DE102016108991A1 - Method and device for characterizing a component - Google Patents

Abstract

It is a method for characterizing a stress-strain behavior of a component having an inhomogeneous material and which is writable in an initial state by an initial model. The method comprises: (a) measuring a strain distribution for a measurement area of the component after applying a force to the component; (b) calculating a strain distribution for the material using a finite element model based on the applied force and the output model; (c) comparing the calculated strain distribution with the measured strain distribution; (d) correcting the initial model based on the comparison (c); and (e) iteratively repeating steps (b), (c) and (d) until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold.

Description

The present invention relates to a method and apparatus for characterizing a component, and more particularly to a method for determining local stress-strain curves defined by material parameters based on measured strains.

background

There is an increasing interest in characterizing inhomogeneous materials or components - especially when exposed to elastic-plastic stresses. Such components include, for example, welds and cold-formed components. As a result, for example, stresses from high temperatures can be simulated, as they occur, for example, in aggregates in the automotive industry or in welded pipes or pipelines in power plant construction or pressure vessels in the chemical apparatus construction.

Experiments on butt-welded tube samples have shown that under high cyclic elastic-plastic stresses (eg at strain amplitudes greater than 1.5%) failure of the component occurs in the base material and not at the seam notch, as is often the case with fatigue stresses is watching. Through systematic investigations of possible influencing factors, it has been found that locally different (cyclic) material strengths are the decisive cause for this behavior. However, in order to be able to correctly model the material behavior in the model, which forms the basis for the life test, the actual existing local material strengths (that is, material strengths, which depend on the location in the component) must be taken into account.

Using such a model, components can be calculated in principle starting from a given initial state taking into account boundary conditions. Such calculations use material laws / non-linear material models, which however contain material parameters that are mostly insufficiently known and which have to be optimized.

The real strain distribution, which can be determined for example by means of gray scale correlation, is a material-mechanically accurate basis for the determination of local material strengths. In order to be able to map the local material behavior, for example, parameters of the (cyclic) stress-strain curve are required, which are not homogeneous but localized in the component and thus take into account material inhomogeneities (such as welds).

• (a) die Spannungsverteilung ist im Querschnitt wegen werkstoffbedingter Inhomogenitäten nicht direkt bestimmbar;
• (b) die gemessene Dehnung ist die totale Dehnung und eine Trennung in elastische und plastische Anteile ist auf Basis der Messdaten schwer möglich (die Kenntnis beider Anteile wäre zur Bestimmung der Spannung geeignet); und
• (c) die Dehnung ist im Bauteilinneren nicht messbar und somit ist auch die Querkontraktion des Bauteiles nur schwer zu bestimmen (die Querkontraktionszahl wäre ansonsten zur Bestimmung der Spannung geeignet).

The following problems arise in determining the strain associated with strain:

• (a) the stress distribution is not directly determinable in cross-section due to material-related inhomogeneities;
• (b) the measured elongation is the total elongation and a separation into elastic and plastic parts is difficult on the basis of the measured data (the knowledge of both parts would be suitable for the determination of the tension); and
• (c) the strain is not measurable in the interior of the component and thus the transverse contraction of the component is difficult to determine (the transverse contraction number would otherwise be suitable for determining the voltage).

For many applications, it is also desirable to allow a direct determination of stresses and thus also of material parameters, without having to take into account errors that occur by approximations. As a result, high safety factors in the design in connection with correspondingly high material usage (thicker components) could be avoided. In addition, an improved simulation saves extensive test series in many different constellations. It would also be desirable if the corresponding trial could be performed on the sample rather than on the entire component. This would be useful in particular for series components, which all have the same inhomogeneities.

The aforementioned problems are so far only insufficiently solved by known methods.

In a conventional approach, the hardness (or load-penetration curves) at many points over an entire inhomogeneous area is determined by means of a penetrator. A disadvantage of this approach is that there is a correlation between hardness and material parameters in the (cyclic) stress-strain curve (ZSD curve) via certain assumptions or approximations. Another disadvantage of the hardness measurement are possible falsifications of hardness values due to residual stresses in the component. The Residual stresses occur particularly on welds and sometimes show a higher hardness in terms of tensile stresses than the underlying material would have without them. For conversion into strengths, however, the hardness of the underlying material is needed. Hardness measurements are also disadvantageous since hardness values are generally determined on unloaded samples, but the material changes over the service life after repeated loads (transient behavior). Renewed hardness measurements are therefore to be carried out after a further 10, 20, 30, etc. cycles of test specimens which are difficult, if at all feasible. The mentioned disadvantages also relate to the correlation between the load-penetration curve and the material parameters of the ZSD curve.

To determine the local material behavior, load-induced elastic strain measurements can also be made. However, such measurements are very expensive and are often carried out under the assumption that the material behaves constantly over the entire cross section, but this is not possible with any inhomogeneities.

Although microfabricated material samples can be worked out locally from an inhomogeneous component, this is expensive and economically of no interest in most cases.

In DE 2009 10 052 967 is presented another method in which strains are measured optically by means of 3D image correlation (gray scale correlation). In this known method, a non-linear optimization is used for the parameter identification, wherein the objective function uses the sum of the squares of the distances between the desired values and the actual values (from measurement and calculation). The optimization used there is a gradient-based method in the context of an iterative calculation method, which provides a direction and a step with respect to the change of the parameters and thus a parameter set for the next iteration step. For the determination of a large number of materials, the mentioned optimization method is very complex or useless. Another disadvantage of this approach is that this method is applicable only for constant voltages in cross section.

Therefore, there is a need for a method for determining local material properties for inhomogeneous materials, for example, to enable life-time prediction.

Summary

The above object is achieved by a method according to claim 1 and an apparatus according to claim 15. The dependent claims relate to advantageous developments of the method according to claim 1.

The present invention relates to a method for characterizing the stress-strain behavior of a component comprising a homogeneous or inhomogeneous material that is writable in an initial state by an initial model. The method comprises the following steps: (a) measuring a strain distribution for a measuring range of the component after applying a force to the component; (b) calculating a strain distribution for the component using a finite element model based on the applied force and output model (eg, taking into account the geometry of the component at the beginning of the strain measurement); (c) comparing the calculated strain distribution with the measured strain distribution; (d) correcting the initial model based on the comparison (c); and (e) iteratively repeating steps (b), (c) and (d) until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold.

In particular, the measuring range can include inhomogeneous regions of the component that can not be described by a homogeneous material. In particular, the correction comprises changing parameters which describe the concrete material at the corresponding position. Thus, the material is effectively changed at this position if it does not correctly counter the measured strain.

A strain distribution is a local assignment of strain values to areas of the component to understand. Thus, the component is not assigned a constant strain value, but individual strain positions or regions of the component are generally assigned different strain values, so that the strain as a function across the component represents a variable variable. In particular, the step of correcting comprises changing the local material based on the strain deviation of strain values, which can first be assigned to the individual regions of the component as part of the initial model, wherein the initial model can have an "arbitrary" start parameter set. Under The assumption of a homogeneous material can first be assigned to at least a portion of the component, which is then corrected based on the calculations after each iterative step. The output values may be, for example, characteristic quantities of the material material which are assigned to the individual regions based on the geometry and / or the material of the component. For example, the material parameters of the base material can be used as starting parameters.

The step of measuring may optionally include assigning measured strain values to areas of the component. These local strain values are determined in further exemplary embodiments by optically detecting a change in a pattern formed on the component by the force (gray-scale correlation or image-value correlation). However, this is merely an example of measuring strain values, and the invention is not intended to be so limited.

For example, in other embodiments, the force is a tensile force that pulls the component apart along a pulling direction, and steps (a) - (c) are repeatedly performed for a plurality of different tensile forces. As a result, different points can be obtained on a stress-strain curve in at least one region of the component. When performing steps (a) - (c), these steps can be performed sequentially for multiple forces (strain states), and subsequently iteratively improves all material parameters. In this way, only points on the local stress-strain curves are determined on the basis of the different force or strain states (this is already a part of step (d)), on which basis the material correction (d) takes place. In this way, the corresponding parameters, which enter into the material law, can be determined. Each area of the component is assigned a curve that represents the expansion behavior as a function of the applied force.

The material can be described by a material law, which represents a relationship between the stress and the strains. However, the material law also depends on parameters that depend on the respective area of the component in the case of inhomogeneous materials. Since the material law can generally depend on several parameters, several measurements are also required to uniquely determine all parameters. This can be done, for example, by applying a plurality of forces, so that the resulting multiple measurements can unambiguously determine the parameters. Therefore, in other embodiments, if the strain distribution is determined by a material law and the material law depends on a number of local material parameters (ie, depends on at least one local parameter), then the plurality of different tensile forces are given by the number of local material parameters Numbers depend on each other or are linked). In this case, the method may further comprise changing the at least one local material parameter in the initial model to thereby better match the FE model to the actual material.

The initial model is determined, for example, by parameters such as the geometry and / or the material of the component. In the initial model, it can be assumed in a first approximation that the material is homogeneous. The inhomogeneities are then determined by the iterative process and by applying multiple forces with increasing accuracy. In order to improve the output model, in further embodiments, the geometry is detected by scanning the component. Scanning provides an optical scan that can automatically capture the geometry of the part and thus provide an accurate initial model. The iterative process then delivers more accurate results.

In other embodiments, the method further includes modeling a history of the component to account for local biases of the material. As a result, the current state of stress can be taken into account.

In further embodiments, the applied force is chosen so strong that the material of the component deforms plastically. Advantageously, all areas of the measuring range (which should be characterized) deform. Since the material can be described at any position in the component by the stress-strain behavior, the material can be determined locally without knowledge of the prehistory or possibly existing prestressing.

In further embodiments, the method further includes partitioning at least a portion of the component into areas to which materials are assigned. The partitioning can be, for example adapted to the geometry or other factors (eg, the material) to improve the initial model.

In further embodiments, the method includes fixed or adaptive partitioning. Fixed partitioning does not change during the iterative process. In adaptive partitioning, at least one area is further subdivided. This can happen, for example, if, within at least one area, a difference between a maximum and a minimum value of a material parameter, which is iteratively improved by the method, is greater than a predefined limit value. However, this is not mandatory.

In further embodiments, the finite element model is based on nodes (FE nodes) and the regions for which strain values are measured include multiple FE nodes or halftone dots / measurement points are associated with a FE node. For example, a distance between adjacent FE nodes may be made larger than a distance between grid measurement points, for example, the positions of the grid measurement points may be selected to achieve this. Step (c) may include comparing the calculated strain value to a combination of a plurality of measured strain values. Therefore, this process can use multiple metrics for a FE node to assign a value to this FE node. In these assignments, for example, an averaging or weighted averaging (eg, depending on the distance from the node) can take place.

In particular, in the adaptive partitioning multiple metrics z. A FE node, in other embodiments in each iteration loop the adaptive partitioning may include changing the number of assigned measurements to a respective FE node.

In further embodiments, surface components of the measured and calculated strain distribution will be compared in step (c). This provides the correct result, since it can be assumed that constant material zones continue into the component interior (use of a symmetry). Thus, in the case of a weld, it can be assumed that the materials present on the surface continue at the end points of the weld seam without major changes to the interior. In particular, if the exemplary welding process takes place automatically, it is to be assumed that constant conditions prevail at different points along the weld, so that inhomogeneities should only form perpendicular to the weld or in the depth.

In other embodiments, when the stress path describes a hysteresis curve for cyclic strains, a number of the applied forces may be selected to calculate and / or measure multiple nodes for a hysteresis branch. Cyclic strains are repetitive stresses (increasing and decreasing the force) of the material, i. H. an applied force is increased up to a maximum force and thereafter the force is successively reduced again to a minimum force (for example, free of forces). Subsequently, the force can be increased again up to a maximum force and again reduced. This process can continue cyclically and leads to cyclic strains.

In other exemplary embodiments, the material has multiaxial stress states and the method comprises detecting additional comparative strains which are sensitive to the multiaxial stress states and which can be calculated, for example, by the Von Mises method.

The method may also be implemented or stored in the form of instructions in software or on a computer program product, the stored instructions being capable of performing the steps of the method as the method runs on a processor. Therefore, the present invention also relates to a computer program product having software code (software instructions) stored thereon configured to perform one of the above-described methods or to provide functions when the software code is executed by a processing unit. The processing unit may be any form of computer or control unit having a corresponding microprocessor capable of executing a software code.

It is understood that the named sequence of the previously defined steps means no chronological sequence or only insofar as it is absolutely necessary. The steps can be performed in any order. Likewise, some of the above steps may be performed after or before iterative enhancement.

The present invention also relates to a device for characterizing a stress-strain behavior of a component which has an inhomogeneous material or a plurality of materials and is writable in an initial state by an initial model. The device comprises a device for measuring a strain distribution for the component after applying at least one force component to the component and a processing unit. The processing unit is configured to iteratively perform the following steps: (b) calculating a strain distribution for the component using a finite element model based on the applied force and the output model; (c) comparing the calculated strain distribution with the measured strain distribution; (d) correcting the initial model based on the comparison (c); and (e) iteratively repeating steps (b), (c) and (d) until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold.

A significant advantage of embodiments is that the history of the component or locally present voltages need not be known because the load path begins at the origin of the hysteresis after load reversal. If the applied force is so strong that the material of the component is plastically deformed, the calculated strain values can be used to uniquely describe the material at the particular point of the component.

Embodiments of the present invention offer the advantage that no correlation of measured values to material parameters is necessary. In addition, the method for any vibrational games to determine the transient material behavior is easy to carry out. Furthermore, a local determination of any materials arranged arbitrarily arranged for non-constant parameters or non-constant voltages in the cross section. Thus, the method is also applicable to inhomogeneous stress / strain states such as notches.

The method according to embodiments also works with static and thus non-cyclic loading, in which the minimum state is the unloaded initial state.

In contrast to the known methods, no elaborate optimization is required in embodiments of the present invention, since both the direction and the step size for the correction of the material are implicitly specified in the context of the method according to the invention.

Further advantages of embodiments are the material-mechanical based approach that allows a direct determination of stresses and thus also of material parameters, without having to take into account errors that would occur by approximations. This leads to a precise calculation. As a result, high safety factors can be avoided during dimensioning in conjunction with correspondingly high material usage (thicker components). In addition, the improved simulation saves extensive test series in many different constellations. Trials may also be performed on the sample rather than on the entire component and are particularly suitable for series components, all of which have the same inhomogeneities.

Brief description of the figures

The embodiments of the present invention will be better understood from the following detailed description and the accompanying drawings of the different embodiments, which should not be construed as limiting the disclosure to the specific embodiments, but for explanation and understanding only.

1 shows a flowchart for a method for characterizing a stress-strain behavior of a material according to an embodiment of the present invention.

2 shows an exemplary component with a measuring range (connection area).

3A , B show three different hysteresis curves for inhomogeneous materials / material and an illustration of forming stress redistributions.

4 shows a cross-sectional view or side view through the component on which a tensile force F is applied.

5 illustrates the division of the measuring range into several grid measuring points.

6A , B show a distribution of nodes of the finite element method and illustrate an assignment of measured values to FE nodes.

7 illustrates a node-based correction of the material (here with a stress point).

8A , B illustrate adaptive and fixed partitioning.

9A , B show stress points from different states along stress-strain curves.

Detailed description

1 shows a flow chart for a method for characterizing a stress-strain behavior of a component with material inhomogeneities. The component is initially described by an initial model in the initial state. The method comprises the steps of measuring S110 an elongation distribution upon application of a force to the component and calculating S120 a strain distribution for the material using a finite element model based on the applied force and the output model. This is followed by a comparison S130 of the calculated strain distribution with the measured strain distribution. For example, this comparison can be done node by node for each FE node. Based on the comparison S130, a correction S140 of the initial model with respect to the material (material parameter) takes place. The steps S120, S130, S140 are repeated until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold value.

Embodiments thus use a combination of a strain measurement, which can be performed for example by means of optical image acquisition (3D image correlation / gray value correlation), and a finite element calculation to determine the stress of the component. The calculation is made iteratively and the material is corrected from calculation to calculation until the measured strain distribution in the model is mapped correctly (within specified limits).

In addition to the given boundary conditions (such as the applied force and the bearing), the local stress in the component is largely determined by the geometry and the material. According to embodiments, the material is determined to be the only unknown quantity using the finite element calculation. All other variables in the finite element model have to be taken into account in a realistic way. For example, the geometry may be modeled based on a scan (eg, an optical scan). Thus, current geometry coordinates are available at the start of the material parameter determination.

However, the material can not be determined directly. It is characterized by one or a plurality of local, ie, localized parameters (depending on the position in the component). Computational dependencies prevail between the local parameters, with the overall solution giving a single measured strain distribution. Unlike the conventional method from the DE 2009 10 052 967 Embodiments provide values of the material parameters for different positions in the component. This determines the parameters for an inhomogeneous material or, in other words, parameters of the material law are determined for different materials.

2 shows an exemplary component 200 that a first section 210 and a second section 220 having, through a connecting portion 230 connected to each other. The connecting section 230 For example, has a weld that the first section 210 with the second section 220 combines. The weld in the connection section 230 represents an example of a material inhomogeneity, along the connecting line between the first section 210 and the second section 220 is formed, but not necessarily perpendicular to it.

The inhomogeneities of the material may relate to intrinsic properties such as another chemical composition or crystallization.

Here, in particular, the resulting strength differences are of importance. Regardless, geometrical inhomogeneities relate, for example, to other forms of the component, such as in the 2 You can see where the component is inside the connection section 230 (Measuring range) has a smaller cross section than in the first section 210 or the second section 220 ,

Stress relocations are also present for homogeneous material and geometric inhomogeneities, when the stress exceeds the linear elastic range. Due to the material inhomogeneities in the cross-section, additional stress redistributions may occur under stresses. For example, the ratios of the instantaneous tangent modules of the local stress-strain curves are responsible for the voltage redistributions. If stresses are already present in the component (eg residual stresses), the load path starts at the beginning of the load at a point on the stress-strain curve that does not lie in the origin. This affects the ratios of the tangent modules. In the calculation, this can be taken into account by recording the "history" of the material in order to capture the local stress correctly.

A special case represents a cyclic load, in which the stress path forms a hysteresis, whereby the hysteresis branches start again after each load reversal in their origin (special case I). In this case, a consideration of the history, which is also the static load of an unloaded component (without residual stress) or at a first load (special case II) is unnecessary. The method according to the present invention is particularly applicable to these two special cases.

3A shows by way of example three different hysteresis curves, a first hysteresis curve 601 , a second hysteresis curve 602 and a third hysteresis curve 603 , Each hysteresis curve describes the dependence of stress σ as a function of strain ε, with each hysteresis curve 601 . 602 . 603 the stress profile at different positions in the component 200 correspond and represents a respective inhomogeneity. Because at the individual positions in an inhomogeneous material 200 tensions already exist, it comes to the various hysteresis curves 601 . 602 and 603 in different areas of the σ-ε-plane. For the aforementioned rearrangement, the local slopes are decisive, which are independent of the position of the reversal point in the stress-strain diagram. It is therefore possible to shift the individual hysteresis branches from their (lower) reversal point to the origin of the σ-ε plane (branches 611 . 612 . 613 on the right in 3A ). A first branch 611 corresponds to the hysteresis curve 601 , a second branch 612 corresponds to the second hysteresis curve 602 and a third branch 613 corresponds to the third hysteresis curve 603 , The initially present linear section represents the usual elastic material behavior.

The geometry of the component 200 (at load reversal) can be determined using scans and / or 3D image correlation software.

3B shows the different stress profile. For example, at a first load 620 go through a curve starting from the origin of the σ-ε-plane. When the maximum force is reached, the hysteresis curve is then cyclically stressed 630 go through, but does not come back to the origin. The 3A shows by way of example 3 possibilities of hysteresis curves for different materials in a component cross section, which ultimately lead to voltage rearrangements. Often the material is characterized by the first load path. For cyclic loads hysteresis branches can be used, whereby the parameters of the hysteresis branch are determined. Since both paths are related (see 3B ), the parameters of the "normal stress-strain curve" of the material can then be calculated (initial load).

4 shows a cross-sectional view / side view of the component 200 , wherein a force F is exerted that the first section 210 from the second section 220 "Pulls away" and thus strains in the measuring range 230 causes. Within the measuring range 230 For example, material inhomogeneities 1, 2, 3,... may be present whose stress-strain behavior is to be analyzed by the method according to the invention. The following can be assumed to be known: the force F, a strain distribution, a geometry (eg in the stress minimum at F = 0). The force does not need to disappear in the minimum strain. For example, all of the forces applied in the calculation may be force differences from a current force of the state under investigation and force in the minimum strain point / reversal point. The same applies analogously to the measured strains. It can always be based on strains that have resulted from the minimum (elongation differences).

According to embodiments, the stress is calculated from the origin (initial state, such as in the 3A right) to another state of stress (after monotone increasing load, F ≠ 0) recalculated. In addition, strains measured in an accompanying experiment are used as a database for the calculation of the strain distribution. In general, strain distributions are calculated for n states, whereby the number n of material parameters to be determined can be adapted. The initial state is needed in particular for a cyclic load, while the initial strains in the special case II (static load without prior loading) are zero (see 3B ).

wobei ε die Dehnung, σ die Spannung, der Werkstoffparameter E der Elastizitätsmodul und n' ein Parameter ist (n' < 1). Die beiden Parameter E und n' können beispielsweise als konstant über das gesamte Bauteil angenommen werden, was für viele Werkstoffgruppen realistisch ist. Daher braucht für einen Werkstoff, der dieser Gleichung genügt, lediglich der freie Werkstoffparameter K' über Messungen der Dehnungsverteilung im Anfangs- und Endzustand ermittelt zu werden. Da der Parameter K' im Allgemeinen eine ortsabhängige Funktion ist, hat er in verschiedenen Bereichen des Bauteils einen anderen Wert.The stress strain curve for cyclic stresses is described, for example, by the following Ramberg-Osgood equation:

where ε is the strain, σ is the stress, the material parameter E is the elastic modulus and n 'is a parameter (n'<1). For example, the two parameters E and n 'can be assumed to be constant over the entire component, which is realistic for many material groups. Therefore, for a material that satisfies this equation, only the free material parameter K 'needs to be determined via measurements of the strain distribution in the initial and final states. Since the parameter K 'is generally a location-dependent function, it has a different value in different areas of the component.

The material law from equation (1) is just one example. However, the invention is not limited thereto. Rather, the invention for any material law, which can be represented by a functional relationship σ = σ (ε, p _i ) or ε = ε (σ, q _i ) can be applied, where p _i , q _i (i = 1 , 2, ...) represents any number of constant or variable parameters.

Even if cyclic loading (special case I) K 'is initially a parameter of the hysteresis branch, this parameter can easily be converted into a parameter of the material law (for the relationship hysteresis and material law cf. 3B ). Since during the initial load (special case II) the load is in accordance with the material law, the original material parameters would be determined directly here. In any case, the iterative method according to the present invention always determines parameters that describe the stress path.

In order to minimize the influence of scattering on the result of the parameters, in further embodiments, several nodes on a given hysteresis branch or on the static stress-strain curve can be used. Thus, several strain distributions (each a strain distribution for a given force) are obtained for different forces. In this case, it is advantageous if the underlying strain distribution (s) represent the material behavior to be imaged with sufficient accuracy. If, for example, parameters are to be determined that describe the elastic-plastic material behavior, strains with elastic-plastic parts are used as a database.

5 illustrates the subdivision of the measuring range 230 in several grid measuring points 301 for which strain values are determined. The grid measurement points 301 may, for example, be nodes of a finite element method, but need not be. The strain measurements are usually only in one measuring range 230 of the component 200 performed, so that the material parameters only in this measuring range 230 be determined. The database for the iterative process is i. A. Surface expansions available, which can be determined for example from an optical measurement when applying the force F. For this purpose, for example, the strains are determined on the basis of a displacement field at the point. Accordingly, the surface expansions are subdivided into different components of the strain tensor, depending on the direction in which the strain occurs (ε_xx, ε_yy, ε_xy), where x and y are tangent vectors on the surface of the material and a local coordinate system with an origin in the respective grid point 301 belong. The measured strain distributions ε are then in the form of discrete values which are grid-shaped in the measuring range 230 are distributed.

6A shows a distribution of nodes 302 of the finite element network (eg in the case of a tetrahedral network) over the measuring range 230 , This distribution becomes, can be adapted to the concrete conditions (eg the geometry). When executing the finite element calculation, the strains are calculated, for example, at integration points. But they can be used as interpolated values at the surface nodes 302 be issued. Thus, in the context of the finite element method, the strains on these nodes 302 determines the strains.

6B shows an assignment of the (measured) strains at the measuring points 301 (please refer 5 ) to the nodes 302 of the finite element model. For example, this may be a circle with a radius R around each node 302 of the finite elements and the measured values of the grid measurement points in the circle 301 the corresponding node 302 be assigned. Optionally, a weighted interpolation of the (measured) strains can be carried out. For the correct assignment of the grid measuring points 301 to the FE nodes 302 be z. For example, the same (global) coordinates and the same positions in space are used in the measurement and calculation. The radius R can, for example, depending on the distance of the finite element node 302 and the available database (distance of grid measurement points 301 ) to get voted. The strain values of the assigned grid measurement points 301 can according to their distance to the FE node 302 weighted averaged so that the node can be assigned a discrete value (per measured strain component). In addition, the spacing of the grid measurement points 301 should be given and should be chosen somewhat smaller than the distance of the nodes 302 of the finite element network, allowing each node 302 at least one measured value can be assigned.

In this way, the measured and calculated values are linked together, with measurements being made to all nodes 302 of the finite element model in the measuring range 230 be assigned (not only to selected identification nodes). However, nodes that are on a partition boundary should be excluded as the result is based on different materials.

After the database in the form of discrete strain values at the FE node 302 From measurements for at least one load condition and the component geometry in the finite element model has been reproduced in detail for the initial state, the iterative determination of the local material parameters can be started.

In a first calculation, you can start with any material (homogeneous or inhomogeneous). The result itself does not depend on the initial state. However, in order to enable a quick determination of the material parameters, it is advantageous to start with suitable starting parameters. For a welded component, these may be, for example, the parameters of the base material with respect to the associated hysteresis branch. The concrete material law, which contains the parameters to be determined, is given for example and can be adapted to the corresponding requirements.

In the calculations, corresponding strain values are determined by force. Defining shift boundary conditions is less suitable because there is often no clear solution for the material. The force-controlled calculation in each case the force F is applied, which was needed in the attempt to move from the initial state to the load state under study. Since it is a continuous load, load conditions can be calculated consecutively to save computation time. For each load state to which a strain field from the measurement was provided as a database, there exists as a result of the calculation a computational solution with stresses σ and strains ε at the individual FE nodes 302 ,

The next step is a target-actual comparison, in which the measured (assigned) strains with the calculated strains at the FE nodes 302 be compared. Since based on the target-actual comparison, the correction of the material 200 for the next calculation, it is ensured that the stress and strain values lie on the stress-strain curve of the respective material (or on the associated hysteresis branch). However, this is the case only for uniaxial stress, where it is possible to work with the uniaxial stress and strain values in a simplified manner. If this is not the case, reference stresses and comparative strains should be used.

Since, as already mentioned, only surface expansions from the measurement are available, but to determine the comparative strain, the complete strain tensor is required, the missing components are supplemented using the finite element calculation.

7 illustrates the correction of the material 200 , The correction becomes the setpoint 702 with the actual value 701 and based on this, the exemplary parameter K 'is adjusted. For example, the determination of the parameter K 'from equation (1) may be done over a single point on the stress-strain curve (if E and n' are constant) or determined by a regression.

In detail, this correction can be done as follows:
If, for example, a deviation of the strains is determined on the basis of the target / actual comparison, the material is corrected, at which the calculated values 701 through the measured values 702 Getting corrected. This correction is carried out mathematically first on a node basis, the correction taking into account the equilibrium in the component cross-section. Thereby the tension becomes at the knot 302 kept constant mentally, in the form that the stress point determined from the finite element calculation 701 horizontally to the measured strain value on the stress-strain curve or on the hysteresis branch 702 is shifted (with multiaxial stress, calculated reference stress and determined comparative strain (see above) are used). In embodiments, exist for each node 302 several points on the stress-strain curve, according to the number of stress conditions used. Using a regression (method of least squares), the new material parameters (eg K ') can now be determined node by node. The corrected material parameters are subsequently used for the next iteration step.

When using comparative strains (multi-axis loading) is the nominal value 702 no pure measured value anymore. With each iteration step in which the material zones in the model approximate reality, the supplementary strain components (for which no measured values are available) are corrected, so that after a few iteration steps local material parameters as well as a complete strain and stress tensor are present.

Partitioning in material zones can be done in two ways: adaptive or with a fixed schedule for all iterations. But it can also be done in other ways.

8A illustrates the adaptive partitioning of the measuring range 230 in which a variable subdivision into areas 801 . 802 . 803 , ... is possible. In this case, there is a horizontal and / or vertical division of the measuring range 230 or an existing partition. This subdivision can be done, for example, when a maximum and a minimum value for a relevant material parameter at the in the areas 801 . 802 , ... lying by an amount that is greater than a predetermined threshold (threshold). For the material law according to equation (1), for example, a further partitioning can take place if the following condition is met: K'_max - K'_min>ΔK'_prescription.

The limit ΔK'_vorgabe can be arbitrarily but meaningfully specified (for example, between 5 and 150 are or about 75 or 15).

For example, the yield strength / yield point can be specified as decisive. However, it is also possible to specify a separate limit value for all material parameters, so that the parameter whose limit value is exceeded is automatically decisive for the change in the partition. The partitioning is carried out until the limit value is adhered to for all partitions, including the newly formed one. This results in different sizes for the individual partition areas 801 . 802 . 803 , The limit should be set so that meaningful partitions can result. For example, scatters and sensitivities of the respective material parameter play a role here. In the same way, it is also possible to adaptively increase the partition in one or both directions (eg if a lower limit for the difference is undershot).

8B shows a fixed division 810 in which, in contrast to the adaptive classification, the partitioning is performed only once before the start of the iteration process. For example, the resulting area division may represent a regular grid. For example, the grating can be chosen to be very fine over the entire measuring range in order to allow the formation of local material zones at every point. This results in a very fine network throughout the measuring range. However, the grid can also be adapted to the geometry of the component. A fixed partitioning has the advantage that the assignment of the measured values to the FE nodes has to be done only once, since partitions and thus also the finite element network remain the same over all iterations. This also makes it easy to check the convergence of the solution.

In both methods, the partition material is determined by regression based on the stress points of all nodes within the partition

9A shows this determination for stress points from different states. In this case, for example, three actual states (actual state 1, actual state 2 and actual state 3) to three target states compared. As a result, the material (or material parameters) in the finite element model is changed so that a subsequent iteration step results in an approximation of the actual states to the desired states.

9B shows stress points from different states (for different strains) and at a node 1 and a node 2 within a partition. Since within a given partition the material is considered to be homogeneous, the stress points also lie with the various ones along the stress-strain curve shown.

In order to be able to determine the material as locally as possible, the adaptive partitioning should have the partitions of equal length (in relation to the considered surface). Therefore, it may be useful to divide a measuring range whose length and width differ greatly before the program start in approximately uniform areas. If, for example, the positions of the material zones are not known in advance - which is often the case - adaptive partitioning has the advantage that large areas of a homogenous material can be coarsely networked since a large partition can be used.

• (1) Zuweisung der Messwerte zu den FE-Knoten (nur bei adaptiver Partitionierung erneut notwendig; berücksichtigt werden nur Knoten innerhalb einer Partition)
• (2) Soll-Ist-Vergleich
• (3) Korrektur des Werkstoffs (zunächst knotenweise)
• (4) Gegebenenfalls adaptive Partitionierung
• (5) Bestimmung eines Partitionswerkstoffs auf Basis der ermittelten Beanspruchungspunkte
• (6) Berechnung mit neu erstellten Werkstoffzonen

Following partitioning, the model can be calculated using the newly created material zones and the process starts from the beginning, by a sequence of steps:

• (1) Assignment of the measured values to the FE nodes (only necessary with adaptive partitioning, only nodes within a partition are considered)
• (2) target-actual comparison
• (3) correction of the material (initially node by node)
• (4) Optionally adaptive partitioning
• (5) Determination of a partition material based on the determined stress points
• (6) Calculation with newly created material zones

This iteration loop is repeated until the deviation from calculated to measured strain for each stress point is below the specified limit value (target / actual comparison).

• 1) Messung der Dehnung im werkstoffinhomogenen Bereich der Probe (des Bauteils) mittels 3D-Bildkorrelation im Zustand minimaler und in zumindest einem Zustand maximaler Belastung der Probe (z. B. auf einem monoton steigenden Beanspruchungspfad). Dies kann beispielsweise über zumindest zwei Bilder erfolgen, die entsprechend ausgewertet werden, um die lokalen Dehnungszustände zu erfassen.
• 2) Modellierung der Bauteilgeometrie entsprechend realer Probe im zugrunde gelegten Ausgangszustand.
• 3) Berechnung der Spannungen und Dehnungen. Dieser Schritt kann beispielsweise von einem homogenen Werkstoff aus gestartet werden.
• 4) Zuweisung der lokal gemessenen Dehnungen zu dem Knoten des Finite-Elemente-Modells.
• 5) Soll-Ist-Vergleich der gemessenen und berechneten Dehnungen.
• 6) Korrektur bzw. eine Bestimmung des Werkstoffs für jeden Knoten (und somit lokal) auf Basis des Soll-Ist-Vergleichs und unter Berücksichtigung des Gleichgewichts (konstant gehaltene Spannung), wozu konventionelle Optimierungsverfahren nicht erforderlich sind.
• 7) Nutzung einer festen oder einer adaptiven Partitionierung, bei der der entsprechende Messbereich oder die Probe in Werkstoffzonen unterteilt werden.
• 8) Berechnung der Spannungen und Dehnungen der Probe mit den neu bestimmten lokalen Werkstoffparametern.
• 9) Wiederholung der Schritte 4) bis 8) solange bis die berechneten Dehnungen den gemessenen Dehnungen entsprechen oder eine entsprechende Abweichung minimal wird oder unterhalb eines Schwellenwertes liegt.

Embodiments of the present invention may be summarized as follows: The invention describes a material-mechanically based iterative method that combines optically measured local strains with local stresses from a finite element simulation of a geometrically detailed mapped component. In the iterative method according to the present invention, the following steps can be carried out:

• 1) Measurement of the strain in the material-inhomogeneous region of the sample (of the component) by means of 3D image correlation in the state of minimum and in at least one state of maximum load of the sample (eg on a monotonically increasing stress path). This can be done for example via at least two images, which are evaluated accordingly to detect the local strain states.
• 2) Modeling of the component geometry corresponding to the real sample in the initial state.
• 3) Calculation of stresses and strains. This step can be started, for example, from a homogeneous material.
• 4) Assignment of locally measured strains to the node of the finite element model.
• 5) Target-actual comparison of the measured and calculated strains.
• 6) Correction or a determination of the material for each node (and thus local) based on the target-actual comparison and taking into account the equilibrium (held constant voltage), which conventional optimization methods are not required.
• 7) Use of a fixed or an adaptive partitioning, in which the corresponding measuring range or sample is divided into material zones.
• 8) Calculation of the stresses and strains of the sample with the newly determined local material parameters.
• 9) repetition of steps 4) to 8) until the calculated strains correspond to the measured strains or a corresponding deviation is minimal or below a threshold value.

Steps 3) to 9) may be implemented in software, for example, and performed automatically.

The order in which the steps were mentioned in the description does not imply a chronological order or only insofar as it is absolutely necessary. Steps can also be performed as sub-steps to previously mentioned steps.

The features of the invention disclosed in the description, the claims and the figures may be essential for the realization of the invention either individually or in any combination.

LIST OF REFERENCE NUMBERS

200

component

210

first section of the component

220

second section of the component

230

Connection area (measuring range)

301

Grid measurement points

302

Node of the finite element network (FE node)

R

Radius for assignment of the measured values to the FE nodes

601, 602, 603

hysteresis

611, 612, 613

assigned hysteresis branches

620

Curve for initial load

630

Curve for a hysteresis

801, 802, 803

adaptive partitioning

810

fixed partitioning

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 200910052967 [0012, 0052]

Claims (15)

Method for characterizing a stress-strain behavior of a component which has a homogeneous or inhomogeneous material or multiple materials and is writable in an initial state by an initial model, comprising the following steps: (A) measuring a strain distribution for a measuring range of the component after applying a force to the component; (b) calculating a strain distribution for the component using a finite element model based on the applied force and the output model; (c) comparing the calculated strain distribution with the measured strain distribution; (d) correcting the initial model based on the comparison (c); and (e) iteratively repeating steps (b), (c) and (d) until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold. The method of claim 1, wherein step (a) comprises the step of: (a1) optically detecting a change of a pattern formed on the component to determine local expansions of the component based on the change. The method of claim 1 or claim 2, wherein the force is a tensile force (F) that pulls the component along a pulling direction, and steps (a) - (c) are repeatedly performed for a plurality of different tensile forces to form different points to obtain a stress-strain curve in at least one region of the component. The method of claim 3, wherein the strain distribution depends on the material and the material law depends on at least one local material parameter and the plurality of different tensile forces are associated with a number of the local material parameters, and wherein step (d) is changing the at least one local material parameter in the Starting model includes. Method according to one of the preceding claims, wherein the step (b) further comprises the following step: (b0) scanning a geometry of the component to improve the output model. Method according to one of the preceding claims, wherein the applied force is chosen so strong that the material of the component deforms plastically. Method according to one of the preceding claims, wherein in step (c) surface components of the measured and calculated strain distribution are compared. Method according to one of the preceding claims, further comprising the following step: (f) partitioning at least a portion of the component into areas, wherein step (b) further comprises the step of: (b2) assigning corrected stress-strain curves based on the measured strain distribution of a strain value to each region to improve the initial model. The method of claim 8, wherein step (f) comprises adaptive partitioning in which at least one region is further subdivided if within the at least one region a difference between a maximum and a minimum value of a material parameter is greater than a predetermined limit. The method of claim 8 or claim 9, wherein the finite element model is node based, and step (a) comprises measuring at measurement points, wherein a distance between the measurement points is selected to be smaller than a distance between adjacent nodes, and the step (8) c) comprises comparing the calculated strain value with a combination of a plurality of measured strain values. The method of claim 10, wherein in each iteration loop, the adaptive partitioning comprises changing the number of assigned measurements to a respective finite element node. Method according to one of claims 8 to 11, when dependent on claim 3, wherein the material law describes a hysteresis curve at cyclic strains and a number of measured or calculated strain distributions for the various applied forces is chosen so that several nodes are calculated or measured for a hysteresis. Method according to one of the preceding claims, wherein the component multiaxial stress states and the method further comprises the step: Capture of comparative strains that are sensitive to multiaxial stress states. A computer program product having software code stored thereon and configured to perform the method of any one of claims 1 to 13 when the software code is executed by a processing unit. Device for characterizing a stress-strain behavior of a component which comprises an inhomogeneous material or multiple materials and is describable in an initial state by an initial model, having the following features: a device for measuring a strain distribution for a measuring range of the component after applying at least one force component to the component; a processing unit configured to iteratively execute the following steps: (b) calculating a strain distribution for the component using a finite element model based on the applied force and the initial model; (c) comparing the calculated strain distribution with the measured strain distribution; (d) correcting the initial model based on the comparison (c); and (e) iteratively repeating steps (b), (c) and (d) until a deviation between the measured strain distribution and the calculated strain distribution is below a threshold.

Images