Systematic tolerance determination relating to samples of series products

Abstract

To determine a tolerance region within an operational design domain (ODD) around a target parameter sample of a simulation model of a product relating to a steer-by-wire steering system or a steering system for highly automated driving.SOLUTION: A method comprises a step of computing a dissimilarity metric for each of multiple pairs, each pair comprising a parameterized simulation model assigned to a target parameter sample and respective one parameterized simulation model of multiple parameterized simulation models, one distance arising for each pair and thereby multiple distances arising; optionally, the dissimilarity metric being based on a gap metric, a ν-gap metric, or an L2 metric; and a step of determining, based on the multiple distances and a maximum permissible distance, the tolerance region within the ODD around the target parameter sample.SELECTED DRAWING: Figure 1

Description

[Background technology]

[0001] Series-produced products (especially large-scale series-produced products), such as steering systems, are subject to variations in product parameters, such as friction, elasticity, and / or inertia, due to manufacturing tolerances and errors. Series-produced products are also subject to additional parameter variations, due to wear and / or aging caused by environmental influences. All value ranges and value combinations of the parameter variations that occur in practice for a series-produced product are the so-called "operational design domain" (ODD) of this series-produced product, e.g., "Betriebsbereich fuer die Auslegung" in German.

[0002] In any simulation model of a (serial) manufactured product, differences between the modeled and real behavior arise based on parameter variations within the entire ODD and due to simplifications made during modeling. Such a simulation model, including model variations and / or model uncertainties within the entire ODD, is the basis for a simulation-based product release. The characterization of the model uncertainties required in this regard is typically performed today on the basis of a few selected product prototypes, since a full characterization is generally too time-consuming. In the following, the target version (also the ideal version) of a prototype is referred to as a product specimen, and its physical realization as a product individual. The selection of a product specimen for characterizing the uncertainty is usually based on expert opinion.

[0003] However, exact realization of the selected (ideal) product sample is practically impossible due to the final manufacturing precision, and as a result, the product sample and the product individual attributed to it have system behaviors that deviate slightly from each other. Typically, acceptable parameter tolerances for the realization of the product sample are defined by expert knowledge. Instead, the aim may be to manufacture and / or finish the product individual as accurately as possible. However, this does not preclude specifying and / or establishing parameter tolerances that are too loose, which may result in significant system behavior differences between the product sample and the product individual. On the other hand, it does not preclude specifying and / or establishing parameter tolerances that are too tight, which may result in unnecessarily high manufacturing costs, even though the system behavior of the product sample and the product individual may be substantially identical.

[0004] The problem to be solved that underlies this disclosure can be found, for example, in providing a method for determining how much a product individual can differ from a predetermined product example in order to fully realize this product example in terms of its system behavior.

[0005] Compared to conventional steering, steer-by-wire (SbW) steering systems and / or steering systems for highly automated driving (HAD) are subject to stricter standard requirements for product release. To ensure that the relatively strict release requirements for (large-scale) serially produced SbW and HAD steering systems do not significantly increase the costs of real-world testing and inspection compared to conventional steering, the field is focusing on simulation-based release processes. For such simulation-based release, a validated and verified simulation model of the steering system with known model uncertainties is essential.

[0006] Within the framework of an in-house and simulation-based release process for SbW and / or HAD steering systems, a characterization of model uncertainty should be performed based on a few selected product individuals.

[0007] The problem before us is therefore to systematically select, using model-based criteria, an attributable product exemplar that is therefore representative of the entire ODD of a (large-scale) serially produced steering system with quantifiable residual uncertainties.

[0008] Currently, there is no systematic method for determining meaningful parameter tolerances for the realization of a product sample, thereby ensuring negligible quantitative system behavior dissimilarity between the product sample and the product individual, thereby avoiding overly tight or loose tolerances.

[0009] In systems theory, various metrics are known to quantify the dissimilarity of two systems (hereafter, systems and products can be considered equivalent) and thereby compare these systems. Below, we will explain the gap metric, the ν-gap metric, and the L2 metric.

[0010] The gap metric quantifies the dissimilarity of the uncontrolled (open-loop) input / output behavior of two systems P1 and P2 relative to their stability and performance characteristics under controlled operation (closed-loop) on a scale within the real interval [0,1]. A metric result close to 0 means that both systems are very similar, and any P1-stabilizing controller will also stabilize system P2 with similar controlled performance. A metric result of 0 means that the considered systems P1 and P2 behave exactly identically. On the other hand, a metric result close to or equal to 1 means that systems P1 and P2 are very dissimilar. Furthermore, the gap metric allows for statements about the robust stability of closed control systems with model uncertainty. An explicit controller concept is required for gap metric evaluation. Further details about the definition and properties of the gap metric can be found in Chapter 17 of the book "Essentials of Robust Control," by Kemin Zhou and John C. Doyle, 1st Edition, Pearson, 1997, ISBN: 9780135258332.

[0011] The systems-theoretic propositions and implications of the ν-gap metric are very similar to the gap metric, although both metrics are defined fundamentally differently. Evaluation of the ν-gap metric requires examining the rotational speeds of the systems P1 and P2 being compared, rather than the controller concepts. Further details about the definition and properties of the ν-gap metric can be found in Chapter 17 of the book "Essentials of Robust Control," by Kemin Zhou and John C. Doyle, 1st Edition, Pearson, 1997, ISBN: 9780135258332, or in the publication "Frequency domain uncertainty and the graph topology," by Glenn Vinnicombe, IEEE Transactions on Automatic Control, vol. 38, no. 9, pp. 1371-1383, September 1993, DOI: 10.1109 / 9.237648.

[0012] The definition of the L2 metric is equivalent to the ν-gap metric, which does not examine the number of revolutions. For this reason, the L2 metric has a lower theoretical importance, although the principle propositions and implications of both metrics are similar. More details about the definition and properties of the L2 metric can be found in Chapter 17 of the book "Essentials of Robust Control", by Kemin Zhou and John C. Doyle, 1st Edition, Pearson, 1997, ISBN: 9780135258332.

[0013] All three introduced dissimilarity metrics have the following well-known properties: - These metrics quantify the dissimilarity of two systems based on their uncontrolled input / output behavior in relation to their stability and performance characteristics under controlled operation. The metric results allow determining whether two systems P1 and P2 are sufficiently similar, i.e. whether P1 can be considered representative of P2. - The metric results can also be explained as the (systematic) distance between the systems being compared. - For all systems to be compared, the following holds: L2 result ≦ ν-gap result ≦ gap result. [Prior art documents] [Non-patent literature]

[0014] [Non-Patent Document 1] Chapter 17 of the book "Essentials of Robust Control," by Kemin Zhou and John C. Doyle, 1st Edition, Pearson, 1997, ISBN: 9780135258332 [Non-patent document 2] Publication "Frequency domain uncertainty and the graph topology", Glenn Vinnicombe, IEEE Transactions on Automatic Control, vol. 38, no. 9, pp. 1371-1383, September 1993, DOI: 10.1109 / 9.237648 Summary of the Invention

[0015] A first general aspect of the present disclosure relates to a computer-implemented method for determining tolerance regions within an operational design domain (ODD) around target parameter samples of a parameterizable simulation model for a product, which may be, for example, a steer-by-wire steering system and / or a steering system for highly automated driving.

[0016] The method includes calculating a dissimilarity metric for each of the multiple pairs, where each pair includes a parameterized simulation model assigned to the target parameter sample and a respective one of the multiple parameterized simulation models, resulting in one distance for each pair, thereby resulting in multiple distances. The dissimilarity metric may be based on, for example, a gap metric, a v-gap metric, and / or an L2 metric.

[0017] The method further includes determining a tolerance region in the ODD around the target parameter sample based on the multiple distances and the maximum allowable distance. A second general aspect of the present disclosure relates to a computer system designed to execute a computer-implemented method for determining tolerance regions within an operational design domain (ODD) around target parameter samples of a parameterizable simulation model for a product according to the first general aspect (or an embodiment thereof).

[0018] A third general aspect of the present disclosure relates to a computer program designed to perform a computer-implemented method for determining tolerance regions within an operational design domain (ODD) around target parameter samples of a parameterizable simulation model for a product according to the first general aspect (or an embodiment thereof).

[0019] A fourth general aspect of the present disclosure relates to a computer-readable medium or signal storing and / or embodying a computer program according to the third general aspect (or an embodiment thereof).

[0020] The method proposed herein according to the first general aspect (or an embodiment thereof) allows for systematically determining tolerance zones based on model-based criteria for each realization (i.e., physical production) of a product sample, so that the system behavior of the product sample and the product individual match, up to an acceptable quantitative dissimilarity. In particular, parameter tolerances for the product sample can be determined. The product sample to be manufactured is assigned to a target parameter sample and is therefore a target version of the prototype (also an ideal version). Thus, in the realization of the product sample, a product individual can be created as a prototype whose parameter sample lies within the tolerance zone around the target parameter sample.

[0021] Among other advantages over the state of the art, the following may be achieved by the method proposed herein according to the first general aspect (or an embodiment thereof): - A systematic procedure for establishing individual parameter tolerances (i.e., the range of permissible values ​​of one or more parameters) for the realization of a product sample; - the calculable individual tolerance zones of the product parameters (i.e. the allowable zones for all parameters) for every product sample, - Quantifiable system behavior dissimilarity between product samples and product individuals based on mismatched parameters within the entire ODD; - possibility of identifying local product parameters or combinations thereof in the vicinity of a product sample which have a particularly strong influence on the system behavior of this product sample.

[0022] The method proposed herein according to the first general aspect (or an embodiment thereof) may be used in the design stage and / or system development (i.e. after the design stage) when developing a real (i.e. physical) product (e.g. in the case of a steer-by-wire (SbW) steering system).

[0023] For example, in the design phase or system development, the present invention can be used to determine individual tolerance regions (e.g., to characterize model uncertainty) for (product) parameters for all product sample realizations, while respecting acceptable system behavior dissimilarities between the product samples and the product individuals. In particular, individual parameter tolerances for all product samples to be realized can be calculated, within which acceptable system behavior dissimilarities are respectively respected. This is important because prototypes whose parameters exactly correspond to target parameter samples are difficult to manufacture and / or require considerable effort for subsequent adjustment (e.g., shimming).

[0024] Moreover, for example, during the design phase, intermediate results of the method based on the first general aspect (or an embodiment thereof) and additional mathematical considerations can identify local product parameters and / or combinations thereof near a product sample that have a particularly strong influence on the system behavior of this product sample.

[0025] The target parameter sample may be one of a number of representative parameter samples in the ODD, i.e., one of a number of parameter samples representative of the entire ODD of a (series) manufactured product with quantifiable residual uncertainty. The realized product specimen (more precisely, the product individual) is in this case the representative product individual. On the other hand, the target parameter sample may not be a representative parameter sample, i.e., may be predetermined as a target for realization, arbitrarily or for other reasons (e.g., as a limit specimen).

[0026] Based on these representative product individuals, further system development can systematically characterize the uncertainty between the real product behavior and its modeled behavior within the entire ODD, due to parameter variations and modeling simplifications. The representative product individuals with characterized model uncertainty can then be used for product control development and / or product release.

[0027] The method proposed here according to the first general aspect (or an embodiment thereof) can be applied, inter alia, within the framework of a process for the release of SbW and HAD steering systems, in which the realization of a representative product specimen of the entire ODD can be utilized in the validation and / or verification of the steering system.

[0028] The method proposed herein according to the first general aspect (or an embodiment thereof) may be implemented wholly or partly numerically, which is advantageous as it does not depend on the existence of a parameterizable simulation model in analytical form. [Brief explanation of the drawings]

[0029] [Figure 1] 1 is a schematic diagram of an exemplary embodiment of a computer-implemented method for determining tolerance regions within an operational design domain (ODD) around target parameter samples of a parameterizable simulation model for a product, particularly a steer-by-wire steering system and / or a steering system for highly automated driving. [Figure 2a] FIG. 2a is a schematic diagram of one embodiment for determining a tolerance region within an ODD based on a selected parameterized simulation model whose respective distances to a parameterized simulation model assigned to a target parameter sample sufficiently correspond to the maximum allowable distance. [Figure 2b] FIG. 2b is a schematic diagram of one embodiment for determining tolerance regions within an ODD based on interpolation of distances for a parameterized simulation model assigned to a target parameter sample, resulting in a distance mapping within the ODD. [Figure 3] FIG. 1 illustrates an exemplary embodiment of the method. [Figure 4a] FIG. 1 illustrates an exemplary ODD with multiple parameterized simulation models, target parameter samples surrounded by tolerance regions. [Figure 4b] FIG. 4b illustrates the example ODD of FIG. 4a without multiple parameterized simulation models. DETAILED DESCRIPTION OF THE INVENTION

[0030] The method 100 proposed in this disclosure focuses on determining a tolerance region 30 within an operational design domain (ODD) around a target parameter sample 10 of a parameterizable simulation model for a product, particularly a steer-by-wire steering system and / or a steering system for highly automated driving. Alternatively or additionally, the method 100 may also focus on manufacturing a product specimen in the form of a product solid that realizes the product specimen.

[0031] The method 100 is wholly or partially numerical and therefore may be applied even when a parameterizable simulation model does not exist in analytical form. First, a computer-implemented method 100 is disclosed for determining a tolerance region 30 within an operational design domain (ODD) around a target parameter sample 10 of a parameterizable simulation model for a product. The product may be, inter alia, a serially manufactured product, i.e., produced continuously. This method 100 (although it may also be applied to non-serial or small-batch production) is particularly useful when a large number of similar products are to be manufactured, although the large number of similar products may differ (e.g., due to production and / or materials). The large number of similar products may include, for example, more than 1e5 products per year, more than 5e5 products per year, or more than 1e6 products per year.

[0032] The product may for example be a steer-by-wire steering system. Alternatively or additionally, the product may be a steering system for automated driving, in particular highly automated driving.

[0033] A parameterizable simulation model can be analytical, but does not have to be analytical. For example, as shown generally in Figure 1, method 100 includes calculating 130 a dissimilarity metric for each of a number of pairs, where each pair includes a parameterized simulation model (always the same for the number of pairs, labeled as Example PUM in Figure 3) assigned to the target parameter sample 10 and a respective one of the number of parameterized simulation models, resulting in one distance for each pair, thereby resulting in a number of distances. Step 130 is labeled "Calculate Dissimilarities Between All PUMs and Example PUMs" in the exemplary embodiment in Figure 3.

[0034] The dissimilarity metric may be based on a gap metric, a v-gap metric, and / or an L2 metric. In particular, the dissimilarity metric may be a gap metric, a v-gap metric, or an L2 metric. Alternatively, the dissimilarity metric may be based on a combination of the gap metric, the v-gap metric, and / or the L2 metric. The dissimilarity metric may output a quantitative measure of the dissimilarity of a pair of parameterized simulation models, i.e., a quantitative measure of how similar or dissimilar both parameterized simulation models of the pair are. In this respect, the dissimilarity metric may also be referred to as a similarity metric or a comparison metric. The quantitative measure output by the dissimilarity metric may be called a distance. If both parameterized simulation models of a pair are similar, the distance may be small. In particular, if both parameterized simulation models of a pair are identical (i.e., maximally similar), the distance may be zero. In contrast, if both parameterized simulation models of a pair are dissimilar, the distance may be large.

[0035] For example, as shown schematically in Figure 1, the method 100 further includes determining 140 a tolerance region 30 in the ODD around the target parameter sample 10 based on the multiple distances and the maximum allowable distance. Step 140 is labeled "Calculate Tolerance Region" in the exemplary embodiment in Figure 3.

[0036] Exemplary results of the determination of the tolerance region 30 around the target parameter sample 10 are shown in FIGS. 4a-b. The target parameter sample 10 and tolerance region 30 may define a product specimen to be manufactured for this product, hereinafter also referred to as the target product specimen.

[0037] 1, method 100 may further include step 150 of outputting requirements for the production of a to-be-manufactured product specimen based on target parameter sample 10 and tolerance zone 30, the requirements being met if (e.g., all) parameters of the to-be-manufactured product specimen are within tolerance zone 30. Step 150 is labeled "Convert tolerance zones to individual parameter tolerances" in the exemplary embodiment in FIG.

[0038] Based on this requirement, production (e.g., prototyping) can produce a product individual. This requirement may include or be an algorithm that can check whether one or more parameters of a product specimen (i.e., product individual) to be produced are within a tolerance region 30 around the target parameter sample 10. Such an algorithm can be useful, for example, when the tolerance region is higher dimensional and deviates significantly from a product space (i.e., a Cartesian product of intervals).

[0039] The determination 140 of the tolerance region 30 within the ODD may be performed, for example, such that within the tolerance region 30, each distance to the parameterized simulation model assigned to the target parameter sample 10 is less than the maximum allowable distance. Alternatively, the determination 140 of the tolerance region 30 within the ODD may be performed, for example, such that within the tolerance region 30, each distance to the parameterized simulation model assigned to the target parameter sample 10 is less than or equal to the maximum allowable distance. For example, it is also conceivable that the determination 140 of the tolerance region 30 within the ODD may be performed such that within the tolerance region 30, each distance to the parameterized simulation model assigned to the target parameter sample 10 is, at most, approximately the maximum allowable distance.

[0040] Furthermore, the determination 140 of the tolerance region 30 within the ODD may be performed such that the tolerance region 30 is exactly or approximately maximal. The tolerance region 30 may be maximized, for example, based on the Lebesgue measure.

[0041] Furthermore, the determination 140 of the tolerance region 30 within the ODD may be performed such that the tolerance region 30 is a connected set within the ODD. Alternatively or additionally, the determination 140 may be performed such that the tolerance region 30 is a convex set within the ODD. In particular, the tolerance region may be a convex connected set within the ODD.

[0042] The tolerance region 30 can be, for example, a manifold, a polytope, in particular a convex polytope, a hyperellipsoid, or a hypercube. A hypercube is particularly advantageous since the parameters and their tolerances are independent of each other. If this is not given, the tolerance region 30 can be defined as an (arbitrary) manifold. To reduce the memory requirements of such a manifold, it may be useful to approximate the manifold by a polytope or a hyperellipsoid.

[0043] As shown, for example, schematically and optionally, in FIG. 2a, determining 140 the tolerance region 30 within the ODD may first include selecting 141 one or more parameterized simulation models whose distance to the parameterized simulation model assigned to the target parameter sample 10 sufficiently corresponds to the maximum allowable distance, and then determining 142 the tolerance region 30 based on the one or more selected 141 parameterized simulation models, optionally based on a polytope, in particular a convex polytope, a hyperellipsoid, or a hypercube. The distance may, for example, sufficiently correspond to the maximum allowable distance if each meets a predetermined criterion. The distance may, for example, sufficiently correspond to the maximum allowable distance if the deviation from the maximum allowable distance is at most 5%. In this case, the tolerance region 30 may, for example, be determined 142 to include the parameter samples assigned to the selected 141 parameterized simulation models. For example, these parameter samples may be corners of the polytope.

[0044] Alternatively, as shown, for example, schematically and optionally in FIG. 2b , determining 140 the tolerance region 30 in the ODD may include first interpolating 143 distances to a parameterized simulation model assigned to the target parameter sample 10, resulting in a distance map in the ODD, and then determining 144 the tolerance region 30 based on the distance map in the ODD, optionally based on a polytope, particularly a convex polytope, hyperellipsoid, or hypercube. The distance interpolation 143 here may be based on a fit function, whose fit parameters are determined, for example, by a Gaussian process. This may be done inductively, but need not be. A Gaussian process (also known as a Gaussian process or GP) is a stochastic process, which may be interpreted, among other things, as a probability distribution in a function space and may thereby be used to solve regression problems. Thus, the mapping of ODD parameters to distances (related to a dissimilarity metric) to a parameterized simulation model of the target product sample under consideration may be expressed using a GP. For this purpose, a parameterizable GP is adapted based on the selection of parameter samples (e.g., "near" the target parameter sample, or within a representative region of the target parameter sample, or within the entire ODD) and the distances attributed to them. Subsequently, for each point in the ODD, the parameterized GP can be used to calculate a normal distribution of distances to the PUM of the target product sample under consideration (see metric mountains with uncertainty in the ODD near the target product sample). Based on the mean or desired percentile (e.g., 95% percentile) of the probability distribution and the acceptable distance, a tolerance region can finally be calculated (see the second-closest contour 40 around the target parameter sample 10 in Figures 4a-b). In this regard, the tolerance region can be approximated by, for example, a polytope or ellipsoid, and its Lebesgue measure can be maximized. The tolerance region can be, but need not be, a region in the mathematical sense.

[0045] As shown schematically and optionally in FIG. 1 , the method 100 may further include initially determining 110 a number of parameter samples 20 in an ODD for the product. Each parameter sample in the ODD may include one or more parameters of the parameterizable simulation model. The determination 110 of the number of parameter samples 20 in the ODD may be performed to ensure sufficiently uniform coverage of the ODD. Such sufficiently uniform coverage may be based, for example, on pseudorandom numbers. Pseudorandom numbers (e.g., via Mersenne Twister) are typically uniformly distributed. Alternatively or additionally, such a sufficiently uniform coverage may be based on Latin Hypercube sampling. Alternatively or additionally, such a sufficiently uniform coverage may be based on Sobol sequences. In particular, such a sufficiently uniform coverage may be based on a combination of pseudorandom numbers, Latin Hypercube sampling, and / or Sobol sequences. Already, assuming sufficiently uniform coverage, a representative parameterized simulation model of the parameterizable simulation model may be better and more efficiently selected 140. Thus, for example, determining 110 the number of parameter samples 20 in the ODD may be based on pseudorandom numbers, Latin Hypercube sampling, and / or Sobol sequences. Alternatively, determining 110 the number of parameter samples 20 in the ODD may relate to only a portion of the ODD (e.g., only around the target parameter sample 10). Step 110 is labeled "Sampling ODD" in the exemplary embodiment in FIG.

[0046] The method 100 may then include generating 120 a parameterized simulation model based on the parameterizable simulation model and a number of parameter samples 20 in the ODD, e.g., as illustrated generally and optionally in Fig. 1. In this regard, for example, the parameterizable simulation model may be evaluated at each of the parameter samples 20. Alternatively or additionally, here, a surrogate model for the parameterizable simulation model may be created for each of the parameter samples 20.

[0047] A parameterizable simulation model can be, but need not be, analytical. If the parameterizable simulation model is not analytical (e.g., in the case of a black-box simulation model), it may not be easily evaluated at each parameter sample. In such a case, for example, a surrogate model for the parameterizable simulation model for each parameter sample may be created using numerical simulation. Therefore, such a surrogate model may also be considered a parameterized simulation model of the parameterizable simulation model. Thus, in this regard, the creation of a surrogate model for each parameter sample may be a parameterization of the parameterizable simulation model. In this regard, a (parameterizable) simulation model may be parameterizable even if it does not exist in analytical form.

[0048] Step 120 is labeled "Determine PUM for all ODD samples" in the exemplary embodiment of Figure 3, where PAM refers to a parameterizable simulation model and PUM refers to multiple parameterized simulation models. "ODD sample" here refers to the parameters of each parameterized simulation model among multiple parameterized simulation models.

[0049] The target parameter sample 10 may be one from a number of target parameter samples representing an ODD. The multiple parameter samples 20 may be within a representative range of the target parameter samples 10. Here, among other things, determining 110 the multiple parameter samples 20 within the representative range may be performed such that the representative range is sufficiently uniformly covered. Again, determining 110 the multiple parameter samples 20 may be based on pseudorandom numbers, Latin hypercube sampling, and / or Sobol sequences.

[0050] As shown schematically and optionally in FIG. 1 , for example, the method 100 may further include identifying 160 one or more parameters within the tolerance zone 30 that have a relatively large effect on the product. This may be done, for example, by sensitivity analysis. This may improve understanding of the product and its behavior. This may be beneficial, especially in the case of higher-dimensional ODDs, where the tolerance zone 30 is no longer directly observable. This may improve system understanding within the tolerance zone. The results may be taken into account in the realization of the target product sample.

[0051] The method 100 may include a step of determining 140 respective tolerance regions in the ODD around the multiple target parameter samples 10. In this case, therefore, around each target parameter sample, an attributed tolerance region in the ODD may be determined. The method 100 may include a step of defining a maximum allowable distance, which is labeled "Define Maximum Allowable Distance" in the exemplary embodiment in FIG. 3. In the case of multiple target parameter samples, this definition may be made globally, i.e., for all target parameters, or may be defined individually for target parameter samples of the multiple target parameter samples.

[0052] As already discussed, the tolerance zone 30 can be used to determine one or more product individuals whose parameters lie within the tolerance zone around the target parameter sample. Alternatively or in addition to the requirement output 150 possibility discussed above, the method 100 can include a step of matching parameter samples of real product individuals stored in a database with the tolerance zone, particularly based on similarity relationships. In the event of an unsatisfactory degree of similarity, one or more instructions can be output on how a sufficiently satisfactory product sample can be manufactured from the real product sample (e.g., by adjustment, shimming, etc.). This product sample can be, but need not be, a representative product sample.

[0053] The product individuals that realize the product samples to be manufactured can be used to design the control of the product. The design of the control of the product can be based, among other things, on a number of product individuals, each realizing a product sample to be manufactured. In particular, if the product samples to be manufactured represent ODDs, the conception and / or design of the robust control of the product can be improved. In particular, this can ensure that the control of the product is designed to be particularly robust. In particular, this can increase the safety of the product.

[0054] Furthermore, one or more requirements for a product can be checked based on individual products that realize a product sample to be manufactured. The check of one or more requirements for a product can be based, among other things, on multiple individual products that each realize a product sample to be manufactured. This ensures that the product is designed robustly. This can also, among other things, increase the safety of the product.

[0055] 3 illustrates one exemplary embodiment of the method 100. In this regard, for example, the following steps may be performed in sequence: - "Sampling ODD", 110 - "Determining PUM for all ODD samples", 120 - "Calculating the Dissimilarity Between All PUMs and Sample PUMs", 130 - "Maximum allowable distance specification" - "Calculating Tolerance Zones", 140 - "Convert Tolerance Zones to Individual Parameter Tolerances", 150 4a illustrates an exemplary ODD having a number of parameterized simulation models (whose parameter samples 20 are shown as dots) and target parameter samples 10 (shown as crosses) surrounded by a tolerance region 30. The ODD is a two-dimensional product space in this example, i.e., represented by two parameters (one parameter in the x-direction and the other parameter in the y-direction), each of which can take on values ​​within an interval. In general, the ODD need not be a product space, but could be an (arbitrary) manifold. The dimensionality of the ODD can be arbitrary. Here, the tolerance region 30 is an exemplary rectangle.

[0056] For this example, in 3D, i.e., in the z direction perpendicular to the plane of the paper, the distance (based on the dissimilarity metric) of each parameter sample 20 in the ODD (or, more precisely, of that parameterized simulation model) to the parameterized simulation model assigned to the target parameter sample 10 could be shown. In 2D, instead, four contour lines (e.g., 40) of this magnitude in the z direction are shown. Each of the four dashed lines is a contour line, i.e., has the same z value. In addition to these contour lines, the boundaries (dashed lines) of the range of influence of each representative parameterized simulation model are drawn.

[0057] FIG. 4b illustrates the exemplary ODD of FIG. 4a without the multiple parameterized simulation models (or more precisely, the parameter samples 20 thereof). In FIGS. 4a-b, for example, to the left of the boundary line (dashed line) of the range of influence, there may be a further target parameter sample, around which a tolerance region may also be determined according to the method 100.

[0058] Additionally, a computer system designed to execute a computer-implemented method 100 for determining a tolerance region 30 within an operational design domain (ODD) around a target parameter sample 10 of a parameterizable simulation model for a product is disclosed. The computer system may include a processor and / or a main memory.

[0059] Additionally, a computer program designed to perform the computer-implemented method 100 for determining a tolerance region 30 within an operational design domain (ODD) around a target parameter sample 10 of a parameterizable simulation model for a product is disclosed. The computer program may exist, for example, in an interpretable or compiled form. The computer program (even in part) may be downloaded into RAM of a computer (e.g., a computer system) for execution, for example, as a bit or byte sequence.

[0060] Further disclosed is a computer readable medium or signal storing and / or embodying a computer program. The medium may include, for example, one of RAM, ROM, EPROM, HDD, SSD, ... and on / in which the signal is stored. [Explanation of symbols]

[0061] 10 Target Parameter Samples 20 Large number of parameter samples 30 Tolerance area 40. The second-closest contour around target parameter sample 10

Claims

1. 1. A computer-implemented method (100) for determining a tolerance region (30) within an operational design domain (ODD) around a target parameter sample (10) of a parameterizable simulation model for a product, particularly a steer-by-wire steering system and / or a steering system for highly automated driving, comprising: calculating (130) a dissimilarity metric for each pair of the multiple pairs, whereby each pair comprises a parameterized simulation model assigned to said target parameter sample (10) and a respective one of said multiple parameterized simulation models, resulting in one distance per pair, thereby resulting in multiple distances; optionally, said dissimilarity metric is based on a gap metric, a v-gap metric and / or an L2 metric; A method (100) comprising a step of determining (140) said tolerance region (30) in said ODD around said target parameter sample (10) based on said multiple distances and a maximum allowable distance.

2. The method (100) of claim 1, wherein the target parameter sample (10) and the tolerance zone (30) define a product specimen to be manufactured for the product.

3. - outputting (150) requirements for the production of the product specimen to be manufactured based on the target parameter sample (10) and the tolerance zone (30), said requirements being met if the parameters of the product specimen to be manufactured are within the tolerance zone (30); 3. The method (100) of claim 2, wherein optionally, said requirements include an algorithm capable of checking whether said parameters of said product specimen to be manufactured are within said tolerance zone (30).

4. 4. The method (100) according to claim 1, wherein the determination (140) of the tolerance region (30) within the ODD is performed such that within the tolerance region (30), each distance to the parameterized simulation model assigned to the target parameter sample (10) is less than the maximum allowable distance.

5. 5. The method of claim 4, wherein the determination of the tolerance region within the ODD is performed such that the tolerance region is strictly or approximately maximal, and optionally the tolerance region is maximized based on a Lebesgue measure.

6. The determination (140) of the tolerance region (30) within the ODD is selecting (141) one or more parameterized simulation models whose distance to the parameterized simulation model assigned to the target parameter sample (10) sufficiently corresponds to the maximum allowable distance, The method (100) according to any one of claims 1 to 5, comprising a step of determining (142) said tolerance region (30) on the basis of said one or more selected (141) parameterized simulation models, optionally on the basis of a polytope, in particular a convex polytope, a hyperellipsoid or a hypercube.

7. The determination (140) of the tolerance region (30) within the ODD is - interpolating (143) the distances for the parameterized simulation model assigned to the target parameter samples (10), resulting in a distance mapping in the ODD; The method (100) according to any one of claims 1 to 6, comprising a step of determining (144) said tolerance region (30) on the basis of said distance mapping in said ODD, optionally on the basis of a polytope, in particular a convex polytope, a hyperellipsoid or a hypercube.

8. 8. The method (100) of claim 7, wherein the interpolation (143) of the distance is based on a fit function, the fit parameters of which are determined by a Gaussian process.

9. The method (100) of any one of claims 1 to 8, wherein the target parameter sample (10) is one from a number of target parameter samples representative of the ODD.

10. determining (110) a number of parameter samples (20) in said ODD for said product, optionally each parameter sample in said ODD comprising one or more parameters of said parameterizable simulation model; generating (120) the parameterized simulation model based on the parameterizable simulation model and the number of parameter samples (20) in the ODD, in particular by evaluating the parameterizable simulation model at each of the parameter samples (20) and / or creating a surrogate model for the parameterizable simulation model for each of the parameter samples (20); 10. The method according to claim 1, wherein the determination of the number of parameter samples in the odd distribution is based on, inter alia, pseudorandom numbers, Latin hypercube sampling, and / or Sobol sequences, so as to ensure a sufficiently uniform coverage of the odd distribution.

11. the plurality of parameter samples (20) are within a representative range of the target parameter sample (10); Optionally, the determination (110) of the multiple parameter samples (20) within the representative region is based on, inter alia, pseudorandom numbers, Latin Hypercube sampling, and / or Sobol sequences, so as to ensure a sufficiently uniform coverage of the representative region.

12. A method (100) according to any one of claims 1 to 11, comprising a step of identifying (160) one or more parameters within said tolerance zone (30) that have a relatively large influence on said product.

13. 13. A computer system designed to execute a computer-implemented method (100) for determining a tolerance region (30) within an operational design domain (ODD) around a target parameter sample (10) of a parameterizable simulation model for a product according to any one of claims 1 to 12.

14. 13. A computer program designed to perform a computer-implemented method (100) for determining a tolerance region (30) within an operational design domain (ODD) around a target parameter sample (10) of a parameterizable simulation model for a product according to any one of claims 1 to 12.

15. A computer readable medium or signal storing and / or embodying a computer program according to claim 14.

Images