections (10) in further sub-steps and determining the match measure in each case; and dividing each feature amplitude spectrum section (10) into new feature amplitude spectrum sections (10) in the further sub-steps until an improvement of the match measure between a preceding sub-step and a current sub-step is smaller than a predetermined improvement value.

16. The method according to claim 11, wherein adjacent feature amplitude spectrum sections (10) overlap by a predetermined amplitude spectrum overlap width.

17. A method for determining a mathematical model (16) that maps model measurement data for at least one target signal sensor (7) based on detected measurement data from a plurality of feature signal sensors (5), comprising: mapping training input measurement data series (19) ascertained by the feature signal sensors (5) onto at least one training output measurement data series (18) ascertained by at least one target signal sensor (7); and filtering, by feature signal filters designed in accordance with claim 11, the training input measurement data series (19) using a data processing system and thereby forming training input data series (17), wherein a training input measurement data series (19) can be filtered with differently designed feature signal filters, such that a plurality of training input data series (17) are formed from a training input measurement data series (19), and wherein the mathematical model (16) is ascertained using the data processing system by a data-based model determination method (20) starting from the training input data series (17) as model input variables and the at least one training output measurement data series (18) as a model output variable.

18. The method according to claim 17, wherein the training input measurement data series (19) are formed by the feature signal measurement data series (2).

19. The method according to claim 17, wherein the feature signal measurement data series (2) have feature signal measurement data points directly following one another in time and thus in each case form a section of the associated feature signal measurement raw data series (1), wherein the sections of the feature signal measurement data series (2) have at least one predetermined minimum number of data points and wherein the section is selected by the data processing system such that a target signal power is maximum in the selected section.

20. The method according to claim 19, further comprising: ascertaining, in order to determine the target signal power for a target signal measurement raw data series (6), a short-term frequency amplitude spectrum by the data processing system for each target signal measurement raw data point; determining a short-term frequency amplitude power for each short-term frequency amplitude spectrum; combining target signal measurement raw data points following one another in time into target signal measurement raw data sections by the data processing system in such a way that a change in the short-term frequency amplitude power of target signal raw data points directly following one another in time is below a predetermined change power; and subsequently forming in each case a target signal power by the data processing system for all combinations of target signal raw data sections following one another in time that have the predetermined minimum number of data points; and selecting the combination as section by the data processing system whose target signal power is maximum.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0028] FIG. 1 is a schematic representation of feature signal measurement raw data series and target signal measurement raw data series and the feature amplitude spectra and target amplitude spectra ascertained by means of the frequency analysis method.

[0029] FIG. 2 is a schematic representation of training input data series used for the determination of the mathematical model as model input variables and the training output measurement data series as the model output variable.

DETAILED DESCRIPTION

[0030] FIG. 1 shows a schematic representation of feature signal measurement raw data series 1 and target signal measurement raw data series 2 and feature amplitude spectra 4 ascertained by a frequency analysis step 3. The feature signal measurement raw data series 1 are determined using feature signal sensors 5. Feature signal measurement data series 2 are ascertained from the feature signal measurement raw data series 1. The at least one target signal measurement raw data series 6 is determined using a target signal sensor 7. At least one target signal measurement data series 8 is ascertained from the target signal measurement raw data series 6. In the frequency analysis step 3, a feature amplitude spectrum 4 is ascertained for each feature signal measurement data series 2 by means of a frequency analysis method, and a target amplitude spectrum 9 is ascertained for each target signal measurement data series 8. The feature amplitude spectra 4 are each divided into a plurality of feature amplitude spectrum sections 10 that are adjacent to one another or partially overlap, in each case comprising a predetermined feature frequency range. The target amplitude spectrum 9 is divided into target amplitude spectrum sections 11, wherein target frequency ranges of the target amplitude spectrum sections 11 correspond to feature frequency ranges. In a match checking process 12, a match measure for each feature amplitude spectrum section 10 is ascertained in each of a plurality of repetitive match checking steps, wherein the match measure is a measure for the matching between the amplitude spectrum of the respective feature amplitude spectrum section 10 and the associated target amplitude spectrum section 11. The feature frequency ranges whose match measure exceeds a predetermined match measure number are selected as selection signal frequency ranges 13. In a subsequent determination step 14, in each case a selection band pass filter 15 is designed for the selection signal frequency ranges 13, such that signal measurement data series filtered with the respective selection band pass filter 15 have signal components lying within the respective selection signal frequency range 13 and signal components lying outside the respective selection signal frequency range 13 are filtered out of the filtered signal measurement data series.

[0031] FIG. 2 shows a schematic representation of training input data series 17 used for determining a mathematical model 16 as model input variables and the training output measurement data series 18 as model output variables. Using the designed selection band pass filter 15, the training input measurement data series 19 ascertained by the feature signal sensors 5, which can also be formed by the feature signal measurement data series 2, are filtered, by which training input data series 17 are formed. The training input measurement data series 19 filtered with the respective selection band pass filter 15 have signal components lying within the respective selection signal frequency range 13. Signal components lying outside the respective selection signal frequency range 13 are filtered out of the filtered training input measurement data series 19. In this case, a training input measurement data series 19 can be filtered with differently designed selection band filters 15, such that a plurality of training input data series 17 are formed from one training input measurement data series 19. By means of a data-based model determination method 20, the mathematical model 16 is ascertained starting from the training input data series 17 as model input variables and the at least one training output measurement data series 19 as a model output variable.