Method for adjusting correction values ​​for use in fuel metering

Abstract

The invention relates to a method for adjusting correction values ​​for use in metering fuel, by means of at least one fuel injector, from a high-pressure accumulator into a combustion chamber (165) of an internal combustion engine, using a learning data set (200) and a correction data set (260), the method comprising: in the correction data set (260), adjusting the correction values ​​(264) based on the learning values ​​(204) of the learning data set (200), wherein the correction value in each active neighboring field is adjusted based on an average neighboring field learning value, wherein the correction value in each field of each field area comprising an active field is adjusted based on an average field learning value, wherein the correction value in each inactive field is adjusted based on a transfer learning value.wherein the transfer learning value is or has been determined according to at least one correlation rule based on the learning values ​​of the active neighboring fields; and providing the correction card (260) with the adjusted correction values ​​for further use.

Description

The present invention relates to a method for adapting correction values for use in metering fuel, by means of at least one fuel injector, from a high-pressure accumulator into a combustion chamber of an internal combustion engine, and to a computing unit and a computer program for carrying it out.BACKGROUND OF THE INVENTIONFor clean, i.e. as low-emission as possible, operation of internal combustion engines, such as diesel engines, for example, the metering of fuel into combustion chambers of the internal combustion engine should always remain as accurate as possible over time or its service life.Disclosure of the InventionAccording to the invention, a method for adapting correction values and a computing unit and a computer program for carrying it out are proposed with the features of the independent patent claims. Advantageous embodiments are the subject matter of the dependent claims and of the following description.The invention is concerned with internal combustion engines such as diesel engines and their clean operation. In particular, internal combustion engines with a high-pressure accumulator are considered here, in which the fuel from the high-pressure accumulator is introduced into the combustion chambers or cylinders by means of fuel injectors. Reference is also made here to so-called common-rail injection systems. The accuracy of the fuel supply, i.e. the fuel quantity to be metered, should be within certain limits over the entire service life in order to meet, for example, the customer and legal requirements.In general, a desired value can be preset for a fuel quantity to be metered by means of a fuel injector in an introduction or injection process, the desired quantity. The fuel injector can then be controlled accordingly in order to meter this setpoint quantity. The quantity actually metered or introduced in this case, the actual quantity, can however deviate from the desired quantity in this case. Reasons for this can be, for example, the so-called specimen scattering, i.e. small deviations between different fuel injectors or else age-related changes. The target amount can therefore be corrected based on a correction value.The deviation between the actual quantity and the desired quantity can, however, change over time, i.e. a correction value should likewise be adapted over time.For this purpose, so-called virtual sensors can be used which detect (or estimate or calculate) the actual flow rate or flow rate (through the fuel injectors), in order to be able to carry out a correction in this way. A direct measurement is generally not possible or too complex. Such virtual sensors can use a fuel pressure in the high-pressure accumulator (the so-called rail pressure signal) detected by means of a sensor in order to calculate the corresponding value. It is therefore possible to determine a calculated actual quantity of fuel to be metered, which would result for a specific desired quantity.The calculation may be based on, for example, trained data models or cause-effect physical relationships, e.g., machine learning models. The correction value can be determined, for example, by comparing the calculated actual quantity with the desired quantity for the current operating point. The correction value is typically filtered by a factor to reduce the noise of the sensor. Such a correction value filtered by a factor can then be a so-called learning value.In view of this, an improved way of adjusting correction values for use in fuel metering is proposed. For this purpose, a learning data set and a correction data set are used.The learning data record and the correction data record each comprise a mutually corresponding multiplicity of fields (i.e. data fields), wherein each of the multiplicity of fields is assigned, for example, in each case to a pressure of fuel in the high-pressure accumulator and to a quantity of fuel to be metered, i.e. in each case to an operating point. Instead of specific pressure values, pressure ranges may also be used, and instead of specific amounts, amount ranges may also be used. However, a different assignment can also be made, for example if fuel properties are to be calculated on the basis of the rail pressure.The learning data set comprises one or more field regions, wherein the one or each of the plurality of field regions comprises in each case one or more, in particular contiguous, of the plurality of fields. However, each of the plurality of fields is assigned to one field area, specifically in particular exactly one field area. In other words, a field area may correspond to one field, but two or more fields may also be combined to form a field area; such combined fields may then be considered together, as will be explained below.In the learning data set, each of the plurality of fields is assigned a learning value and a status, and in the correction data set, each of the plurality of fields is assigned a correction value. The term learning map is also to be used below for the term learning data set and the term correction map is to be used below for the term correction data set. These are slightly more illustrative terms, a map can be displayed visually, for example, with the fields in the manner of a matrix. ultimately, however, the learning data set and the correction map are each a data set with corresponding values. This applies correspondingly to other types of cards which are mentioned below.The status of each of the plurality of fields of the learning card is in each case a status from a status list, wherein the status list comprises at least the following states: active field, active adjacent field, inactive field, and in particular inactive field in a field area. However, further states can also be provided.An active field is a field in which the learning value is such that adjustment is to be made here. In one embodiment, it is provided that, if one or one of a plurality of predefined activation criteria is present, the status of a field is changed to an active field. Such an activation criterion may be considered, for example, that a number of learning events (generally, i.e. together for all fields) is higher than a predefined threshold value. It is also conceivable that, for example, a number of learning events for an individual field is higher than a predefined threshold value. Or else, for example, that a time duration or travel distance, for an individual field, since a comparison time or a comparison travel distance is greater than a predefined threshold value. For example, it can be provided that an adaptation should always take place every two months, or every 1000 km. It is also conceivable that a number of operating cycles of the internal combustion engine is counted, and if this number exceeds a threshold value, an activation criteria is considered to be present.The virtual sensor can supply continuously calculated values (correction values). Due to various boundary conditions, however, this is generally valid (or trained) only for specific or defined ranges and boundary conditions and is sufficiently accurate there. As mentioned, criteria for the validity of learning values can be used (so-called learning enables). If all relevant criteria or learning enables are fulfilled, the learning value can be written into the relevant field. This can then be considered a learning event.In one embodiment, it is provided that, if one or one of a plurality of predefined deactivation criteria is present, the status of a field is changed to inactive field. This can be, for example, after a predefined time period or travel distance in which no learning event has occurred for the relevant field since the change to active field.An active adjacent field is a field that adjoins a field region in which an active field is present. If a field becomes an active field, it can also be provided at the same time that the status of the active adjacent field applies to adjacent fields of the field area which comprises an active field. Thus, for example, if a field is set active, respective neighbor fields become active neighbor fields unless they are already. If a field is set to inactive, the adjacent fields can become an inactive field if they are not also adjacent fields of other active fields or, if appropriate, of a field region or correlation rule apply, as will be described below.An inactive field is a field that is inactive and that is not an active neighbor field, and that is also not part of a field area in which there are one or more active fields. An inactive field in a field region is then a field in a field region that is not active.Based on the learning values in the learning map, the correction values in the correction map are then adjusted. This is effected in this case as a function of the respective status of the learning values; in this respect, the status serves only to decide how the learning values are used.The correction value in each of the active adjacent fields is adjusted based on a mean adjacent field learning value, respectively. The mean neighbor field learning value is determined based on the learning value of the one or more active neighbor fields. It can be an arithmetic mean value, for example. If there is only one adjacent field, the average adjacent field learning value corresponds to the adjacent field learning value.The correction value in each field of each field region, which comprises an active field (optionally also a plurality of active fields), is adjusted based on a mean field learning value. The average field learning value is determined based on the learning values of the active fields of the field region. It can be an arithmetic mean value, for example. If there is only one active field in the field area, the average field learning value corresponds to the field learning value. In particular, the correction values of those fields which are inactive fields in the field range are thus also adapted in this way.The correction value in each inactive field is adjusted based on a transmission learning value. The transmission learning value is determined according to at least one correlation rule based on the learning values of the active neighbor fields. This can be, for example, a correlation rule applicable to new parts, or a correlation rule applicable to parts arriving at the end of a lifetime. The correlation rules can also be calibrateable. Thus, for example, a value can be determined as a function of a running performance of the internal combustion engine from or according to both correlation rules.It is conceivable that before the adaptation of the correction values, the learning values are transmitted according to the above rules into an adaptation map (which also comprises the plurality of fields). The correction value can then be adjusted in each field based on the learning values in the adaptation map.With the proposed procedure, therefore, not only one or a few correction values are adapted, but very many. In particular, correction values can also be adapted in this way, the operating points of which are reached rarely or never. Certain changes, for example age-related changes of the fuel injectors, however, have also proven to have an effect at other or even all operating points. If such operating points are then reached later, an accurate metering of fuel can also take place there.The correction map with the adjusted correction values is then provided for further use.In one embodiment, the fields that the one or more field regions comprise are adjusted as needed. For example, fields having similar learning values can be combined in order to thus enable simpler adaptation of the correction values, especially since it can then be assumed that changes also affect these other fields. If it is found that learning values in fields still change again over time, the summary can be adapted again to the field area.In one specific embodiment, an adaptation of a setpoint quantity predefined for the metering of fuel based on the correction values includes the determination of an interpolated correction value, specifically based on at least two correction values (or fields) which are adjacent to the setpoint quantity and the current pressure of the fuel in the high-pressure accumulator (i.e., the current operating point). The interpolated correction value is then used for the adjustment.As already mentioned, there is one correction value per field, namely in particular exactly one correction value. When a field corresponds to a specific operating point with pressure and quantity, it may and will occur that a quantity to be measured is to be corrected at an operating point that does not exactly correspond to one of the correction map. The interpolation (e.g. linear) thus allows a particularly simple and accurate correction; this is in particular simpler and more accurate than if e.g. a plurality of correction values were provided per field. In addition, a uniform transition between operating points is achieved thereby.In one embodiment, for determining a respective learning value, it is provided that an estimated actual quantity is determined based on a setpoint quantity predefined for the metering of fuel, in particular using a machine learning model. An adjusted actual quantity is then determined based on the estimated actual quantity and the respective correction value. A deviation amount is then determined based on the adjusted actual amount and the target amount. The deviation amount is then adjusted based on at least one correction amount, and the learning value is determined based on the adjusted deviation amount and a learning factor. The learning factor in this case indicates, for example, a speed at which the correction factor is learned or adapted. The smaller the learning factor, the smaller the learning value based on a certain correction factor, and the smaller the adjustment of the correction factor fails.It is also possible to expand the correlation maps and to separate the adaptation map (or the determination of the correction values) according to different fuel injectors or fuel injector characteristics. Thus, for example, a division into ballistic and nonballistic ranges or into specific quantity levels can be carried out on the basis of the operating points (for example pilot quantity, partial load, high load range).With the proposed procedure, correction values can therefore also be learned in other regions in which the vehicle is operated. The drift or correction requirement due to the causes (e.g. wear) typically occurs in the entire map area. This is not taken into account with previous concepts. In addition, although learning values have been created, the learning values are now expired because the area has not been used for a long time. Also, in current concepts, some areas are discarded by dominant areas. The virtual sensor used to detect the flow rate or quantity may have an inhomogeneous topology via the correction map. This has an additional tolerance contribution and negative influence on the adaptation, which can be taken into account with the proposed procedure.A computing unit according to the invention, e.g. a control device of a motor vehicle, is configured, in particular by programming, to carry out a method according to the invention.The implementation of a method according to the invention in the form of a computer program or computer program product with program code for carrying out all method steps is also advantageous since this causes particularly low costs, in particular if an executing control device is also used for further tasks and is therefore present in any case. Finally, a machine-readable storage medium is provided with a computer program stored thereon, as described above. Suitable storage media or data carriers for providing the computer program are, in particular, magnetic, optical and electrical memories, such as hard disks, flash memories, EEPROMs, DVDs, among others. Download of a program via computer networks (Internet, intranet, etc.) is also possible. Such a download can be effected in a wired or wired or wireless manner (e.g. via a WLAN network, a 3G, 4G, 5G or 6G connection, etc.).Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawing.The invention is schematically illustrated in the drawing on the basis of an exemplary embodiment and is described below with reference to the drawing.Brief Description of the DrawingsFIG. 1 schematically shows an internal combustion engine with a common rail system, which is suitable for carrying out a method according to the invention. FIGS. 2 and 3 show a sequence of a method in one embodiment.Embodiment(s) of the InventionFIG. 1 schematically shows an arrangement 100 with an internal combustion engine 160 which is suitable for carrying out a method according to the invention. By way of example, the internal combustion engine 160 comprises three combustion chambers or associated cylinders 165. Each combustion chamber 165 is assigned a fuel injector 170 which is in turn connected to a high-pressure accumulator 175, a so-called (common) pressure accumulator, and via which it is supplied with fuel. It is understood that a method according to the invention can also be carried out in an internal combustion engine having any other number of cylinders, for example one, two, four, five, six, eight, ten or twelve cylinders, etc.Furthermore, the high-pressure accumulator 175 is supplied with fuel 197 from a fuel tank 195 via a high-pressure pump 161. The high-pressure pump 161 is coupled to the internal combustion engine 160, specifically, for example, in such a way that the high-pressure pump is driven via the internal combustion engine.The fuel injectors 170 are controlled for metering or injecting fuel into the respective combustion chambers 165 via a computing unit designed as an engine control unit 180. For the sake of clarity, only the connection from the engine control unit 180 to a fuel injector 170 is shown, but it is understood that each fuel injector 170 is correspondingly connected to the engine control unit. Each fuel injector 170 may be specifically controlled. Furthermore, the engine control unit 180 is configured, for example, to detect the fuel pressure in the high-pressure accumulator 175 by means of a pressure sensor 190.FIG. 2 shows a sequence of a part of a method in one embodiment, specifically adaptation of correction values based on learning values. How learning values can be obtained is shown in FIG. 3. If maps are mentioned here, this can be understood in particular as data sets which contain the corresponding data.Here, a learning map 200 and a correction map 260 are shown. The learning map comprises a plurality of fields, wherein, by way of example, each of the plurality of fields is associated in each case with a pressure of fuel in the high-pressure accumulator and with an amount of fuel to be metered. The pressure is plotted as axis 210 and the quantity is plotted as axis 212 next to the learning map 200. By way of example, a field 202 is shown in the center of the learning map 200.The learning map 200 can comprise, for example, m x n fields, where m indicates the number of different pressures and n indicates the number of different sets; a matrix is thus obtained, each field being associated with an operating point with pressure and set.The correction map 260 also includes a plurality of fields corresponding to the learning map. By way of example, a field 262 is shown in the center of the correction map 260.The learning map can comprise one or more field regions, wherein the one or each of the plurality of field regions comprises in each case one or more, in particular contiguous, of the plurality of fields. Each of the plurality of fields is associated with a field area. It is conceivable that each field corresponds to a field area, but two or more fields can also be combined to form a field area. This can be adapted, for example, if required. For this purpose, a configuration map 214 is shown by way of example, on the basis of which the learning map or the field regions thereof can be adapted. Illustratively, nine fields are shown to the right above the training map 200 (as part 200' of the training map 200), including the field 202, where four of the fields are merged into a field region 206, i.e., the field region 206 includes four fields.The configuration card 214 can be used to configure the learning card 200 once. For a specific motor project, this configuration will generally not change over the lifetime.An engine map may include areas that have equal correction demands. These areas or the corresponding fields in the learning map can then be combined to form a field area. Both cards 214 and 220 can have the same size for this purpose. Fields in the map 214 may also be configured with "0"; such fields are then excluded from learning; the corresponding fields in the learning map are not used for learning, i.e., for example, are not set to active fields. However, such fields can be taken into account as an active adjacent field or otherwise with correlation rules. However, it is also conceivable that such fields are generally not taken into account, i.e. also not in the correction map.In the learning map 200, each of the plurality of fields 202 has a learning value associated therewith, here exemplarily denoted 204. In addition, a status is assigned to each field in the learning card 200, here denoted by 222 in a status list 220 by way of example. The status of each field is a status from the status list 220, wherein the status list comprises the following states: active field, active neighbor field, inactive field, inactive field in a field area. In the correction map 260, a correction value is assigned to each of the plurality of fields 262, denoted here by 264 by way of example. There may also be other states, namely how the transmission values are determined for the inactive fields. This may be different, for example, starting from a field region with an active field, depending on whether the method is moved to the left, right, upward or downward (in the map). In addition, according to the above statements, one or more fields may be ignored, or fields may be deactivated, but taken into account for a transmission value.In the correction map, the correction values are now adjusted based on the learning values of the learning map. Then, the correction map with the adjusted correction values is provided for further use.The states of the fields in the learning map 200 may change. For example, if one or more predetermined activation criteria 224 are present, the status of a field is changed to an active field; this may be the case, for example, if a number of learning events is greater than a predetermined threshold. For fields adjacent to the field area that comprises an active field, the status of the active adjacent field then applies. Thus, for example, if (only) field 202 is or becomes an active field, the fields of portion 200' that are outside field region 206 are, for example, active neighbor fields. The fields of the portion 200' that are within the field area 206 except for the active field 202 may then be inactive fields in the field area. All other fields (not shown here) may be inactive fields.If the correction values are now adjusted in the correction map 260, the learning values of the learning map 200 can be used. For this purpose, the learning values, optionally with adaptation, can first be transmitted into an adaptation map 250 (which has the plurality of fields corresponding to the learning map and correction map). In the adaptation map 250, there is then a (optionally adapted) learning value for each field, based on which learning value the correction value in the corresponding field of the correction map 260 is adapted.For this purpose, in each of the active adjacent fields, an average adjacent field learning value can be generated in the adaptation card 250 in each case. The mean neighbor field learning value is determined based on the learning values of the plurality of active neighbor fields. In other words, learning values of the learning map 200 are thus averaged in the active adjacent fields, and this average value is entered in the adaptation map 250 in each active adjacent field.In each field of each field area that comprises an active field, i.e. for example in each of the four fields of the field area 206, a respective average field learning value is generated in the adaptation map 250. The mean field learning value is determined based on the learning values of the active fields of the field region, i.e., corresponds here to the learning value 202, for example. In the specific example, the learning value 202 is thus entered in the adaptation map 250 into each field of the field area 206.This carry of learning values with possibly averaging for field areas with active fields and also active adjacent fields can be effected, for example, on the basis of an averaging adaptation map 230.In each inactive field, a transmission learning value is generated in the adaptation card 230 or then in the adaptation card 250 respectively. The transmission learning value is determined, for example, using a correlation rule 240 and a correlation rule 242, in each case based on the learning values of the active adjacent fields. The correlation rule 240 may apply to new parts, the correlation rule 242 may apply to parts arriving at the end of a lifetime.A factor can be assigned to all or even just some fields via the correlation rule(s), for example. If a transmission value is now to be determined starting from a field (output field) for which an (adapted) learning value is present (this can be a neighbouring field or an inactive field for which a transmission learning value is already present), for an adjacent field (at least if a factor is present there), this can be carried out on the basis of the factors of the two fields concerned and the (adapted) learning value or transmission value of the output field, for example via a quotient of the two factors. If the factor in the adjacent field is smaller than in the output field, the transmission value will become smaller there and vice versa.In the case of two correlation rules (new parts and end of the lifetime), which each specify different factors for the fields, it is possible to interpolate between these two correlation rules. Such interpolation can be carried out, for example, depending on the number of learning values, the km reading, the operating hours of the internal combustion engine or otherwise.Based on the learning values thus generated (adjusted) in the adaptation map 250, the correction values can then be adjusted. In particular, therefore, all correction values of the correction map are also adapted in the case of only one active learning value in the learning map. It is also conceivable that the (adjusted) learning values generated (adjusted) in the adaptation map 250 are used directly as correction values. In this case, the adaptation map 250 can be used as a correction map, or the generated (adapted) learning values are not generated in a separate adaptation map 250, but directly in the correction map.This process of adapting the correction values can be carried out repeatedly, e.g. regularly at predetermined time intervals, or depending on the route traveled by the relevant vehicle or also on the basis of other criteria. For example, each time the one or more activation criteria are met, an operation of adjusting the correction values may be performed. This can then be associated with the change of a field to be active, for example, unless the relevant field is still active in any case. It is also conceivable for this to take place at predetermined time intervals of, for example, 100 ms, defined on the engine control unit, for example.In order to adapt a setpoint quantity predefined for the metering of fuel on the basis of the correction values, an interpolated correction value can then be used, for example. In the part 260' of the correction map, four correction values are shown by way of example, which are each associated with a pressure and a quantity. However, if the target amount does not now correspond to an amount of the correction values, but lies between the correction value 262 and the adjacent correction value 262', for example, the interpolated correction value 272 can be determined based on the correction values 262 and 262'. Such an interpolation between two correction values can be effected linearly, for example. These are adjacent correction values or fields. If the current pressure does not correspond to one of the correction values, an interpolation can be carried out between two other correction values or else three or four correction values. Here, it is then possible, for example, to try to interpolate between three or four points.FIG. 3 shows how learning values can be obtained, as can be used, for example, in the learning map 200 according to FIG. 2. For this purpose, actual quantities of a desired quantity of fuel to be metered at a current pressure in the high-pressure accumulator can be estimated in each case for different operating points. These estimated actual amounts may be entered into, by way of example, each of two adaptation maps, a global adaptation map 300 and a local adaptation map 310.These operating points may be divided according to the plurality of fields of the learning map and the correction map (see FIG. 2 ). In the adaptation map 300, an operating point is denoted by 304 by way of example. By way of example, each of the operating points is assigned in each case to a pressure of fuel in the high-pressure accumulator and to a quantity of fuel to be metered. The pressure is plotted as axis 310 and the quantity is plotted as axis 312 adjacent to the fit map 300. The same applies to the adaptation map 310.The global adaptation map 300 and the local adaptation map 310 differ in resolution, for example. Thus, the global adaptation map 300 may include more fields than the local adaptation map 310. In this way, inhomogeneities (i.e., for example missing information for specific fields) of the virtual sensor can be taken into account by first using the global adaptation map to correct errors. Remaining fields may then be corrected with the local adaptation map. The shading required for this can take place once and remain constant over the lifetime.Here, an estimated actual quantity 302 is now determined for the operating point 304 or ultimately for each operating point. As mentioned, this is done, for example, using a machine learning model or a virtual sensor. This estimated actual quantity 302 can be referred to as Q sensor= f (Q des, p rail) and is therefore a function of the desired quantity Q des( axis 312) and the pressure in the high-pressure accumulator p rail( axis 310).Using the adaptation maps 300, 310, a corrected or adapted actual quantity 320 or Qsensor_adjustedbestimmt is calculated from the estimated actual quantity 302; thus, errors of the virtual sensor are corrected in particular with this. From this, the desired quantity Q then becomes des or. 322 so as to obtain a deviation amount Q deviation= Q des- Qsensor adjusted, referred to as 330.The deviation quantity 330 or Q deviation is then additionally corrected with a first correction quantity 342 from, for example, a first correction quantity map 340 and subtractively with a second correction quantity 352 from, for example, a second correction quantity map 350. Both the first correction quantity and the second correction quantity can each depend on the operating point (this can be effected via the respective correction quantity map).With the first correction quantity, it is possible, for example, to take account of when the current operating point does not exactly correspond to the midpoint in the relevant field at which the correction value is present; here, interpolation can take place in the direction of the adjacent field. The second correction quantity can be used, for example, to take into account which value the correction value of the field in which the operating point is located has exhibited at the beginning (or before).The deviation quantity adjusted in this way is then multiplied by a learning factor 360 so that the learning value 370 (for the respective operating point) is obtained. The learning factor thus ultimately indicates how strongly or how quickly the correction value is corrected later (as explained with reference to FIG. 2 ) with the learning value obtained thereby. The learning factor is determined, for example, depending on noise of the virtual sensor. The stronger the noise, the smaller the learning factor can be selected (e.g. between 0 and 1).

Claims

Method for adapting correction values for use in metering fuel, by means of at least one fuel injector (170), from a high-pressure accumulator (175) into a combustion chamber (165) of an internal combustion engine (160) using a learning data set (200) and a correction data set (260), wherein the learning data set (200) and the correction data set (260) each comprise a mutually corresponding plurality of fields (202, 262), wherein each of the plurality of fields is respectively associated with a pressure or a pressure range of fuel in the high-pressure accumulator and a quantity or a quantity range of fuel to be metered, wherein the learning data set (200) comprises one or more field ranges (206), wherein the one or each of the plurality of field ranges each comprises one or more, in particular contiguous, of the plurality of fields, and wherein each of the plurality of fields is associated with a field region, and wherein in the learning data set (200) a learning value (204) and a status (222) are associated with each of the plurality of fields, and wherein in the correction data set (260) a correction value (264) is associated with each of the plurality of fields, wherein the status of each of the plurality of fields of the learning data set is in each case a status from a status list (220), wherein the status list comprises at least the following states: active field, active adjacent field, inactive field, the method comprising: in the correction data set (260), adjusting the correction values (264) based on the learning values (204) of the learning data set (200), wherein the correction value in each of the active adjacent fields is in each case adjusted based on a mean adjacent field learning value, wherein the mean neighbor field learning value is or has been determined based on the learning value of the one or more learning values of the plurality of active neighbor fields, wherein the correction value in each field of each field region comprising an active field is or has been adjusted based on a mean field learning value, wherein the mean field learning value is or has been determined based on the learning values of the active fields of the field region, wherein the correction value in each inactive field is or has been adjusted based on a transmission learning value, wherein the transmission learning value is or has been determined according to at least one correlation rule based on the learning values of the active neighbor fields; and providing the correction map (260) with the adjusted correction values for further use.The method of claim 1, wherein if one or more predetermined activation criteria is present, the status of a field is changed to active field, and wherein for fields adjacent to the field area comprising an active field, the status of the active neighbor field applies, respectively.The method of claim 2, wherein the one or more predetermined activation criteria is at least one of: - a number of learning events is greater than a predetermined threshold, - a number of learning events for an individual field is greater than a predetermined threshold, - a time duration or travel distance, for an individual field, since a comparison time or travel distance is greater than a predetermined threshold.The method of any preceding claim, wherein if one or more predetermined deactivation criteria are present, the status of a field is changed to inactive field.Method according to one of the preceding claims, wherein the at least one correlation rule comprises one or more of the following rules: - a correlation rule applicable to new parts, - a correlation rule applicable to parts arriving at the end of a lifetime.Method according to one of the preceding claims, further comprising, for determining a respective learning value: determining an estimated actual quantity (302) based on a target quantity predetermined for the metering of fuel, in particular using a machine learning model, determining an adjusted actual quantity (320) based on the estimated actual quantity (302) and the respective correction value, determining a deviation quantity (330) based on the adjusted actual quantity (320) and the target quantity (322), adjusting the deviation quantity (330) based on at least one, preferably two, correction quantities (342, 352), and determining the learning value (370) based on the adjusted deviation quantity and a learning factor (360).The method of any preceding claim, wherein the fields that the one or more field areas comprise are adjusted as needed.Method according to one of the preceding claims, wherein each of the plurality of fields is respectively associated with a pressure of fuel in the high-pressure accumulator and a fuel to be metered, and wherein an adaptation of a setpoint quantity predefined for the metering of fuel based on the correction values comprises: determining an interpolated correction value based on at least two correction values adjacent to the setpoint quantity and the current pressure of the fuel in the high-pressure accumulator, wherein the interpolated correction value is used for the adaptation.Arithmetic unit (180) which is configured to carry out all method steps of a method according to one of the preceding claims.A computer program that causes a computing unit to perform all method steps of a method according to any one of claims 1 to 8 when executed on the computing unit.A machine readable storage medium having stored thereon a computer program according to claim 10.

Images