DE10328787A1 - Populating of an engine characteristic map for a combustion engine whereby discrete interpolation points are determined by carrying out regression calculations based on a measurement series of reference points - Google Patents

Abstract

Method for correlation-based populating of an engine characteristic map for the control unit of a combustion engine, whereby the characteristic map is divided into discrete interpolation points. The interpolation points are populated directly from measurement series for virtual reference points. For at least a fixed reference point, a regression calculation is carried out, whereby the available correlation points are treated as equally as possible. The invention also relates to a corresponding control unit for controlling a combustion engine, whereby the control unit has at least a discretized engine characteristic map.

Description

The The invention relates to a method for correlation-based calculation a discretized characteristic map, in particular in a control unit of an internal combustion engine and an operating according to this method control unit a Internal combustion engine according to the preambles the respective independent Claims.

In Modern automotive technology increasingly comes fuel metering systems used in which the injection of fuel into one Combustion chamber (cylinder) of an internal combustion engine by means of a High-pressure accumulator takes place, with the pressure generation and injection are decoupled from each other. A very common type of such Kraftstoffzumesssystems are the so-called "common rail systems", in which said High-pressure accumulator is referred to as "rail". The injection pressure becomes independent generated by the engine speed and the injection quantity and is in the high-pressure accumulator for the injection ready. The injection time and the injection quantity become calculated in an electronic engine control unit and injectors each cylinder of the internal combustion engine implemented via controlled valves.

at the manufacture of said injectors or other components the respective Kraftstoffzumesssystems about occurring manufacturing tolerances cause differences in operating characteristics of the various injectors of the fuel metering system. These differences occur especially over the Life of the injectors or the fuel metering system on or be during the Lifespan even enhanced. In this case, have the injectors of a single Kraftstoffzumesssystems in particular different quantity maps, i. different dependencies between the injection quantity, the rail pressure and the activation time the injectors. this leads to to that, despite a precise Control of the injectors every single injector the respective Combustion chamber with different amounts of fuel fills.

So goes out of the DE 102 15 610 A1 a method for correcting said injection behavior of injectors, in which the injectors are measured after production at characteristic operating points on their injection quantity and classified into classes. This class information is stored, for example, in memory means arranged at the respective injector, these values being read by the engine control unit from said storage means of the injector and being used in the subsequent operation of the internal combustion engine or of the respective fuel metering system.

The mentioned classification of the injectors can be done, for example. that the injectors at several test points regarding the injection metering checked become. Are the measured actual values at all test points within a predetermined tolerance window, so does the injector rated as "good" the actual value of a measuring point is used to determine the injectors in to divide a number of tolerance classes. The tolerance windows The respective classes are only valid at the respective checkpoint a fraction of the total tolerance. Because between the checkpoints only an insufficient one Correlation is, however, at the other checkpoints such a narrowing of tolerances at all not possible.

are the injectors installed on the internal combustion engine, so will the determined class affiliation in the control unit programmed. The control unit then leads a correction of the injection quantity by means of a correspondingly pre-rated map by. Due to the mentioned insufficient correlation between the operating points or the test points is a correction only in the range of one used for classification checkpoint possible. Furthermore Operating range can at best based on statistical mean value shifts between the classes a small adjustment of the quantity allocation respectively.

There is also the possibility, by means of the engine control unit individually for each injector, depending on the target amount and the rail pressure, to correct the driving time compared to a nominal map to come as close as possible to the target amount. For this purpose, the control unit receives during installation per injector several, preferably four test values (VL, EM, LL and VE) from the production. From these quantities, a correction quantity map based on discrete interpolation points is created (see 2 ). In this case, from the deviations of the injection quantities from their nominal values at the said four checkpoints, a quantity correction is carried out for a series of pressure / control combinations with the aid of which a correlation of the injection quantity to the injection quantity at the respective other test points is determined for each test point. Thus, the control unit can be filled with known values for the quantity deviations (.DELTA.VL, .DELTA.EM, .DELTA.LL and .DELTAVEVE) at the respective test points of the correction quantity map with numerical values.

From the DE 102 15 610 A1 goes further to determine the said correction amount by linear regression of multiple comparisons of the desired values with the actual values at the multiple test points of an injector. Furthermore, it describes the effects of measuring errors of the test values on injection by correlating several test points reduce it.

One special problem with the maps described is their Bedatung with the mentioned test values where the test values because of the high cost usually not determined directly from measurement series can be. Therefore, most of the interpolation points are mapped using correlations, focusing on a few, carefully select fixed points support. The selection of the fixpoints and their correlations currently takes place purely empirically.

Of the Invention is based on the object, a method and a control device of the initially to provide said type, which the said disadvantages of Avoid prior art and a precise and at the same time as possible automatable parameterization of the maps affected here.

These Task is solved by the characteristics of the independent Claims. Advantageous embodiments and developments are the subject of Dependent claims.

Advantages of invention

The peculiarity of the method according to the invention and of the control unit operating according to this method is that the determination of the interpolation points is optimizing on the basis of the so-called coefficient of determination R ² of regressions applied to the measured values, maximum values being required for the coefficient of determination. In the method, first all nodes are neutral (ie equivalent) combined and their measured values correlated with each other. In the case of the respective regression, evenly distributed fixed points of the characteristic field are determined completely freely and the best correlations are likewise assigned by means of regression according to the criterion maximum R ² .

The correlation calculation performed preferably on the basis of a linear regression of the measured values by means of a symmetrical correlation matrix supplies three result matrices with results from the correlations, one for the straight line slope k, one for the y-axis section b and one for the coefficient of determination R ² . The result matrices are each occupied on one side and have an unoccupied main diagonal. These matrices are effective due to their then present, described in more detail below symmetry and therefore time-saving evaluable. The necessary in the evaluation inversion of said matrices can advantageously be carried out according to simple rules. The result matrix R ² is exactly symmetric, so that a matrix element yx can be populated with the same value as an element xy. The result matrix for the straight line slope k can therefore be inverted according to the simple rule k_xy · k_yx = R ² .

The according to the invention, strictly methodical approach allows Advantageously, a very structured Kennfeldbedatung from the viewpoint the result optimization with a high degree of automation. Thereby arises opposite the prior art, a significant time savings and a clear reduced error potential, because the mentioned process steps by means of suitable calculation modules including the reading in of the data, the evaluation of the read data and the output of the calculated Data are fully automatable. In particular, the invention enables any number of map support points the most suitable bases free to determine and correlate the remaining reference points optimally set.

The Invention can be used with any type of map data, provided that the mapping to correlations of absolute values or Deviations (delta values) to fix points to support, such as on the the aforementioned field of automotive engineering and in particular in the field of there known methods of Nullmengenkalibrierung (NMK) and injector quantity matching (IMA), the latter in usually already before a vehicle delivery at low engine speed carried out in particular, the smooth running at idle the engine to optimize.

drawing

The Invention will be described below with reference to the drawings preferred embodiments described in more detail, from which further characteristics, characteristics and advantages of the invention.

In show the drawing in detail

1 a schematic representation of part of a common rail injection system according to the prior art;

2 a discretized quantity correction field with respect to the injection quantity of an injector as a function of the rail pressure in one in the 1 shown common rail injection system;

3 a discretized map suitable for correlation-based assessment according to the invention in a simplified representation;

4 a typical correlation matrix illustrating an inventive combination method used;

5 a typical result of a correlation calculation according to the invention by means of linear regression between two individual correlation points for a part lot;

6 a typical result matrix for the coefficient of determination R ² as a result of a correlation calculation according to the invention over all possible correlations;

7 a typical result matrix for the correlation constants as a result of a correlation calculation according to the invention over all possible correlations; and

8a and 8b a typical map data of an IMA map according to the invention.

description

In the 1 shows the high pressure part of a common rail accumulator injection system. In the following, only the main components and those components which are essential for the understanding of the present invention will be explained in more detail. The arrangement comprises a high-pressure pump 10 , which have a high pressure line 12 with the high-pressure accumulator ("rail") 14 communicates. The high-pressure accumulator 14 is connected to the injectors via additional high-pressure lines. In the present illustration are a high pressure line 16 and an injector 18 shown. The injector 18 is installed in the engine of a motor vehicle. The illustrated system is powered by a motor control unit 20 controlled. Through the engine control unit 20 in particular, a control of the injector 18 ,

At the injector 18 is a facility 22 provided for storing information which is customized to the injector 18 Respectively. The information in the facility 22 can be stored by the engine control unit 20 be taken into account, so that an individual control of each injector 18 can be done. Preferably, the information is correction values for the quantity map of the injector 18 , The device 22 for storing the information can be used as a data store, as one or more electrical resistors, as a barcode, by alphanumeric encryption or by one at the injector 18 arranged integrated semiconductor circuit can be realized. The engine control unit 20 may also be a semiconductor integrated circuit for evaluation in the device 22 have stored information.

In 2 a diagram for explaining the invention is shown. The diagram shows a quantity correction map MKK, wherein one of the injector 18 metered amount M is plotted against a rail pressure p _rail . The quantity correction map MKK is based on several injection points (VL, EM, LL, VE). The adjustment values .DELTA.VL, .DELTA.EM, .DELTA.LL and .DELTA.VE are used for quantity correction M, which are determined by comparing desired values with actual values at different rail pressures p _rail at different test points. If necessary, a correction factor KW _{(n) is} assigned to the adjustment values ΔVL, ΔEM, ΔLL and ΔVE. By way of example, the injection quantity M at a test point P is assigned the adjustment value ΔEM as a function of a pressure (rail pressure / activation duration combination) of the injection EM, from which a correction quantity ΔQ _(n) for the control unit in the respective test point is determined. The calculated correction amounts ΔQ _(n) are based on the adjustment values, which are determined from quantity deviations ΔVL _{Abw. (N)} , ΔEM _{Abw. (N)} , ΔLL _{Abw. (N)} and ΔVE _{Abw. (N)} in the respective test points, and the associated determined correction factors KW _(n) . In 2 For example, a correction factor KW _{(n) is} assigned to the test point P ΔEM.

It can also be seen that numerous test points P for an injector 18 may be provided, which arise over the entire operating range and the quantity correction map MKK. Between the reference points defined by test points P, the adjustment values can also be linearly interpolated so that ultimately a reliable fuel quantity measurement can be carried out over the entire operating range.

starting point for the described below correlation-based calculation of a discretized Kennfeldes is a series of measurements for a parts lot, from which measured values for support points of the to be optimized Map result. The measurement series is systematically evaluated where all available in the map reference points equivalent to each other, to the underlying To correlate measured values with each other. This is in particular Fixed points and their correlation areas within the map established.

That in the 3 shown embodiment of a map comprises thirty (5 · 6) discretized nodes, which are presently shown in the form of grid points. The arrangement of the interpolation points is carried out according to appropriate specifications, for example. After an equidistant requirement or according to a requirement, after which the interpolation points have to meet a certain accuracy specification. The four blackened support points 300-315 represent support points for the parts lot of an existing measurement series, in the present case referred to as reference points "RP". In contrast, the remaining, not blackened grid points represent by means of a correlation calculation according to the invention still to be used, presently as correlation points "KP" designated support points. The correlation calculation preferably takes place by means of the correlations between all KPs and in each case a fixedly assumed RP on the basis of the measured values of the underlying measurement series.

The 4 shows a typical correlation matrix for illustrating the applied combination method, according to which all nodes of an underlying map are initially considered as KPs and the KPs are correlated with each other so that all possible or necessary combinations are taken into account. Since, due to the correspondence of a correlation KPx-KPy with its mirror-symmetric correlation KPy-KPx, each of the combinations KPx-KPy has to be considered only once and, moreover, correlations of the form KPx-KPx, ie correlations between identical KP values, are of no significance here the correlation matrix is a square, with respect to the main diagonal (HD) symmetric matrix, the HD is unoccupied. For KPs of the number n, n / 2 · (n-1) combination possibilities thus result in a manner known per se.

The on the basis of in the 4 In the present exemplary embodiment, correlation calculation takes place by means of linear regression of the measured values at said support points, resulting in three result matrices, namely a matrix for the resulting straight line slopes k, a matrix for the resulting y-axis sections b and one of the resulting coefficients of certainty R ² having matrix. The evaluation of the linear regressions carried out is based on the criterion "maximum R ² ", ie on the maximum frequency or the maximum sum value, uniformly distributed fixed points of the characteristic field being determined completely independently (free) and the best correlations according to the criterion "maximum R ² "assigned. To check the results, the matrix with the y-axis sections can also be used. Finally, the mapping of the map is done, as described in more detail below, with the resulting values of the k-matrix corresponding to the optimally selected correlations.

by virtue of the shown matrix symmetry and the associated relative These matrices are effective and therefore resource-efficient. and time-saving evaluable.

In the 5 is a typical result of a previously described linear regression-based correlation calculation between two in the 4 represented correlation points shown. Correlation values that result after passing through all the possible combinations for the present parts lot using the example of a combination for the injector quantity adjustment (IMA) of a ballistic injector are entered in the diagram. The correlation here relates to the two correlation points KP10 and KP11 of a lot of forty-two parts, which are correlated by the linear regression.

The 6 Figure 9 shows a typical graphically represented correlation matrix resulting from the correlation calculation at all correlation points (twenty-two correlation points in the present example). In the correlation matrix, the R ² values resulting from the above-described correlation calculation are plotted in the form of contour plots, with the value ranges shown in the legend shown on the right. The correlation matrix shown also has the already in the 4 shown symmetry. The R ² values give a measure of the quality of a respective correlation.

The in the 6 Contours shown are limited to certain correlation areas, since in the evaluation a threshold R ² _{limit is} specified and values R ² <R ² _{limit are} not taken into account in the correlation matrix. As we can see from the legend for the contours, the value for R ² _limit is 0.6. At present twenty-two KPs there are a total of 231 usable combinations KPx-KPy. In the correlation matrix, two evaluation lines extending in the two coordinate directions are also drawn, which in the present case illustrate for the KP13 the procedure for the correlation evaluation, according to which, assuming fixed KP13, its correlation with all other KPs is evaluated.

The 7 shows a typical result field, where the already mentioned correlation constant k, namely the linear regression resulting from the linear regression, exceeds that in the 6 shown matrix for the correlation points KPx and KPy. On the basis of these values for the correlation constant, the actual parameterization of the underlying characteristic field now takes place. For the best correlations KPx-KPy selected from the evaluation according to the criterion "maximum R ² " with the present four fixed points KPx, the matrix elements k_xy are read out of the result matrix for the straight line slope k and stored as values or Factors used for the metering in Mengenkorrekturkenntfcld (MKK) directly. If, as described, the point KPx is not suitable as a fixed point, but KPy, then due to the symmetry of the correlation matrix, the correction constant does not have to be recalculated, but only inverted by the relevant support point xy into the support point yx, in accordance with the rule k_xy · K_yx = R ² . Therefore, this method allows a resource-conserving control of the MKK with the factors k of the best correlations, selected from the totality of all weighting factors k, as in the 7 shown.

Based on the tables 8a and 8b, a real re-data of an IMA adjustment map of an internal combustion engine will now be illustrated by means of the method according to the invention, wherein in the 8a the result of a condition according to the prior art and in the 8b the result of a Bedatung according to the invention can be seen. Registered in the two tables are the coefficients of determination R ² , which represent a measure of the quality of the adjustment to be made of the quantity correction. The 8a shows according to the prior art previously fixed preset fixed points that are not selected in your position according to an optimization criterion. Assigned to them are correlation points whose coefficient of determination R ^{2 has} significantly lower values than at the same point in the equation 8b because there the assignment is done purely empirically and not methodically optimized. The 8b contains determinants that fulfill the criterion "maximum R ² " in the sum or frequency, from which one can directly derive the four fixed points that are best suited as correlations for correlations in your interpolation environment and, on the other hand, the demarcation of the correlation regions itself, so that only the qualitatively best correlations in the correction map (in the present case the IMA adjustment map) are used.

By comparing the two tables results on the one hand, that the VE 'point shown undergoes such a shift due to the invention according to the invention that small injection quantities of the present injector, as in a pilot injection, VE' are attributed, can be displayed accurately. By shifting the fixed point VE to higher pressure, small quantities can be corrected even at high pressures of good quality (ie maximum R ² ), while maintaining a sufficient correction quality even in the middle pressure range (eg 800 bar). On average, the quality of the comparison is thus increased, thus proving that the ballistic injector in the middle pressure range (ie the mentioned 800 bar) over the entire range from the small amount to the maximum amount with a fixed point, EM 'and sufficient quality durchkorreliert or . can be corrected.

To the Another causes the shift of the 'VL' point to be seen into one Middle position, that at medium rail pressure values one compared to the state the technique improved quantity correlation takes place. From the application of the invention the optimized map data also results in a shift of the 'VL' point away from a raildruckseitigen edge layer (in this case 1600 bar), so throughout the high load range (i.e., at high rail pressure and high Injection quantity) in the sum or on average a better quality of the quantity correction he follows. The acceptable quality loss at full load (i.e., at maximum rail pressure and maximum injection quantity) in favor of a significant increase in quality at maximum rail pressure and medium injection quantities.

The The method described above is preferably in the form of a program or control codes either at a control station for an engine control unit Internal combustion engine or in such a control device itself realizable.

Claims (9)

Method for the correlation-based calculation of a discretized map, in particular in a control unit of an internal combustion engine, characterized in that the map is divided into discrete nodes, where from an existing series of measurements for a Teilelos reference points are directly fedat and wherein for at least one fixed assumed reference point, a regression calculation is performed , where the available correlation points are treated as equally as possible. A method according to claim 1, characterized in that the regression is performed on the basis of a maximum coefficient of determination R ² . Method according to claim 2, characterized in that that the regression calculation based on a linear Regression occurs. Method according to Claim 3, characterized in that the linear regression is carried out by means of a symmetric correlation matrix and supplies three result matrices, one for the straight line slope k, one for the y-axis section b and one for the coefficient of determination R ² , each occupied on one side are and have an unoccupied main diagonal. Method according to one of the preceding claims, characterized in that the Re is performed on the basis of a correlation matrix providing all possible combinations of the correlation points. Method according to claim 5, characterized in that that the correlation matrix is arranged symmetrically to the main diagonal is, with the main diagonal is blank. Method according to one of the preceding claims, characterized characterized in that the division of the interpolation points according to a specification, preferably according to the specification of the equidistance or the gradient dependence, he follows. Method according to one of the preceding claims, characterized characterized in that in the regression calculation fixed assumed Reference points over the support points preferably equally distributed become. control unit for controlling an internal combustion engine, wherein the control unit at least a discretized map, characterized in that the mapping of the map is correlation-based according to the method according to claim 1 takes place.