DE102014116034A1 - Device and method for determining mechanical properties of a pliable object to be measured - Google Patents

Abstract

The invention relates to a device for determining mechanical properties, for example the natural oscillation frequency, the inherent oscillation shape, the modal mass and / or the modal damping, of a pliable measurement object (210), comprising: a first fixing device (202), with which the measurement object (210) A second fixing device (203) with which the measurement object (210) is releasably fixable on its side opposite the first side, wherein the measurement object (210) in a fixed state in the first fixing device (210) 202) and the second fixing device F2 (203) and otherwise free-hanging, wherein the first fixing device (202) is connected to a first actuator (204) and mechanical through the first actuator (202) via the first fixing device (202) Suggestions in the measuring object (210) can be introduced, a control unit (206) for controlling the first actuator (206), an optical sensor system (201) for scanning a 3D surface (209) of the measurement object (210) in the fixed state and for determining a number n 3D measurement points (xi (t), yi (t), zi (t) representing the surface (209) )) with i = 1, 2, ... n, and time t, a first evaluation unit (207), with the basis of the 3D measuring points (xi (t), yi (t), zi (t)) a Time series of models M (t) of the 3D surface (209) of the measuring object (210) is determined, and a second evaluation unit (208), based on the models M (t), the mechanical properties of the measuring object (210) by means of a Modal analysis can be determined.

Description

The invention relates to a device and a method for determining mechanical properties, for example, the natural vibration frequency, the natural vibration mode, the modal mass, the modal damping, and / or the deflection of a pliable object to be measured.

Such properties of measurement objects or components are used, for example, in the context of quality control during component production, or in robot-assisted production planning.

The object is to specify an apparatus and a method for determining mechanical properties, for example the natural oscillation frequency, the inherent oscillation form, the modal mass, the modal damping, and / or the deflection of a pliable object to be measured which reliably and quickly determine the mechanical properties of the object DUT allows.

The term "measuring object" is to be interpreted broadly. In particular, it comprises planar components and components made of fiber composite materials (for example CFRP components), which are increasingly being used today in the automotive industry or in the aerospace industry.

The invention results from the independent claims. Advantageous developments and refinements are the subject of the dependent claims. Other features, applications and advantages of the invention will become apparent from the following description, as well as the explanation of embodiments of the invention, which are illustrated in the figures.

According to a first aspect of the invention, an apparatus for determining mechanical properties, for example the natural vibration frequency, the natural vibration mode, the modal mass and / or the modal damping, of a pliable measurement object is proposed. The device comprises a first fixing device with which the object to be measured is releasably fixable on a first side; a second fixing device with which the measurement object is releasably fixable on its second side opposite the first side, wherein the measurement object is fixed in a fixed state in the first fixing device and the second fixing device and otherwise hangs freely. The first fixing device is connected to a first actuator, wherein mechanical excitations can be introduced into the measurement object by the first actuator via the first fixing device. The device further comprises a control unit for controlling the first actuator; an optical sensor system for scanning a 3D surface of the measurement object in the fixed state and for determining a number n of the 3D measurement points representing the surface (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t; a first evaluation unit with which a time series of models M (t) of the 3D surface of the measurement object is determined on the basis of the 3D measurement points (x _i (t), y _i (t), z _i (t)); and a second evaluation unit, with which the mechanical properties of the test object are determined by means of a modal analysis on the basis of the models M (t).

The fixing devices can be designed, for example, as clamps, screw connections or as a gripper (end effector of a robot arm). They can serve or be designed for point fixation or fixation of an edge or a section of the measurement object.

The first actuator advantageously permits a 1D, 2D or 3D movement of the first fixing device and thus a corresponding movement of the part of the measuring object articulated by the first fixing device. By this movement, the measurement object is mechanically excited to vibrate.

In a simple embodiment variant, the second fixing device is arranged stationary so that mechanical excitations are introduced into the measuring object only via the first fixing device, while the part of the measuring object fixed in a stationary manner by the second fixing device essentially reflects incoming mechanical oscillations. Alternatively, the second fixing device or its mechanical linkage can be designed in such a way that it essentially yields mechanical excitations of the measurement object as a quasi "free end". As a result, in the extreme case, no reflection of mechanical waves at the second fixing device.

The control of the first actuator is carried out by the control unit, according to a predetermined task position. Advantageously, the first actuator is controlled in such a way that the first fixing device performs a deflection in accordance with a step signal, a rectangular signal or a sine signal. Of course, the mechanical stimuli (signal shapes, frequencies, etc.) that can be introduced into the object to be measured via the first fixing device can basically be specified virtually as desired.

Advantageously, the second fixing device is connected to a second actuator, wherein mechanical excitations can be introduced into the measurement object by the second actuator via the second fixing device, and wherein the control unit is designed and set up to control the second actuator. The second actuator advantageously allows a 1D, 2D or 3D movement of the second fixing device and thus a corresponding movement of the by the second fixing device articulated part of the measurement object. By a correspondingly coordinated control of the first and the second actuator, a multiplicity of different mechanical excitation states can be generated in a targeted manner, and their evaluation, as described below, determines different mechanical properties of the measurement object.

The optical sensor system includes an optical sensor which is advantageously a 3D laser scanner, or an optical 3D camera, for example., A Microsoft ^® Kinect RGB-D sensor, for scanning the 3D surface (eg. A top or bottom or a portion thereof ) of the measurement object. On the basis of the sampling data generated during the sampling, the sensor system determines a number n of the sampled surface 3D measuring points (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t.

Is the sensor system more than the 3D surface. If a part of it is detected, if, for example, additionally the fixing devices or an environment are detected, then the captured image data is advantageously segmented in such a way that only the region of interest of the 3D surface is provided for further evaluation. The measuring points (x _i (t), y _i (t), z _i (t)) are advantageously detected or determined only in the region of interest.

The first evaluation unit, based on the acquired time series of 3D measurement points (x _i (t), y _i (t), z _i (t)), determines a time series of models M (t) or approximations of the 3D surface of the measurement object ,

In order to determine such a model M (t) or the approximation of the 3D shape of the surface, the first evaluation unit is embodied and set up in an advantageous development such that subsequent steps are carried out. In a first step, a polynomial-based approximation A ₁ (t) of the 3D measuring points (x _{i '} (t), y _i (t), z _i (t)) is determined by means of a least-squares method. In a second step, the 3D measuring points (x _i (t), y _i (t), z _i (t)) are smoothed to produce smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)), where k = 1, 2, ... m and m ≤ n. In a third step is carried out for each point of the smoothed 3D measuring points (x _k (t), y _k (t), for _k (t)) determining a normal direction n _k (t) ((x _k (t), y _k (t), z _k (t)) with respect to the determined first approximation A ₁ (t) for the smoothed 3D measurement points (x _k (t), y _k (t), z _k (t)) iteratively executing a RANSAC algorithm for determining a polynomial-based approximation A ₂ (t) of the 3D surface, which is represented by an im Subset ( _xf (t), _yf (t), _zf (t)) of the 3D measurement points ( _xk (t), _yk (t), _zk (t)) determined by the RANSAC algorithm with f = 1, 2, ... l and l ≤ m. In a fifth step, a polynomial-based approximation A ₃ (t) is determined for the subset (x _f (t), y _f (t), z _f (t)) by means of a least squares method. In a sixth step, an iterative reduction of the order of the polynomial-based approximation A ₃ (t) to determine an optimized approximation A ₃ * (t) and to determine an associated optimized subset (x _g (t), y _g (t), z _g (t)), which is determined by comparing the respective approximation A ₃ * (t) with the 3D measuring points (x _k (t), y _k (t), z _k (t)), where g = 1, 2, ... p and p ≥ l. Finally, in a process loop, the preceding steps are repeated starting with the fifth step, wherein the approximation A ₃ (t) instead of for the subset (x _f (t), y _f (t), z _f (t)) for the optimized subset (x _g (t), y _g (t), z _g (t)) is determined until the number p of the points of the subset (x _g (t), y _g (t), z _g (t )) does not increase further compared with the previous determination of the subset (x _g (t), y _g (t), z _g (t)), whereby the model M (t) of the 3D surface of the measurement object is characterized by the subset thus optimized (x _g (t), y _g (t), z _g (t)) 'associated approximation A ₃ *' (t) is given.

This embodiment and device of the first evaluation unit also solves the problem of a three-dimensional point cloud, as determined by the 3D measurement points (x _{i '} (t), y _i (t), z _i (t) determined by the optical sensor system). and the highly noisy measurements may be to find the largest connected, smooth surface. At the same time, a polynomial model A ₃ * '(t) is determined whose parameters mathematically describe the found 3D surface of the measurement object.

The above processing steps advantageously implemented in the first evaluation unit can basically be used for all areas in which 3D surfaces are to be extracted from three-dimensional point clouds and / or a (mathematical) model is to be found which describes this 3D surface. The method can be used in particular for modeling and segmenting a carbon-fiber-plastic (CFRP) structure from a point cloud obtained with a 3D optical sensor (for example, by a Kinect sensor). In the present case, this modeling is used in particular to obtain material properties of a measurement object, for example a CFRP structure, from the measurement data.

For this purpose, a series of time-discrete models M (t ₀ + Δt · q), with q = 0, 1, 2, 3,..., And Δt time increment, are typically determined by the first evaluation unit and made available for further processing. Of course, the choice of the time increment .DELTA.t determines the evaluable frequency spectrum of mechanical vibrations of the measurement object. In this respect, the frequency spectrum of the mechanical excitation of the measurement object and the choice of the time increment .DELTA.t are to be matched advantageously.

The models M (t ₀ + Δt · q) thus indicate a time series of approximated 3D shapes of the surface of the measurement object. Thus, a time series of deflections is also known for each pixel of the optical sensor or each pixel of the approximated 3D shapes. Depending on the size of a pixel, this can be, for example, several tens or hundreds of thousands etc. points. By contrast, the number of n 3D measurement points (x _i (t), y _i (t), z _i (t)) detected by the optical sensor system or the time series of deflections of these 3D measurement points detected thereby is typically a few orders of magnitude smaller , then for n typically: n ≥ 36.

The second evaluation unit now determines the mechanical properties of the DUT by means of modal analysis on the basis of the models M (t) = M (t ₀ + Δt · q). The term "modal analysis" is well-known and includes the experimental or numerical characterization of the dynamic behavior of vibratory systems with the help of their natural vibration quantities (modal parameters) natural frequency, natural mode, modal mass and modal damping. For further explanation, reference is made to the prior art.

The proposed device can be used in particular for the quality control of surfaces. It is also conceivable to use the proposed device for the detection, identification and localization of workpieces.

An advantageous development of the device is characterized in that the first evaluation unit is designed and set up such that the least-squares method in the first step and / or in the fifth step is a moving-least-square method or a weighted-least squares Procedure is / are.

An advantageous development of the device is characterized in that the first evaluation unit is designed and set up such that the smoothing by projection of each 3D measuring point (x _i (t), y _i (t), z _i (t)) to a local area of the approximation A ₁ (t) takes place, which represents a local environment of the respective 3D measuring point (x _i (t), y _i (t), z _i (t)).

An advantageous development of the device is characterized in that the first evaluation unit is designed and set up such that in the RANSAC algorithm for each 3D measuring point (x _k (t), y _k (t), z _k (t)) an Euclidean distance _Dk (t) ( _xk (t), _yk (t), _zk (t)) between the measuring point ( _xk (t), _yk (t), _zk (t)) and a corresponding point (x _k (t), y _k (t), z _k (t)) _{A2 of} the approximation A ₂ (t), and an angular difference W _k (t) (x _k (t), y _k ( t), z _k (t)) between the determined normal direction n _k (t) (x _k (t), y _k (t), z _k (t)) and a normal direction n _{k, A2} (t) (x _k (t), y _k (t), z _k (t)) of the corresponding point (x _k (t), y _k (t), z _k (t)) _{A2 of} the approximation A ₂ (t) is determined.

An advantageous development of the device is characterized in that the first evaluation unit is designed and set up such that the subset (x _f (t), y _f (t), z _f (t)) only those 3D measuring points (x _k (t), _yk (t), _zk (t)), for which: D _k (t) (x _k (t), y _k (t), z _k (t)) <G ₁ , and W _k (t) (x _k (t), y _k (t), z _k (t)) <G ₂ , where G ₁ and G _{2 are} predetermined limits.

An advantageous development of the device is characterized in that the first evaluation unit is designed and set up such that a segmentation of the optimized subset (x _g (t), y _g (t), z _g (t)) 'from the 3D Measuring points (x _k (t), y _k (t), z _k (t)) takes place.

A second aspect of the invention relates to a method for determining mechanical properties, for example the natural oscillation frequency, the inherent oscillation form, the modal mass and / or the modal damping, of a limp test object. The proposed method comprises the following steps. In a first step, the measuring object is fixed with a first side to a first fixing device and with its second side opposite the first side to a second fixing device, wherein the measuring object otherwise hangs freely. In a second step, an introduction of a mechanical excitation into the measured object takes place via the first fixing device. In a third step, a 3D surface of the measurement object is scanned in the fixed state with an optical sensor system for determining a number n of the 3D measurement points representing the surface (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t. In a fourth step, a time series of models M (t) of the 3D surface of the measurement object is determined on the basis of the determined 3D measurement points (x _i (t), y _i (t), z _i (t)). In a fifth step, on the basis of the models M (t), the mechanical properties of the test object are determined by means of a modal analysis.

An advantageous development of the above method is characterized in that determining the time series of models M (t) of the 3D surface of the measurement object comprises the following steps: determining a polynomial-based approximation A ₁ (t) of the 3D measurement points (x _{i '} ( t), y _i (t), z _i (t)) by means of a least squares method; Smoothing the 3D measuring points (x _i (t), y _i (t), z _i (t)) to produce smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)), with k = 1, 2, ... m and m ≤ n; for each point of the smoothed 3D measurement points ( _xk (t), _yk (t), _zk (t)) determining a Normal direction n _k (t) ((x _k (t), y _k (t), z _k (t)) with respect to the determined first approximation A ₁ (t); for the smoothed 3D measuring points (x _k (t) , y _k (t), z _k (t)) iteratively executing a RANSAC algorithm for determining a polynomial-based approximation A ₂ (t) of the 3D surface, which is determined by a subset (x _f (t) determined as part of the RANSAC algorithm ), y _f (t), z _f (t)) of the 3D measuring points (x _k (t), y _k (t), z _k (t)) is determined, with f = 1, 2, ... l and l ≤ m; determining a polynomial-based approximation A ₃ (t) for the subset (x _f (t), y _f (t), z _f (t)) by a least-squares method; iteratively reducing the order of polynomial-based approximation A ₃ (t) for determining an optimized approximation A ₃ * (t) and for determining an associated optimized subset (x _g (t), y _g (t), z _g (t)) obtained by comparing the respective ones Approximation A ₃ * (t) with the 3D measuring points (x _k (t), y _k (t), z _k (t)) is determined, with g = 1, 2, ... p and p ≥ l; and repeating the steps beginning with the determination of the polynomial-based approximation A ₃ (t), wherein the approximation A ₃ (t) instead of for the subset (x _f (t), y _f (t), z _f (t)) for the optimized subset (x _g (t), y _g (t), z _g (t)) is determined as long as the number p of the points of the subset (x _g (t), y _g (t), z _g ( t)) with respect to the previous determination of the subset (x _g (t), y _g (t), z _g (t)) does not increase further, wherein the model M (t) of the 3D surface of the measurement object by the thus optimized Subset (x _g (t), y _g (t), z _g (t)) 'associated approximation A ₃ *' (t) is given.

A third aspect of the invention relates to a method for determining a model of a 3D surface of a measurement object, comprising the following steps. In a first step, the 3D surface of the measurement object is scanned with an optical sensor system for determining a number n of the 3D measurement points (x _i , y _i , z _i ) representing the surface with i = 1, 2,... In a second step, a polynomial-based approximation A _{1 of} the 3D measuring points (x _i , y _i , z _i ) is determined by means of a least-squares method. In a third step, the 3D measuring points (x _i , y _i , z _i ) are smoothed to produce smoothed 3D measuring points (x _k , y _k , z _k ), where k = 1, 2, m ≤ n. In a fourth step, for each point of the smoothed 3D measurement points (x _k , y _k , z _k ), a normal direction n _k (x _k , y _k , z _k ) is determined based on the determined first approximation A ₁ . In a fifth step is performed for the smoothed 3D measuring points (x _k, y _k, z _k), an iterative performing a RANSAC algorithm for determining a polynombasierten approximation A ₂ of the 3D surface that determined by a part of the RANSAC algorithm Subset (x _f , y _f , z _f ) of the 3D measuring points (x _k , y _k , z _k ) is determined, with f = 1, 2, ... l and l ≤ m. In a sixth step, a polynomial-based approximation A ₃ for the subset (x _f , y _f , z _f ) is determined by means of a least-squares method. In a seventh step, the order of the polynomial-based approximation A ₃ is iteratively reduced in order to determine an optimized approximation A ₃ * and to determine an associated optimized subset (x _g , y _g , z _g ) which is determined by comparing the respective approximation A ₃ * with the 3D measuring points (x _k , y _k , z _k ) is determined, with g = 1, 2, ... p and p ≥ l. Furthermore, in an eighth step, the steps are repeated starting with the sixth step, with the approximation A ₃ now replacing the subset (x _f , y _f , z _f ) for the optimized subset (x _g , y _g , z _g ). is determined until the number p of the points of the subset (x _g , y _g , z _g ) over the previous determination of the subset (x _g , y _g , z _g ) does not increase further, the model of the 3D surface of the Measuring object through which the thus optimized subset (x _g , y _g , z _g ) 'associated approximation A ₃ *' is given.

A development of the method is characterized in that the smoothing by projection of each 3D measuring point (x _i , y _i , z _i ) takes place on a local surface of the approximation A ₁ , which is a local environment of the respective 3D measuring point (x _i , y _i , z _i ).

A development of the method is characterized in that in the RANSAC algorithm for each 3D measuring point (x _k , y _k , z _k ) a Euclidean distance D _k (x _k , y _k , z _k ) between the measuring point (x _k , y _k , z _k ) and a corresponding point (x _k , y _k , z _k ) _{A2 of} the approximation A ₂ , and an angular difference W _k (x _k , y _k , z _k ) between the determined normal direction n _k (x _k , y _k , z _k ) and a normal direction n _{k, A2} (x _k , y _k , z _k ) of the corresponding point (x _k , y _k , z _k ) _{A2 of} the approximation A _{2 is} determined.

A further development of the method is characterized in that the subset (x _f, y _f, z _f) only those 3D measuring points (x _k, y _k, z _k) comprises, in which: D _k (x _k, y _k , z _k ) <G ₁ , and W _k (x _k , y _k , z _k ) <G ₂ , where G ₁ and G _{2 are} predetermined limits.

A development of the method is characterized in that, after step eight, the optimized subset (x _g , y _g , z _g ) is segmented from the 3D measuring points (x _k , y _k , z _k ).

A development of the method is characterized in that the approximation A ₃ * 'is provided for further processing.

A refinement of the method is characterized in that the 3D surface of the measurement object is time-changeable, the optical sensor is a time series of 3D measurement points representing the surface (x _i (t), y _i (t), z _i (t) ), wherein for each time point t of the time series an approximation A ₃ * '(t) is determined, and on the basis of the approximation A ₃ *' (t) material properties of the measurement object are determined.

With the proposed method, it is possible to perform a segmentation and simultaneous modeling of a surface. The best match of noisy measurement data with a parametric model is also found in the case of obscurations and disturbing objects in the foreground and background.

Due to fitting polynomials with a RANSAC method, any (non-closed) surface can be approximated, so no prior knowledge of the shape of the object is necessary. An optimization step of the polynomial degree allows a more efficient and more precise modeling. This RANSAC segmentation and modeling technique is used as the seed of a maximization-expectation loop, which finds the global optimum, the model that best describes the object. This reduces inaccuracies in the initial estimate.

Initializing with a RANSAC procedure ensures, on the one hand, that a globally optimal model of the structure is found, even if the entire structure was not segmented in the first RANSAC step. On the other hand, this offers a time advantage, as a purely RANSAC-based algorithm is an iterative algorithm that needs to be run very often to ensure that the optimal model is found.

Due to the expectation-maximization loop, the proposed method represents a closed optimization loop with which it can be ensured that a global optimum can be found despite very noisy data.

Flexibility of the procedure

The method allows modeling a structure without having any prior knowledge of the shape of the structure. In contrast to known RANSAC methods, in which the form of the structure to be modeled must be known, this makes possible a very broad application possibility to all structures that can be modeled as a bivariate polynomial, whose surface thus represents a non-closed, smooth surface. This includes many surface primitives such as planes.

On the other hand, prior knowledge about the form of the structure can be incorporated into the algorithm, for example, the maximum degree of the model polynomial in each direction can be limited independently. This is the case when it is known that a material has only one deflection in one direction. This allows a faster and more accurate calculation of the model.

clustering

This feature ensures that the largest contiguous area is extracted.

Polynomial degree Optimization

The degree optimization step gradually reduces the maximum degree of the model polynomial and then re-optimizes the model. As a result, the model that describes the model with the smallest degree is preferred. This prevents higher polynomial terms from being used as necessary, which is a speed advantage in the calculation.

Another aspect of the invention relates to a computer system having a data processing device, wherein the data processing device is configured such that a method as described above is performed on the data processing device.

Another aspect of the invention relates to a digital storage medium having electronically readable control signals, wherein the control signals may interact with a programmable computer system such that a method as described above is performed.

Another aspect of the invention relates to a computer program product having program code stored on a machine-readable medium for carrying out a method as described above when the program code is executed on a data processing device.

Another aspect of the invention relates to a computer program having program codes for carrying out a method as described above when the program runs on a data processing device. For this, the data processing device can be configured as any known from the prior art computer system.

Further advantages, features and details will become apparent from the following description in which - where appropriate, with reference to the drawings - at least one embodiment is described in detail. The same, similar and / or functionally identical parts are provided with the same reference numerals. Show it:

1 a schematic structure of a proposed device for detection mechanical properties, such as the natural vibration frequency, the natural mode, the modal mass and / or the modal damping, a pliable object to be measured,

2 a schematic process flow of a proposed method for determining mechanical properties, such as the natural frequency, the natural mode, the modal mass and / or modal damping, a pliable object to be measured, and

3 a schematic process flow of a proposed method for determining a model of a 3D surface of a measurement object.

1 shows a schematic structure of a proposed device for determining mechanical properties, such as the natural vibration frequency, the natural vibration mode, the modal mass and / or the modal damping, a pliable object to be measured.

The device comprises a first fixing device 202 with which the measurement object 210 in the present case, a CFRP plate, which is releasably fixed to a first side, a second fixing device 203 with which the measurement object 210 is fixed releasably fixed at its second side opposite the first side, wherein the measurement object 210 Otherwise, it hangs freely in the illustrated fixed state, ie, if necessary, it deflects slightly due to gravity (not shown). The first fixing device 202 is an end effector of a 3-dimensionally movable first actuator 204 , in this case, a first arm of an industrial robot 211 , Through the first actuator / robot arm 204 are about the first fixing device 202 mechanical excitations in the measurement object 210 can be introduced, for example, by vertical deflections, as indicated by the black arrows shown. The second fixing device 203 is also an end effector of a 3-dimensionally movable second actuator 205 , in this case a second arm of the industrial robot 211 , Through the second actuator 205 are about the second fixing device 203 also mechanical excitations in the measurement object 210 recoverable.

The device further comprises a control unit 206 for controlling the industrial robot 211 or the first actuator 204 and the second actuator 205 , as well as an optical sensor system 201 for scanning a 3D surface 209 of the measurement object 210 in the fixed state and to determine a number n the surface 209 representing 3D measurement points (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t.

The device further comprises a first evaluation unit 207 in which a time series of models M (t) of the 3D surface are determined on the basis of the 3D measuring points (x _i (t), y _i (t), z _i (t)) detected by the sensor system ( 209 ) of the measurement object 210 is determined, and a second evaluation unit 208 , with the basis of the determined models M (t), the mechanical properties of the DUT 210 be determined by a modal analysis.

To determine the models M (t) of the 3D surface 209 of the measurement object 210 is the first evaluation unit 207 for carrying out the following steps: determining a polynomial-based approximation A ₁ (t) of the 3D measuring points (x _{i '} (t), y _i (t), z _i (t)) by means of a least squares method; Smoothing the 3D measuring points (x _i (t), y _i (t), z _i (t)) to produce smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)), with k = 1, 2, ... m and m ≤ n; for each point of the smoothed 3D measurement points ( _xk (t), _yk (t), _zk (t)) determining a normal direction _nk (t) (( _xk (t), _yk (t), z _k (t)) with respect to the determined first approximation A ₁ (t); for the smoothed 3D measurement points (x _k (t), y _k (t), z _k (t)) iteratively executing a RANSAC algorithm for determination a polynomial-based approximation A ₂ (t) of the 3D surface, which is determined by a subset (x _f (t), y _f (t), z _f (t)) of the 3D measurement points (x _k ) determined as part of the RANSAC algorithm (t), y _k (t), z _k (t)), with f = 1, 2, ... l and l ≤ m; finding a polynomial-based approximation A ₃ (t) for the subset (x _f (t), y _f (t), z _f (t)) by means of a least-squares method; iteratively reducing the order of the polynomial-based approximation A ₃ (t) to obtain an optimized approximation A ₃ * (t) and for determination an associated optimized subset (x _g (t), y _g (t), z _g (t)) obtained by comparing the respective approximate on A ₃ * (t) with the 3D measurement points (x _k (t), y _k (t), z _k (t)) is determined, with g = 1, 2, ... p and p ≥ l; and repeating the steps beginning with step: determining a polynomial-based approximation A ₃ (t), wherein the approximation A ₃ (t) now substitutes for the subset (x _f (t), y _f (t), z _f (t)) for the optimized subset (x _g (t), y _g (t), z _g (t)) is determined until the number p of the points of the subset (x _g (t), y _g (t), z _g (t)) does not increase any further compared to the previous determination of the subset (x _g (t), y _g (t), z _g (t)), whereby the model M (t) of the 3D surface ( 209 ) of the measurement object through which the thus optimized subset (x _g (t), y _g (t), z _g (t)) 'associated approximation A ₃ *' (t) is given.

The device allows the determination of material characteristics of the pliable object to be measured 210 in the sense of system identification. On the one hand interested in the sagging of the pliable object to be measured 210 during transport with the aid of a robot portal, as well as the determination of the material properties natural vibration frequency, natural vibration mode, modal mass and modal damping. As measuring objects to be examined are basically all limp components in question, such as carbon fiber reinforced plastic (CFRP) -Halbzeuge.

The test object to be examined 210 is how in 1 shown at both ends to the end effectors 202 . 203 of the industrial robot 211 clamped. The robot arms 204 . 205 are used below to stimulate the component with movements. It is assumed that the coordinate relationship of the optical sensor of the sensor system 201 to the robot arms or positions of the end effectors 202 . 203 is known and thus a common coordinate system is defined.

This is advantageously ensured by performing a so-called "hand-eye calibration". Explanations on this are known from the prior art. It is also assumed that the positions of the first 202 or second 203 Fixing device and thus the fixing points of the measurement object 210 are known. This is ensured, for example, with the aid of the optical sensor system 201 can be determined.

The performed hand-eye calibration is the transformation between the robot coordinate system and the coordinate system of the optical sensor of the sensor system 201 known. As a result, for example, a coordinate system can be defined whose origin is in a fixation point of the measurement object 210 and whose x-axis points in the direction of the second fixing point, and whose y-axis lies along, for example, a suspension beam of the looped-in component. This coordinate system will be in the following coordinate system of the DUT 210 called.

The sagging of the test object 210 depends on the distance of the fixation points. The sagging of the test object 210 is now, for example, for different distances A _p between two Roboterendeffektoren 202 . 203 determined by determining the surface model M (A _p ) for each of these distances A _p . If the model is present, the sag can be determined by determining the extreme value of the polynomial (transformed into the component coordinate system). The sag is then the amount of the z-coordinate at the extreme point found.

Note: The sag determined in this way is based on a free-floating fixation of the DUT 210 out. In the case of non-free fixation / restraint, the sag depends on the material's characteristics (stiffness). The method described above nevertheless represents a worst-case situation since it describes the maximum possible sagging.

The determined sag allows, for example, a path planning of a production robot, which ensures that there is no collision of the component (object under test) with an environment due to the sag.

To determine the material characteristics natural vibration frequency, natural vibration mode, modal mass and modal damping, the known method of modal analysis is used. For this the measuring object becomes 210 in the fixed / clamped state excited with a known mechanical movement, for example. A pulse movement, and detects the structural response. For this purpose, the model M (t) of the surface of the measurement object is at different times t 210 certainly. Thus, the resulting (decaying) mechanical vibration over time can be determined.

Classically, these material characteristics are determined by attaching accelerometers and measuring accelerations over time. By the proposed device and the optical sensor system 201 is a purely optical measurement possible without attaching sensors or other structures on the workpiece. For this purpose, the surface structure determined as 3D model M (t) is scanned at computationally determined equal intervals and the acceleration is calculated for each of the scanned points. This gives you a very dense network of "virtual" accelerometers. With the calculated accelerations, the material characteristics can now be determined in the classical way, ie as with accelerometers.

Note: Very high-frequency vibrations can not be measured with an optical sensor system, since the maximum frequency is limited by the refresh rate of the sensor. In the case of a limp component, however, this usually poses no problem, since only relatively low-frequency main vibrations occur here.

2 shows a schematic process flow of a proposed method for determining mechanical properties, such as the natural vibration frequency, the natural vibration mode, the modal mass, and / or the modal damping of a pliable object to be measured 210 with the following steps. In a first step 301 a fixation of the measurement object takes place 210 with a first side on a first fixing device 202 and with its second side opposite the first side to a second fixing device 203 , where the measurement object 210 otherwise free hanging. In a second step 302 takes place via the first fixing device 202 an introduction of mechanical excitations in the measurement object 210 , In a third step 303 a scanning of a 3D surface is done 209 of the measurement object 210 in the fixed state with an optical sensor system 201 to determine a number n the surface 209 representing 3D measurement points (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t. In a fourth step 304 Based on the 3D measurement points (x _i (t), y _i (t), z _i (t)), a time series of models M (t) of the 3D surface is determined 209 of the measurement object 210 , In a fifth step 305 Based on the models M (t) a determination of the mechanical properties of the DUT is carried out 201 by means of a modal analysis.

3 shows a schematic process flow of a proposed method for determining a model M of a 3D surface of a measurement object 210 with the following steps: scanning 101 the 3D surface of the measurement object 210 with an optical sensor system 201 for determining a number n of the 3D-measuring points representing the surface (x _i , y _i , z _i ) with i = 1, 2, ... n; Determine 102 a polynomial-based approximation A _{1 of} the 3D measuring points (x _i , y _i , z _i ) by means of a least squares method; Smooth 103 the 3D measurement points (x _i , y _i , z _i ) for generating smoothed 3D measurement points (x _k , y _k , z _k ), where k = 1, 2, ... m and m ≤ n; for each point of the smoothed 3D measurement points (x _k , y _k , z _k ) 104 a normal direction n _k (x _k , y _k , z _k ) with respect to the determined first approximation A ₁ ; for the smoothed 3D measurement points ( _xk , _yk , _zk ) iterative execution 105 of a RANSAC algorithm for determining a polynomial-based approximation A _{2 of} the 3D surface, which is determined by a subset (x _f , y _f , z _f ) of the 3D measurement points (x _k , y _k , z _k ) determined as part of the RANSAC algorithm ), with f = 1, 2, ... l and l ≤ m; Determine 106 a polynomial-based approximation A ₃ for the subset (x _f , y _f , z _f ) by means of a least squares method; iteratively reducing the order of the polynomial-based approximation A ₃ for detection 107 an optimized approximation A ₃ * and for determining an associated optimized subset (x _g , y _g , z _g ), which is determined by comparing the respective approximation A ₃ * with the 3D measuring points (x _k , y _k , z _k ) with g = 1, 2, ... p and p ≥ l; To repeat 108 the steps starting with step 106 ., Where the approximation A ₃ instead of the subset (x _f , y _f , z _f ) for the optimized subset (x _g , y _g , z _g ) is determined, until the number p of the points of the subset (x _g , y _g , z _g ) does not rise any further compared to the previous determination of the subset (x _g , y _g , z _g ), wherein the model M of the 3D surface of the measurement object is characterized by the subset (x _g , y _g , z _g ) 'associated approximation A ₃ *' is given.

Although the invention has been further illustrated and explained in detail by way of preferred embodiments, the invention is not limited by the disclosed examples, and other variations can be derived therefrom by those skilled in the art without departing from the scope of the invention. It is therefore clear that a multitude of possible variations exists. It is also to be understood that exemplified embodiments are really only examples that are not to be construed in any way as limiting the scope, applicability, or configuration of the invention. Rather, the foregoing description and description of the figures enable one skilled in the art to practice the exemplary embodiments, and those skilled in the art, having the benefit of the disclosed inventive concept, can make various changes, for example, to the function or arrangement of individual elements recited in an exemplary embodiment, without Protected area, which is defined by the claims and their legal equivalents, such as further explanation in the description.

Claims (10)

Device for determining mechanical properties, for example the natural oscillation frequency, the inherent oscillation shape, the modal mass and / or the modal damping, of a limp test object ( 210 ), comprising: - a first fixing device ( 202 ), with which the measuring object ( 210 ) is releasably fixable on a first side, - a second fixing device ( 203 ), with which the measuring object ( 210 ) is fixable fixed releasably on its side opposite the first side, wherein the measurement object ( 210 ) in a fixed state in the first fixing device ( 202 ) and the second fixing device F2 ( 203 ) is fixed and otherwise hangs freely, wherein the first fixing device ( 202 ) with a first actuator ( 204 ) and by the first actuator ( 204 ) via the first fixing device ( 202 ) mechanical excitations in the measurement object ( 210 ), - a control unit ( 206 ) for controlling the first actuator ( 206 ), - an optical sensor system ( 201 ) for scanning a 3D surface ( 209 ) of the test object ( 210 ) in the fixed state and for determining a number n the surface ( 209 ) representing 3D measuring points (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t, - a first evaluation unit ( 207 ), on the basis of the 3D measuring points (x _i (t), y _i (t), z _i (t)) a time series of models M (t) of the 3D surface ( 209 ) of the test object ( 210 ), and - a second evaluation unit ( 208 ), on the basis of the models M (t) the mechanical properties of the test object ( 210 ) are determined by means of a modal analysis. Device according to Claim 1, in which the first evaluation unit ( 207 ) is executed and set up to carry out the following steps: 2.1. Determining a polynomial-based approximation A ₁ (t) of the 3D measuring points (x _{i '} (t), y _i (t), z _i (t)) by means of a least squares method, 2.2. Smoothing the 3D measuring points (x _i (t), y _i (t), z _i (t)) to produce smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)), with k = 1, 2, ... m and m ≤ n, 2.3. for each point of the smoothed 3D measurement points ( _xk (t), _yk (t), _zk (t)) determining a normal direction _nk (t) (( _xk (t), _yk (t), z _k (t)) with respect to the determined first approximation A ₁ (t), 2.4 for the smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)) iteratively executing a RANSAC algorithm for determining a polynomial-based approximation A ₂ (t) of the 3D surface, which is determined by a subset (x _f (t), y _f (t), z _f (t)) of the 3D measuring points determined by the RANSAC algorithm ( x _k (t), y _k (t), z _k (t)), where f = 1, 2, ... l and l ≤ m, 2.5 Determining a polynomial-based approximation A ₃ (t) for the Subset (x _f (t), y _f (t), z _f (t)) by means of a least squares method, 2.6 iteratively reducing the order of the polynomial-based approximation A ₃ (t) to obtain an optimized approximation A ₃ * (t) and for determining an associated optimized subset (x _g (t), y _g (t), z _g (t)) obtained by comparing the respec approximation A ₃ * (t) with the 3D measuring points (x _k (t), y _k (t), z _k (t)) is determined, with g = 1, 2, ... p and p ≥ l , 2.7. Repeating the steps beginning with step 2.5, wherein the approximation A ₃ (t) is now substituted for the subset (x _f (t), y _f (t), z _f (t)) for the optimized subset (x _g (t ), y _g (t), z _g (t)) is determined until the number p of the points of the subset (x _g (t), y _g (t), z _g (t)) compared to the previous determination of Subset (x _g (t), y _g (t), z _g (t)), where the model M (t) of the 3D surface ( 209 ) of the measurement object through which the thus optimized subset (x _g (t), y _g (t), z _g (t)) 'associated approximation A ₃ *' (t) is given. Device according to one of claims 1 to 2, wherein the second fixing device ( 203 ) with a second actuator ( 205 ), and by the second actuator ( 205 ) via the second fixing device ( 203 ) mechanical excitations in the measurement object ( 210 ) are insertable, and wherein the control unit ( 206 ) for controlling the second actuator ( 205 ) is designed and furnished. Device according to one of claims 2 to 3, wherein the first evaluation unit ( 207 ) is designed and set up such that the least squares method in step 2.1. and / or in step 2.5. a moving-least-square method or a weighted-least-squares method is / are. Device according to one of claims 2 to 4, wherein the first evaluation unit is designed and arranged such that the smoothing by projection of each 3D measuring point (x _i (t), y _i (t), z _i (t)) to a local Surface of the approximation A ₁ (t) takes place, which represents a local environment of the respective 3D measuring point (x _i (t), y _i (t), z _i (t)). Device according to one of claims 2 to 5, wherein the first evaluation unit ( 207 ) is designed and set up such that in the RANSAC algorithm for each 3D measuring point (x _k (t), y _k (t), z _k (t)) an Euclidean distance D _k (t) (x _k (t ), _yk (t), _zk (t)) between the measuring point ( _xk (t), _yk (t), _zk (t)) and a corresponding point ( _xk (t), _yk (t), z _k (t)) _{A2 of} the approximation A ₂ (t), and an angular difference W _k (t) (x _k (t), y _k (t), z _k (t)) between the determined normal direction _nk (t) ( _xk (t), _yk (t), _zk (t)) and a normal direction _{nk, A2} (t) ( _xk (t), _yk (t), _zk ( t)) of the corresponding point (x _k (t), y _k (t), z _k (t)) _{A2 of} the approximation A ₂ (t) is determined. Apparatus according to claim 6, wherein the first evaluation unit ( 207 ) is executed and set up such that the subset (x _f (t), y _f (t), z _f (t)) only those 3D measuring points (x _k (t), y _k (t), z _k ( t)) for which: D _k (t) (x _k (t), y _k (t), z _k (t)) <G ₁ , and W _k (t) (x _k (t), y _k (t), z _k (t)) <G ₂ , where G ₁ and G _{2 are} predetermined limits. Device according to one of claims 2 to 7, wherein the first evaluation unit ( 207 ) is designed and arranged such that after step 2.7. segmentation of the optimized subset (x _g (t), y _g (t), z _g (t)) 'takes place from the 3D measurement points (x _k (t), y _k (t), z _k (t)) , Method for determining mechanical properties, for example the natural oscillation frequency, the inherent oscillation form, the modal mass and / or the modal damping, of a limp test object ( 210 ), with the following steps: 9.1. Fix ( 301 ) of the measurement object ( 210 ) with a first side on a first fixing device ( 202 ) and with its first side opposite side to a second fixing device ( 203 ), wherein the measurement object ( 210 ) otherwise free hanging, 9.2. via the first fixing device ( 202 ) Introducing ( 302 ) mechanical excitations into the measurement object ( 210 9.3. Scanning ( 303 ) of a 3D surface ( 209 ) of the test object ( 210 ) in the fixed state with an optical sensor system ( 201 ) for investigation ( 304 ) a number n the surface ( 209 ) representing 3D measurement points (x _i (t), y _i (t), z _i (t)) with i = 1, 2, ... n, and time t, 9.4. on the basis of the 3D measuring points (x _i (t), y _i (t), z _i (t)) 304 ) a time series of models M (t) of the 3D surface ( 209 ) of the test object ( 210 ), and 9.5. on the basis of models M (t) 305 ) of the mechanical properties of the test object ( 201 ) by means of a modal analysis. The method of claim 9, wherein determining the time series of models M (t) of the 3D surface of the measurement object comprises the steps of: 10.1. Determining a polynomial-based approximation A ₁ (t) of the 3D measuring points (x _{i '} (t), y _i (t), z _i (t)) by means of a least squares method, 10.2. Smoothing the 3D measuring points (x _i (t), y _i (t), z _i (t)) to produce smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)), with k = 1, 2, ... m and m ≤ n, 10.3. for each point of the smoothed 3D measurement points ( _xk (t), _yk (t), _zk (t)) determining a normal direction _nk (t) (( _xk (t), _yk (t), z _k (t)) with respect to the determined first approximation A ₁ (t), 10.4 for the smoothed 3D measuring points (x _k (t), y _k (t), z _k (t)) iteratively executing a RANSAC algorithm for determining a polynomial-based approximation A ₂ (t) of the 3D surface, which is determined by a subset (x _f (t), y _f (t), z _f (t)) of the 3D measuring points determined by the RANSAC algorithm ( x _k (t), y _k (t), z _k (t)), with f = 1, 2, ... l and l ≤ m, 10.5 Determining a polynomial-based approximation A ₃ (t) for the Subset (x _f (t), y _f (t), z _f (t)) using a least-squares method, 10.6 iteratively reducing the order of the polynomial-based approximation A ₃ (t) to obtain an optimized approximation A ₃ * (t) and for determining an associated optimized subset (x _g (t), y _g (t), z _g (t)) obtained by comparing the jewe iligen approximation A ₃ * (t) with the 3D measuring points (x _k (t), y _k (t), z _k (t)) is determined, with g = 1, 2, ... p and p ≥ l , 10.7. Repeating the steps beginning with step 10.5, wherein the approximation A ₃ (t) is now substituted for the subset (x _f (t), y _f (t), z _f (t)) for the optimized subset (x _g (t ), y _g (t), z _g (t)) is determined until the number p of the points of the subset (x _g (t), y _g (t), z _g (t)) compared to the previous determination of Subset (x _g (t), y _g (t), z _g (t)) is no longer increasing, the model M (t) of the 3D surface of the measurement object being characterized by the thus optimized subset (x _g (t), y _g (t), z _g (t)) 'associated approximation A ₃ *' (t) is given.

Images