WO2017178015A1 - Method and device for determining a technical system model - Google Patents

Abstract

The aim of the invention is to determine a technical system model in a reduced amount of time and at the same time increasing the quality of the resulting parameters of the determined model. Said aim is achieved, according to the invention, in particular by means of a method for determining a technical system model, wherein output variables of the technical system are determined based on input variables/combinations of input variables. Said method comprises the following steps: a) specifying input variable limits, b) specifying output variable limits, c) determining input variables/combinations of input variables within the input variable limits and output variables associated therewith, by adjusting input variables/combinations of input variables and measuring output variables on the technical system, wherein at least one input variable/combination of input variables is determined which drives an output variable lying on an output variable limit, d) determining a first convex covering using the determined input variables/combinations of input variables, some or all input variables/combinations of input variables being the threshold points of the first convex cover, the hyperplanes being applied to the threshold limits, the hyperplanes containing the amount of the determined input variables/combinations of input variables, e) adjusting other input variables/combinations of input variables and measuring output variables on the technical system, the other input variables/combinations of input variables lie within the input variable threshold, and within the first convex covering and/or outside the first convex covering, excluding input variables/combinations of input variables which lie within at least one other convex covering, for each other convex covering, only one threshold point of the first convex covering and the points of intersection of the hyperplanes only applied to the threshold point comprise the input variable limits and optionally existing intersection points of the input variable limits as a threshold point of at least one other convex covering, wherein the only threshold point corresponding to an input variable/combinations of input variables which drives an output variable arranged on an output variable limit, f) training the parameters of the technical system model using the determined input variables/combinations of input variables and output variables.

Description

 Method and device for determining a model of a technical system

 description

The present invention relates to a method and a device for determining a model of a technical system with the features according to the claims.

 For example, according to Schreiber, A. (2010): "Dynamic engine measurement using various methods and models"; In: Electronic management of motor vehicle drives: electronics, modeling, control and diagnostics for internal combustion engines, transmissions and electric drives; 1. Aufl. Wiesbaden: R. Isermann , Pp 167-199 it is known that the most accurate knowledge of the manipulated variable ranges of an internal combustion engine, a necessary

Prerequisite for the measurement of the internal combustion engine is. In particular, it is known from this that not all possible combinations of manipulated variables are applicable. On the contrary, there are limits whose exceeding leads to damage or to undesirably high levels

Emissions leads. In any case, based on these limits results in a so-called variation space. To determine the variation space, several methods are described, for example a star-shaped departure in steps, until the respective limit is reached. Also shown is a method for determining the variation space, wherein the permissible

Manipulated variable combinations are displayed in the form of a convex hull. Based on this now known variation space can then be made an adaptation of the experimental design used.

Such a procedure is also from Renninger, P., K. v.. Pfeil, D. Hofmann, R. Isermann (2005): "Dynamic Models and their Application in the Optimization of N FZ Engines"; In: 14th Aachen Colloquium - Vehicle and Engine Technology, 1st Edition Aachen, p. 1205- 1222. In particular, it is described there that initially a basic measurement for

Determination of limit points is made. On the basis of these boundary points then the calculation of a convex hull, which describes the space of variation. However, with the

Basic survey only a part of the actual variation space was determined, starting from this convex hull, the measurement of the variation space continues.

Starting from the midpoint of each hyperplane of the convex hull, a further determination of boundary points takes place in the direction of the respective normal vector. Based on the so determined

Boundary points is again formed a convex hull. Starting from this convex hull, as described in a next step, a new determination of boundary points takes place, so that the convex hull, which in each step corresponds only to the respectively determined variation space, approaches the actual space of variation. The methods described rely on extensive upstream measurements of the

Variationspacees advance. As a result of this upstream and extensive process step of stepwise measurement of the variation space, valuable time is needed. A further disadvantage is also that the quality of the resulting parameters of the model to be determined is reduced, since the position of the experimental points is optimized for the experimental space finding.

 It is therefore an object of the present invention to make the determination of a model of a technical system less time-consuming and thereby to increase the quality of the resulting parameters of the model determined.

 This object is achieved in particular by means of a method for determining a model of a technical system, wherein by means of the model based on

Input variables / input value combinations output variables of the technical system can be determined, with the following steps:

a) specification of input quantity limits,

b) specification of output limits,

c) Determination of Input Quantities / Input Size Combinations within the

Input size limits and associated output variables, by setting

Input variables / input variable combinations and measurement of output variables on the technical system, wherein at least one input variable / input variable combination is determined which causes an output variable which lies on an output limit,

d) Determination of a first convex hull on the basis of the determined

Input variables / input size combinations, with some or all

 Input variables / input quantity combinations Limit points of the first convex hull are, whereby hyperplanes lie at the boundary points, wherein the hyperplanes include the set of the determined input variables / input value combinations,

e) setting of further input variables / input variable combinations and measurement of output variables on the technical system, the others

Input variables / input quantity combinations are within the input size limits, as well as within the first convex hull and / or outside the first convex hull, excluding input variables / input quantity combinations lying within at least one further convex hull, wherein each further convex hull is merely a boundary point of the first convex hull Shell and the intersections of the present at only one boundary point hyperplanes with the input size limits and optionally existing intersections of the input quantity limits are the boundary points of at least one other convex hull, where which corresponds to only one limit point of an input quantity / input quantity combination which causes an output quantity which lies on an output quantity limit,

f) training of the parameters of the model of the technical system based on the determined

Input variables / input variable combinations and output variables.

 According to the invention, the determination of a test space / a convex casing takes place on the basis of previously performed measurements in individual steps or at any time. In the process, this convex hull is widened by the described strategy in such a way that a new convex region arises in which the planning of a test point takes place. After measuring this

The test point is again the convex hull of all previously made measurements, this is extended with the described strategy, a new point planned and so on. Compared to the prior art, only a minimal initial design plan is measured, in any case eliminates the costly determination of system boundaries in an upstream process step.

 According to the invention, therefore, the convex hull is iteratively extended in an online process, so that points outside the convex hull are also planned. When an outside point has been measured, with the position of the point in turn depending on the behavior of the technical system under consideration, a new convex hull is calculated and expanded again. The end result of the expansion of the convex hull according to the invention is a non-measured region, but this indicates the maximum possible variation space at this time, provided that a convex variation space is assumed. Thus, there is a range that will be used for the further design of the experiment, which aims to provide the evaluable model area (convex hull after the entire

Surveying) and to optimize the model quality itself. It is advantageous that, by means of the method according to the invention, a refinement of the model as well as the boundary measurement are combined by expanding the respectively current convex hull and carrying out a planning within the extended area.

 According to the invention, therefore, the release of a further area, which is to be used for the experimental design. The combination of determining the maximum possible convex envelope size (extension strategy) and the experimental design within this space on the basis of other criteria (eg a distance-based criterion) results in a synergy effect according to the invention. Ie. the coupling between finding the boundaries of the variation space and selecting suitable experimental points for modeling is resolved.

In summary, there is an extension of the variation space for an experimental design or a maximization of the usable range for a later model evaluation. The result is an increase in acceptance of such methods, since a process step is eliminated. Further advantageous embodiments of the present invention will become apparent from the following embodiment and the dependent claims.

 As is well known, a technical system has inputs and outputs. The technical system can be any engine, for example one

Heat engine or an electric machine. In particular, the technical system may be an internal combustion engine. The internal combustion engine can according to the Ottoprinzip or the

Diesel principle are operated. For example, the internal combustion engine has a variable

Valve control on. In particular, the internal combustion engine has both an adjustable

Inlet camshaft and an adjustable exhaust camshaft on. Ie. a double variable camshaft spread, so the timing of the inlet and the outlet can be adjusted relative to each other. In any case, this results in two input variables, ie one

Input variable combination of the technical system, which is here an internal combustion engine, namely the (variable) timing of the intake valves and (variable) timing of the

Exhaust valves. Changes in these two input variables in turn cause changes in engine output, as is well known. Thus, in particular by a change (at least one) of the two input variables, the amount of residual gas or the

Residual gas in a combustion chamber of the engine influenced or changed or so results in a new input variable combination. Ie. the residual gas content may be a first output. The term "input variable combination" denotes a combination of two or more input variables

Formulation "input quantity combination" correspond.

As indicated in FIG. 1 by means of the arrow, the proportion of residual gas increases as a function of the separation of the intake and exhaust valves, ie the later the exhaust valves close and the earlier the intake valves open, the greater the overlap and thus the residual gas content. It is questionable now how much residual gas is expedient in an operating point of the internal combustion engine. For this purpose, for example, a further output variable can be considered, namely the smoothness of the internal combustion engine or the extent of the so-called cyclic fluctuations or the standard deviation of the indicated mean pressure of the internal combustion engine. The aim in the calibration or application of an internal combustion engine, it is known to find optimum settings, in this case, the overlap and thus adjust the residual gas content so that the residual gas content is as high as possible and consequently z. B. the fuel consumption is as low as possible (fuel consumption may be a further output variable) and / or the internal combustion engine as low as possible releases nitrogen oxides (the nitrogen oxide concentration in the exhaust gas may be yet another output variable) and / or the internal combustion engine is not unacceptably many hydrocarbons (the hydrocarbon concentration in the exhaust gas can be yet another output variable), However, the smoothness of the internal combustion engine may not be deteriorated so that a predetermined limit is violated. In any case, there is a course of a limit value GW relating to the smooth running of the internal combustion engine, which must not be violated, see FIG. 1. This profile of a limit value GW is initially unknown. The aim is to determine a model with which it is possible, based on input variables, here on the basis of the variable timing of the

Inlet valves and the variable timing of the exhaust valves or the overlap to describe the influence of these input variables on output variables - here the smoothness - and thus to determine or reveal the respective limit GW or the course of the limit GW. As the expert knows, these considerations also apply to others

Output variables, ie the fuel consumption or the nitrogen oxide or

 Hydrocarbon concentration in the exhaust gas, d. H. there is also a (initially unknown and to be modeled) functional relationship between input and output variables or

Input variables and the respective limit value GW or the course of the limit value GW.

 For a determination of a model that is as accurate as possible (that is, parameters that are as applicable as possible to this data model), it may be best to carry out experiments, ie. H. Settings of input variables and measurements of output variables on the technical system, over the entire relevant range of the input variables or the entire relevant input variable space (also

Variation space called) to distribute. Anyway, it's not beneficial to the location of the

Examination test points in the variation space (ie the combinations of set input variables / input variable combinations and measured output variables), as described in the prior art, only in terms of finding the

Input size space / input size range to optimize or align accordingly.

According to the bill is taken into account, as will become clear later.

 Initially, input size limits are specified, i. H. the maximum variation space of the input variables is staked out. As shown in FIG. 1, two input quantity limits initially result from the fact that the range of adjustability of an input variable - here the variable timing of the intake valves - is determined by a latest possible intake valve timing or latest possible opening of the intake valves (left vertical line in FIG 1) is limited and by an earliest possible timing of the intake valves or an earliest possible opening the

Inlet valves (right vertical line in Figure 1) is limited. Two input size limits further result from the fact that the range of adjustability of the second input variable - here the variable timing of the exhaust valves - by an earliest possible timing of the exhaust valves or the earliest possible closing of the exhaust valves (lower horizontal line in Figure 1) is limited and by a latest possible control time of the exhaust valves or a latest possible closing of the exhaust valves (upper horizontal line in Figure 1) is limited. Furthermore, output limits are specified, ie in particular minimum and / or maximum values which may not be exceeded or fallen short of as a result of a setting / variation of the input variables. These may be so-called hard or soft limits, see above in the cited prior art. Following the described embodiment, there is a

Output limit in particular in that due to the variation of the timing of the intake and exhaust valves of the limit GW regarding the smoothness must not be violated. Also conceivable here again (alternatively or additionally) is the specification of the limits GW (as output variable limits) with regard to the fuel consumption and / or the nitrogen oxide. Hydrocarbon concentration in the exhaust gas.

 According to the invention, in a further step, a determination of input variables takes place within the previously specified input quantity limits, that is to say a determination of the valve timing

Inlet valves and timing of the exhaust valves or combinations thereof and a determination of the associated output variables of the technical system, so here the internal combustion engine. Ie. there is a determination of associated input variables and output variables, wherein the output quantities result just because input variables are set on the technical system and the technical system by its properties or by running physical / chemical processes quasi out of the input variables produces specific output variables, which is generally is known. The setting of the

Input variables / input variable combinations are preferably carried out by means of a

Automation system.

 As shown in Figure 1, a total of four

 Input variables / input variable combinations determined within the specified

Input quantity limits, namely four combinations of timing of the intake valves and timing of the exhaust valves, highlighted in Figure 1 by circular rings. In other words, four different overlaps are given. As already described, the determination of these input variables within the input quantity limits as well as the output variables caused thereby or caused by the technical system takes place by setting these four combinations of intake valve timing and exhaust valve timing on the internal combustion engine (especially in conjunction with FIG a suitable test equipment / test bench and known measurement technology) and by measuring the output variables. As already described, for example, the smoothness, the fuel consumption or the concentration of nitrogen oxides / hydrocarbons in the exhaust gas can be such an output variable. In any case, at least one input quantity / input quantity combination is determined which causes an output quantity which lies on a (previously defined) output size limit. As described, only smoothness can be a possible starting point in this respect is looked at. As shown in Figure 1, even two input variables or two combinations of intake valve timing and timing of the exhaust valves are determined (G1, G3), each causing an output that the smoothness corresponds to the limit GW, ie the respective output is on an output quantity limit or on the course of the limit value GW shown in FIG

Internal combustion engine.

 According to the invention, the determination of input variables within the

Input size limits and the determination of the associated outputs, by setting inputs and measuring outputs on the technical system not only / not exclusively that the determination of at least one input that causes an output that is on an output limit, directly by setting

Input variables and measurement of output quantities takes place on the technical system, ie that the technical system or the internal combustion engine actually has to be brought to a limit range in order to determine the input variable which causes an output variable which lies on an output limit. Rather, this determination can also be made indirectly. This can be done according to the invention, that although an adjustment of input variables and measurement of output variables on the technical system / the internal combustion engine takes place, but not immediately at least one output variable is sought / measured on a

It is initially a training (the parameter) of the model of the technical system based on the (determined) input variables and (non-critical, since the internal combustion engine does not have to be brought to a limit range) output variables and then based on the model determination / determination at least an input that causes an output that is at an output limit (determined by the model).

 In the further course takes place on the basis of the determined as described above

 Input variables / input variable combinations the determination of a variation space and that the determination of a first convex hull. As is known, it is possible to include a convex set - here the set of detected inputs - by hyperplanes that lie on the edge of the convex set. Ie. some or all of the identified

 Input variables / input variable combinations can be boundary points of a convex hull. A convex hull is known to be the smallest convex polygon covering the set of all points (here, input sizes / input size combinations).

 Assuming only four, as shown in FIG

Input variables / input variable combinations determined, then all are determined Input variables / input quantity combinations Limit points of the first convex hull. To limit the convex amount of four shown in FIG

 Input variables / input quantity combinations are used hyperplanes, d. H. four hyperplanes lie on the four boundary points (G1 - G4) of the convex set. By way of example, of the four hyperplanes, first of all the two hyperplanes H1 and H2 are highlighted in FIG. The first hyperplane H1 abuts the boundary point G1 and the boundary point G2. The second hyperplane H2 is located at the boundary point G1 and at the boundary point G3. Anyway, that is through the

Collaboration of the four hyperplanes according to Figure 1 according to the invention a convex hull on the basis of the determined input variables determined / calculated, these input variables

Limit points of the first convex hull are, to which hyperplanes abut, so that the

Hyperplanes include the convex set of detected inputs.

 According to the invention, for the further determination of the variation space or for finding further input variables / input variable combinations for the following experimental design, an extension of the first convex hull takes place by input variables which lie outside the currently present first convex hull. Ie. it is explicitly determined outside the first convex hull of the set of input variables previously determined as described above, other input variables. For this purpose, a setting of other input variables and measurement of

Output variables in the technical system, wherein the other input variables are always within the input size limits, but both within the first convex hull and outside the first convex hull. As already described, the (current) first convex hull is formed by the hyperplanes adjacent to the four boundary points (G1-G4). As illustrated in FIG. 1 by means of the hatching, all the input variables now become one more

Experimental design / extraction of input-output variable combinations used for the creation of a model of the technical system, which lie in the hatched area, ie also outside the previously determined first convex hull.

 Ie. It is also possible to determine or determine input variables which are located in the region / space (the envelope, which is likewise convex here, it can also be a concave envelope), which lies between the boundary point G1, the boundary point G2 and the intersection S2 (FIG. which results from the input size limit and the second hyperplane H2 intersecting) and the

Input size limit (here the right vertical line, which represents the maximum advance of the intake camshaft) by the hyperplanes H1 and H2 and the other

Input size limit is included. Another boundary point of this likewise convex hull is the intersection point E1 of the input size limits with each other. Input variables can also be determined in the space / region or within the convex hull whose boundary points are the intersection E2 of the input quantity limits, the boundary point G2 and the boundary point G4.

 Input variables can also be determined in the space / region or within the convex hull whose / their boundary points the intersection E3 of the input size limits, the boundary point G4, the intersection S3 (resulting from the fact that the input size limit and the second hyperplane H2 cut) and are the boundary point G3.

 Input variables can also be determined in the space / region or within the convex hull whose boundary points the intersection point S4 (which results from the fact that the input size boundary and a hyperplane intersect), the intersection S1 (which results from that the input size limit and the first hyperplane H1 intersect), the

Limit point G1, the limit point G3 and the intersection E4 of the input quantity limits.

According to the invention, however, the two areas / spaces / convex hulls not hatched in FIG. 1 whose boundary points S1, S2 and G1 as well as S3, S4 and G3 are excluded from the further determination of input quantities and measurement of output quantities on the technical system, although these further Input variables are within the input quantity limits or outside the first convex hull (which is bounded by the boundary points G1-G4), as in the previously described cases.

 As shown in Figure 1 in particular by the course of the limit GW of quiet running is clear, the technical system, here the internal combustion engine, then up to and including a

Limit value GW of output variables operable when

 Input variables / input size combinations that are part / element of an amount that is enclosed by a convex hull. It is now concluded that the input variables which cause output variables beyond the limit value GW are only suitable for setting further input variables and measuring output variables on the technical system if the previously formed first convex hull (G1-G4) has a further (here) convex hull (G3, S4, E4, S1, G1) is extended, but according to the invention just the part sizes are excluded which cause outputs beyond the limit GW, which are part / element of a set, which in turn convex Sheath (S1, S2, G1 and S3, G3, S4) is included.

In practice, an exclusion of input quantities which lie within a further convex envelope (S1, S2, G1 and S3, G3, S4) takes place, see FIG. 1, ie there are two such further convex envelopes here. Of course, depending on the application, there may be more than two such further convex hulls. These further convex hulls have the following limit points. The boundary point G1 of the previously determined first convex hull is likewise a boundary point of a convex hull (in FIG. 1 on the right, not hatched), which comprises input variables to be excluded (in the further step of the method). In this way, the reference is made to the output output limit because the first boundary point G1 of the previously determined first convex hull of a

Input size / input size combination that causes an output that is on an output size limit. In any case, it is important that only one boundary point of the (previously determined) first convex hull is also a boundary point of one convex hull, which includes / excludes further input variables and not two or more boundary points of the (previously determined) first convex hull , Thus, the two boundary points G1 and G3 according to FIG. 1 are not part of the one convex hull, which includes input variables to be excluded in the further course, but exclusively only the boundary point G1. Another boundary point of the one convex hull (in FIG. 1 right, not hatched ), which comprises (in the further step of the method) input variables to be excluded, is the point of intersection S1 which is formed by the fact that the hyperplane H1 which adjoins this only one boundary point G1 and the right input quantity boundary intersect. Another boundary point of a convex hull (in FIG. 1 on the right, not hatched), which comprises input variables to be excluded (in the further step of the method), is the point of intersection S2, which is formed by the hyperplane applied to this only one boundary point G1 H2 and the right input limit. In any case, by the hyperplane H1 between the boundary point G1 and the intersection S1, the input size boundary between the intersection S1 and the intersection S2 and the hyperplane H2 between the intersection S2 and the boundary point G1 this is a convex hull (in Figure 1 right, not hatched ) educated.

In this way, the second convex hull (in FIG. 1 on the left, not hatched), which comprises input variables to be excluded (in the further step of the method), is also formed.

 It is also conceivable that these further convex hulls continue to have boundary points that are intersections of input size limits with each other. Although not shown in FIG. 1, it could be the case that the point of intersection S1 moves to the left, ie to the left of the point of intersection E4 of the input quantity limits, so that the above-described convex envelope having the input variables to be excluded not only three but also includes four boundary points, namely in addition to the intersection E4 of the input quantity boundaries with each other.

As shown, on the one hand, input variables and associated output variables were determined, wherein the input quantities are part / elements of the quantity which according to the invention has been expanded within the input size limits, that is to say with the exclusion of defined input variables a training of the parameters of the model of the technical system is carried out on the basis of these (determined) input variables and output variables, as is generally known.

 Preferably, the inventive method for determining a model of a technical system is performed stepwise, d. H. the first convex hull is iteratively extended.

 As described, according to the invention, a setting of more

 Input variables / input variable combinations and measurement of output variables on the technical system.

 In this case, at least one input variable / input variable combination is again determined which causes an output variable which lies on an output limit.

 Based on the first convex hull, the determination of an expanded convex hull is then performed on the basis of the now determined input variables / input quantity combinations, again some or all now determined input quantities / input quantity combinations being boundary points of the extended convex hull and in turn abutting the boundary points hyperplanes and the hyperplanes the set the now determined

Include input variables / input variable combinations.

 Anyway, a new setting of more

 Input variables / input variable combinations and measurement of output variables on the technical system, the other input variables / input variable combinations within the

Input size limits lie, as well as within the extended convex hull and / or outside the widened convex hull, to the exclusion of

 Input variables / input quantity combinations which lie within at least one further convex hull, wherein each further convex hull is merely a boundary point of the extended convex hull and the intersections of the hyperplanes adjoining this only one boundary point with the input size limits and optionally existing intersection points of the

Input size limits are boundary points of the at least one further convex hull, wherein the only one boundary point of the extended convex hull corresponds to an input quantity / input quantity combination determined in the previous step, which causes an output quantity which lies on an output size limit.

 This procedure is repeated, with the first convex hull coinciding with each

Repetition extended, so that training the parameters of the model of the technical system based on the number of detected with each repetition

Input variables / input variable combinations and output variables takes place, whereby the process is repeated until a specific criterion is met. The criterion is, for example, that a given accuracy of the model of the technical system is achieved. According to the invention, a device for determining a model of a technical system is furthermore provided. This is characterized in that a computer designed for carrying out the method according to the invention has a CPU and a

machine-readable storage medium is provided. In particular, the computer with a CPU is a computer that is part of a well known one

Test bench automation is or is associated with it. This test stand automation or the automation system is in turn connected to the considered technical system. The test bed automation is also designed so that not only the setting of

Input variables / input variable combinations with it takes place, but also measurements of output variables at the technical system take place, whereby the obtained measured values are provided for a further processing. The test bench automation can thus be a

Be computer program. The further processing of the measured values obtained, d. H. In particular, the determination of convex hulls provided according to the method according to the invention and the training of the parameters of the model of the technical system are carried out by means of a

Computer program to solve mathematical problems or to graphically display the results. On the storage medium of the computer is thus stored at least one computer program comprising steps of the inventive method, wherein the

Computer program or for carrying out the method according to the invention

interacting computer programs are executed by the CPU.

Claims

claims 1. A method for determining a model of a technical system, wherein by means of the model on the basis of input variables / input variable combinations output variables of the technical system are determined, comprising the following steps: a) specification of input quantity limits, b) specification of output limits, c) Determination of Input Quantities / Input Size Combinations within the Input size limits and associated output variables, by setting Input variables / input variable combinations and measurement of output variables on the technical system, wherein at least one input variable / input variable combination is determined which causes an output variable which lies on an output limit, d) Determination of a first convex hull on the basis of the determined Input variables / input size combinations, with some or all  Input variables / input quantity combinations Limit points of the first convex hull are, whereby hyperplanes lie at the boundary points, wherein the hyperplanes include the set of the determined input variables / input value combinations, e) setting of further input variables / input variable combinations and measurement of output variables on the technical system, the others  Input variables / input quantity combinations are within the input size limits, as well as within the first convex hull and / or outside the first convex hull, excluding input variables / input quantity combinations lying within at least one further convex hull, wherein each further convex hull is merely a boundary point of the first convex hull Shell and the intersections of the present at this only one boundary point hyperplanes with the input size limits and optionally existing intersections of the input quantity limits are boundary points of at least one other convex hull, which is only a boundary point of the first convex hull a  Input size / input size combination that causes an output that is on an output size limit, f) training of the parameters of the model of the technical system based on the determined Input variables / input variable combinations and output variables. 2. The method according to claim 1, wherein if in step e) an adjustment of further input variables / input variable combinations and measurement of output variables on the technical system has been carried out and again at least one input variable / input variable combination has been determined, which causes an output that is on an output limit, based on the first convex hull, an expanded convex hull is determined on the basis of the now determined input variables / input quantity combinations, again some or all now determined input variables / input quantity combinations being boundary points of the extended convex hull, again at the boundary points hyperplanes are present, the hyperplanes being the set include the now determined input variables / input variable combinations, wherein step e) is repeated, so that a new setting of further input variables / input variable combinations and M Output quantities are measured on the technical system with the other input variables / input value combinations within the input size limits, as well as within the extended convex envelope and / or outside the extended convex envelope, excluding  Input variables / input quantity combinations which lie within at least one further convex hull, wherein each further convex hull is merely a boundary point of the enlarged convex hull and the intersections of the hyperplanes adjoining this only one boundary point with the input size limits and any intersection points of the Input size limits are boundary points of the at least one further convex hull, which is only a boundary point of the extended convex hull of a  Input size / input size combination that causes an output that is on an output size limit.  3. The method of claim 2, wherein the method is repeated according to claim 2, wherein the first convex hull expands with each repetition, so that according to step f) a training of the parameters of the model of the technical system on the basis of each Repetitive increasing number of determined input variables / input variable combinations and output variables is carried out, the process is repeated until a certain criterion is met.  4. The method according to claim 3, wherein the criterion is that a predetermined accuracy of the model of the technical system is achieved.  5. The method according to claim 1 to 4, wherein the determination of  Input variables / input variable combinations within the input size limits and the determination of the associated output variables is carried out by a setting of Input variables / input variable combinations and measurement of output variables on the technical side System is performed, but not directly at least one output variable is measured, which is on an output size limit, but first a training of the model of the technical system based on the set input variables / input quantity combinations and measured outputs and then based on the model, the determination of at least one Input variable / input quantity combination that causes an output that is on an output limit.  6. The method according to claim 1 to 5, wherein the technical system is an engine. 7. The method of claim 6, wherein the engine is an internal combustion engine.  8. The method according to claim 1 to 7, wherein the setting of  Input variables / input variable combinations by means of an automation system takes place. 9. The method according to claim 8, wherein in the setting of  Input variables / input variable combinations continuous monitoring takes place, if an output limit is reached.  10. A device for determining a model of a technical system, characterized in that a prepared for carrying out the method according to one of claims 1 to 9 computer provided with a CPU and a machine-readable storage medium, wherein on the storage medium at least one computer program is stored, the Steps of the Method according to one of claims 1 to 9, wherein the computer program is executed by means of the CPU.