DE102018124569A1 - Control method for an additive manufacturing device - Google Patents

Abstract

In order to be able to influence process-induced mechanical properties of an additively manufactured component (20) during production, a control method is proposed in which component layer data (32) of a component layer (26) of the component is non-destructively by means of a component sensor device (30), for example a CT device manufacturing component (20) are detected. The expected mechanical component properties (40) are determined in real time during production from the component layer data (26) and from replacement models for material (42) and component (44). Due to the deviation of the expected component properties (40) from the specified component properties (48), the manufacturing process parameters are changed so that the deviation decreases. Processing in real time is made possible in particular by prepared replacement models (42, 44) trained by machine learning. The replacement models (42, 44) and the additive manufacturing device (10) are each based on artificial neural networks (50) and can be trained further during the manufacture of the component (20) in order to avoid defects that affect the mechanical properties of the component (20 ) can affect.

Description

The invention relates to a control method for controlling an additive manufacturing device designed for layer-by-layer additive manufacturing of a component. The invention further relates to a material replacement model determination method and a component replacement model determination method. The invention further relates to an additive manufacturing device and a control device therefor.

Additive manufacturing, sometimes referred to casually as 3D printing, is one way of producing components at the push of a button. Initially mainly used for rapid prototyping, additive manufacturing is increasingly finding its way into industrial series production. One advantage of this is the more gentle way of working compared to classic subtractive manufacturing processes such as turning, milling, grinding, etc.

Due to the specific properties of additive manufacturing processes, sometimes significantly more complex components can be manufactured more easily. So far, the components to be manufactured have been designed, printed out and then subjected to non-destructive testing. It turned out that the component properties of two identically constructed components can sometimes differ significantly from one another if they were printed out on different machines or under slightly different environmental conditions.

In the aviation sector in particular, finished components are subjected to stringent tests. As a result, components are classified as not permitted in case of doubt. For additively manufactured components, this means that the entire machine runtime and the material used can be wasted.

In addition, the reduction in the overall mechanical properties of the additively manufactured component due to process inaccuracies has so far not been able to be determined in detail during manufacture. This makes it difficult to assess manufacturing deviations or errors more precisely. As a result, there is a two-time iteration between design and manufacturing, as well as a final component that may not meet the structural requirements. Furthermore, it has not previously been provided to control the manufacturing system to meet the mechanical demands on the component.

So far, neither a methodology is known for determining the process-induced mechanical properties of the component during additive manufacturing, nor for controlling the manufacturing system against the required structural properties and effective process deviations.

The invention has for its object to improve additive manufacturing processes.

The object is solved by the subject matter of the independent claims. Preferred developments are the subject of the dependent claims.

1. a) mittels einer Bauteilsensoreinrichtung: zerstörungsfreies Erfassen einer Bauteilschicht des zu fertigenden Bauteils zwecks Erhalt von Bauteilschichtdaten oder des Bauteils zwecks Erhalt von Bauteildaten, wobei die Bauteildaten und Bauteilschichtdaten jeweils indikativ für Bauteileigenschaften des Bauteils sind;
2. b) mittels einer Erwartungsbauteileigenschaftenermittlungseinrichtung: Ermitteln von Erwartungsbauteileigenschaften aus den Bauteilschichtdaten/Bauteildaten, aus wenigstens einem Materialersatzmodell und aus wenigstens einem Bauteilersatzmodell;
3. c) in Abhängigkeit von den Erwartungsbauteileigenschaften mittels einer Steuereinrichtung:
   • - Steuern der additiven Fertigungsvorrichtung, einen additiven Fertigungsprozess mit unveränderten Fertigungsprozessparametern fortzusetzen; oder
   • - Verändern der Fertigungsprozessparameter basierend auf den Erwartungsbauteileigenschaften und Vorgabebauteileigenschaften zwecks Annähern der Erwartungsbauteileigenschaften an die Vorgabebauteileigenschaften und Steuern der additiven Fertigungsvorrichtung, den Fertigungsprozess mit den veränderten Fertigungsprozessparametern fortzusetzen.

The invention provides a control method for controlling an additive manufacturing device designed for layer-by-layer additive manufacturing of a component, comprising the steps:

1. a) by means of a component sensor device: non-destructive detection of a component layer of the component to be manufactured for the purpose of obtaining component layer data or the component for the purpose of obtaining component data, the component data and component layer data being indicative of component properties of the component;
2. b) by means of an expected component property determination device: determining expected component properties from the component layer data / component data, from at least one material replacement model and from at least one component replacement model;
3. c) depending on the expected component properties by means of a control device:
   • - Controlling the additive manufacturing device to continue an additive manufacturing process with unchanged manufacturing process parameters; or
   • - Changing the manufacturing process parameters based on the expected component properties and default component properties in order to approximate the expected component properties to the default component properties and controlling the additive manufacturing device to continue the manufacturing process with the changed manufacturing process parameters.

In order to be able to influence process-induced mechanical properties of an additively manufactured component during production, a control method is proposed in which component layer data of a component layer of the component to be manufactured are recorded non-destructively by means of a component sensor device, for example a CT device. The expected mechanical component properties are determined during production from the component layer data and from replacement models for material and component, preferably in real time. Due to the deviation of the expected component properties from the specified component properties, the manufacturing process parameters are changed so that the deviation decreases. The Real-time processing is made possible in particular by prepared replacement models trained by machine learning. The replacement models and the additive manufacturing device are each based on artificial neural networks. The artificial neural networks can be trained further during the production of the component in order to avoid defects which can impair the mechanical properties of the component or the component behavior. The control process can be used to adapt the manufacturing process in such a way that the component to be manufactured or manufactured has the desired properties, for example with regard to rigidity, breaking strength, damage tolerance, component fatigue, etc. The expected component properties are determined from the component data or component layer data and the replacement models for material and component.

It is preferred that the component sensor device is selected from a group comprising an imaging device, a camera, a VIS camera, an IR camera, an X-ray camera, a CT device, an ultrasound imaging device, a thermal imaging device, a thermometer device , an eddy current measuring device and a vibration analysis device. Different sensor devices allow different component (layer) defects to be detected. Cameras for visible light (VIS), infrared (IR) or X-rays (X-Ray) in particular enable the measurement of the dimensions of any material defects, so-called voids. Acquisition by means of tomography techniques or ultrasound is also possible. Temperature measurements, eddy current measurements or vibration measurements are also conceivable. Depending on the building material and the desired information, one or the other method may be preferred.

It is preferred that in step b) the material replacement model is determined by a material replacement model determination method described here. It is preferred that in step b) the component replacement model is determined by a component replacement model determination method described here. Substitute models allow the control process to be carried out much more quickly in real time. The special methods described here for determining the replacement models have proven to be particularly advantageous.

It is preferred that in step c) the last component layer produced is revised with the changed manufacturing process parameters. If a sufficient deviation between the component to be created or its expected component properties is determined, the last component layer can be corrected in this way.

It is preferred that in step c) the changed manufacturing process parameters are determined by an artificial neural network. This methodology allows continuous adaptation to changed data and parameters. Furthermore, faster processing is also possible than with conventional methods of determination.

It is preferred that the artificial neural network is trained with manufacturing process parameters acquired during the manufacturing process and component behavior determined during the manufacturing process. The control process can thus be better and better adapted to new and previously unknown components.

It is preferred that the component behavior is determined using a preferred component replacement model determination method.

1. a) Bereitstellen einer Mehrzahl von Materialeinheitszellen mit unterschiedlichen Konfigurationen von Materialdefekten in der Materialeinheitszelle;
2. b) Ermitteln von Materialersatzparametern für jede Materialeinheitszelle aus den Materialparametern und unterschiedlichen Lastfällen;
3. c) Zuordnen jeder in Schritt a) bereitgestellten Materialeinheitszelle zu den in Schritt b) ermittelten Materialersatzparametern;
4. d) Trainieren des Materialersatzmodells, insbesondere durch maschinelles Lernen, mit den Materialeinheitszellen und den zugehörigen Materialersatzparametern, so dass das Materialersatzmodell indikativ ist für das Materialverhalten des zu modellierenden Materials bei unterschiedlichen Materialdefektdichten und unterschiedlichen Materialdefektgrößen.

The invention provides a material replacement model determination method for determining a material replacement model from material parameters of the material to be modeled and material defects, in particular material defects and / or material cracks, for the purpose of using the material replacement model in a component replacement model determination method, with the steps:

1. a) providing a plurality of material unit cells with different configurations of material defects in the material unit cell;
2. b) determining material replacement parameters for each material unit cell from the material parameters and different load cases;
3. c) assigning each material unit cell provided in step a) to the material replacement parameters determined in step b);
4. d) Training the material replacement model, in particular by machine learning, with the material unit cells and the associated material replacement parameters, so that the material replacement model is indicative of the material behavior of the material to be modeled with different material defect densities and different material defect sizes.

By forming a material replacement model, the material behavior can be determined faster compared to a simulation. Several or a large number of material replacement models can be trained without the need for production. The material replacement models determined in this way can then simply be called up in the control and allow processing in real time. The material replacement model is not limited to a specific structure but serves as a general model for using the additive Manufacturing process manufacturable materials. In other words, the material replacement model indicates which material properties a material has with a certain material defect density and size. Material defects are in particular the porosity and / or material defects and / or material cracks. For this purpose, different load cases are simulated for, for example, cube-shaped material unit cells that have different configurations of material defects. The deformation and other changes to the material unit cell can be determined. The change in the material cracks, for example their propagation, can also be determined. The material replacement parameters are then determined from these changes. A material replacement model can thus be formed which, for example, immediately outputs the corresponding material behavior upon input of a load case.

It is preferred that in step a) the provision is made by computer-generated data. This enables material replacement models to be pre-trained without having to run production.

It is preferred that in step a) the provision is made by component layer data acquired during an additive manufacturing process. This allows the material replacement model to be better adapted to actual manufacturing processes or materials.

It is preferred that in step b) each material replacement parameter is selected from a group that contains material rigidity, material thickness, material fatigue, material elongation at break, material elasticity module, material compression module, material thrust module. The material replacement parameters can also be referred to as effective material parameters. The material replacement parameters can be indicative of the material behavior of the model material.

1. a) Bereitstellen von Bauteildaten, die indikativ sind für Bauteilabweichungen von dem Bauteilmodell, und/oder Bauteilschichtdaten, die indikativ sind für Bauteilschichtabweichungen von einer Bauteilmodellschicht des Bauteilmodells;
2. b) Ermitteln eines zu einem Bauteilbereich passenden Materialersatzmodells für jeden Bauteilbereich und Zuordnen des Materialersatzmodells zu diesem Bauteilbereich zwecks erhalt einer Bauteilersatzstruktur;
3. c) Ermitteln einer Bauteilstrukturreaktion für unterschiedliche Lastfälle anhand der Bauteilersatzstruktur;
4. d) Zuordnen der in Schritt a) bereitgestellten Bauteildaten zu jeder in Schritt c) ermittelten Bauteilstrukturreaktion;
5. e) Trainieren des Bauteilersatzmodells, insbesondere durch maschinelles Lernen, mit den Bauteildaten und den zugehörigen Bauteilstrukturreaktionen, so dass das Bauteilersatzmodell indikativ ist für das Bauteilverhalten, insbesondere Bauteilfestigkeitskriterien/Bauteilversagenskriterien, wie beispielsweise die Schadenstoleranz und/oder Bauteilermüdung, des zu fertigenden Bauteils.

The invention provides a component replacement model determination method for determining a component replacement model from at least one material replacement model and a component model of a component to be manufactured for the purpose of using the component replacement model in a control method for controlling an additive manufacturing device designed for layer-by-layer additive manufacturing of a component, with the steps:

1. a) providing component data which are indicative of component deviations from the component model and / or component layer data which are indicative of component layer deviations from a component model layer of the component model;
2. b) determining a material replacement model suitable for a component area for each component area and assigning the material replacement model to this component area in order to obtain a component replacement structure;
3. c) determining a component structure reaction for different load cases on the basis of the component replacement structure;
4. d) assigning the component data provided in step a) to each component structure reaction determined in step c);
5. e) Training the component replacement model, in particular by machine learning, with the component data and the associated component structure reactions, so that the component replacement model is indicative of the component behavior, in particular component strength criteria / component failure criteria, such as the damage tolerance and / or component fatigue, of the component to be manufactured.

By forming a component replacement model, the component behavior can be determined faster compared to a simulation. Component replacement models can be trained without the need for manufacturing. The component replacement models determined in this way can then simply be called up in the control and allow processing in real time. The component data contains information about where defects exist in the component compared to the specified component. The same applies to the component layers. The material defects can therefore be determined from the component data. Certain component areas that have substantially similar material defects can be assigned a material replacement model that corresponds to the material defects. In other words, the material replacement model takes the place of the actual material defects in this component area. A component replacement structure can thus be formed, which essentially comprises component regions of different material replacement parameters. Different load cases are applied to the component replacement structure, and the corresponding mechanical component structure reaction can be determined directly from the component replacement structure and the load cases. A component replacement model can thus be formed which, for example, immediately upon input of a load case, outputs the corresponding component behavior, in particular component strength criteria / component failure criteria, such as the damage tolerance and / or component fatigue.

It is preferred that in step a) the provision is made by computer-generated data. This enables component replacement models to be pre-trained without the need for production to run.

It is preferred that in step a) the provision is made by component layer data or component data acquired during an additive manufacturing process. This allows the component replacement model be better adapted to actual manufacturing processes or materials.

The invention provides a control device for controlling an additive manufacturing device designed for layer-by-layer additive manufacturing of a component, the control device being designed for carrying out one of the preferred methods described herein, the control device comprising a component sensor device which can be attached to the additive manufacturing device and which is used for the non-destructive detection of a component layer of the component to be manufactured for the purpose of obtaining component layer data or the component that is indicative of component properties of the component, an expected component property determination device that is designed to determine expected component properties from the component layer data, from a material replacement model and from a component replacement model, and a device that can be connected to the additive manufacturing device Includes control device, which is designed, depending on the expected component properties, the ad Control the additive manufacturing device such that the additive manufacturing device continues an additive manufacturing process with unchanged manufacturing process parameters or that the additive manufacturing device changes the manufacturing process parameters based on the expected component properties and default component properties to approximate the expected component properties to the default component properties, and the additive manufacturing device, the manufacturing process with the changed Manufacturing process parameters continues.

It is preferred that the component sensor device is selected from a group comprising an imaging device, a camera, a VIS camera, an IR camera, an X-ray camera, a CT device, an ultrasound imaging device, a thermal imaging device, a thermometer device , an eddy current measuring device and a vibration analysis device.

The invention provides an additive manufacturing device which is designed for layer-by-layer additive manufacturing of a component and which is designed to carry out a preferred method described herein and / or which comprises a preferred control device.

Different non-destructive investigation (NDI) methods have been developed to examine additively manufactured components. Machine learning and approximation methods for knowledge-based systems are also known in principle. The ideas presented here assess the NDI-based additive manufacturing deviations with regard to the mechanical properties of the component and compare them with the requirements for the finished component. This enables a quick, knowledge-based assessment of the effects of defects on component quality. This can also increase the robustness of the manufacturing process.

It should be noted that the invention is explained by way of example using a device for selective laser sintering, but other additive manufacturing devices, in particular FDM printers and selective laser fusion printers, are also suitable for the application of the invention.

• 1 ein Ausführungsbeispiel einer additiven Fertigungsvorrichtung;
• 2 ein Ausführungsbeispiel eines Steuerungsverfahrens;
• 3 ein Ausführungsbeispiel eines Materialersatzmodellermittlungsverfahrens; und
• 4 ein Ausführungsbeispiel eines Bauteilersatzmodellermittlungsverfahrens.

Exemplary embodiments are explained in more detail below with the aid of the schematic drawings. It shows:

• 1 an embodiment of an additive manufacturing device;
• 2nd an embodiment of a control method;
• 3rd an embodiment of a material replacement model determination method; and
• 4th an embodiment of a component replacement model determination method.

It will start on 1 Reference which is an additive manufacturing device 10th shows. The additive manufacturing device 10th is, for example, a selective laser sintering device 12th . The additive manufacturing device 10th comprises a metal powder bed device 14 , a laser sintering device 16 and a metal powder coater 18th . In the metal powder bed facility 14 becomes a component 20th sintered. The laser sintering device 16 is movably mounted and sends a laser beam 22 out, the metal powder 24th warmed and thus sintered. After a component layer 26 has been finished sintered by the metal powder coating device 18th a new layer of metal powder 24th upset.

The additive manufacturing device further comprises 10th a control device 28 which controls the manufacturing process. The control device comprises a component sensor device 30th . The component sensor device 30th can at least the top of the component layers 26 capture to component layer data 32 to obtain. The component sensor device 30th can for example a camera 34 , especially for X-rays. The component sensor device 30th can also have other sensor devices.

The control device 28 further comprises an expected component property determination device 38 , which is formed from the component layer data 32 Expected component properties 40 to determine which are indicative of the expected Component properties. The expected component property determination device 38 accesses pre-trained material replacement models 42 and component replacement models 44 back.

The control device 36 further comprises a control device 46 , which is designed to control the manufacturing process or the manufacturing process parameters depending on the determined expected component properties 40 , more precisely the deviation of the expected component properties 40 of default component properties 48 to change to reduce the deviation. The control device 46 can be an appropriately trained artificial neural network 50 have the control. The artificial neural network 50 can be trained in advance with manufacturing process parameters and corresponding component behavior.

Below is on 2nd Reference is made to an example of a control method for the additive manufacturing device 10th shows. In one production step 100 becomes a component 20th by sintering component layers lying on top of each other 26 generated.

In one acquisition step 102 detects the component sensor device 30th Component layer data 32 , which in particular information about the porosity, defect density and / or size and the dimensions of the component layer 26 contain. Furthermore, the manufacturing process parameters in the acquisition step 102 detected.

In an expected component property determination step 104 become from the component layer data just acquired 26 and all component layer data previously recorded during production 26 as well as material replacement models 42 and component replacement models 44 the expected component properties 40 determined. The expected component properties 40 indicate the expected component properties, which the component based on the recorded component layer data 26 expected to have.

In a validation step 106 become the expected component properties 40 with the default component properties 48 compared. If the deviation of the expected component properties 40 from the default component properties 48 is within a specified tolerance range in a continuation step 108 continued (branch Y ); otherwise (branch N ) is in an interrupt step 110 continued.

In the continuation step 108 the manufacturing process parameters for the next shift are set and otherwise left unchanged. Then in a control step 112 continued.

In the interrupt step 110 the manufacturing process is interrupted, with a subsequent change step 114 the manufacturing process parameters are changed such that the deviation of the expected component properties 40 from the default component properties 48 is within the tolerance range. How to change the manufacturing process parameters for this is done, for example, by the artificial neural network 50 determined. Then in the control step 112 continued.

In the control step 112 becomes the additive manufacturing device 10th on the manufacturing step 100 prepared and the changed manufacturing process parameters transmitted to them. Furthermore, the control data for producing the next component layer 26 to the additive manufacturing device 10th transmitted. Then in the manufacturing step 100 continued.

The process runs until the component 20th is finished.

Furthermore, the control method can be a learning step 116 include. In the learning step 116 can the manufacturing process parameters and the expected component properties 40 processed and for training the artificial neural network 50 be used. This changes the manufacturing process parameters in terms of approximation to the default component properties 48 visibly better.

Below is on 3rd Reference, which shows an example of a material replacement model determination process.

In one step 200 become component data or component layer data 32 by the component sensor device 30th were entered. The component data / component layer data 32 provide information about material defects, in particular porosity and / or defect density or size.

In one go to the input step 200 subsequent modeling step 202 is first a plurality of material unit cells 52 , for example in the form of a cube, each with different configurations of material defects 54 exhibit. The material unit cells 52 represent a piece of a predefined material that corresponds to the manufacturing material, for example the metal powder 24th .

In one stress step 204 are different load cases 56 on each of the material unit cells 52 applied and a corresponding loaded material unit cell 58 determined. In particular material rigidity 60 and material thickness 62 (e.g. until breakage) are used as material replacement parameters 64 determined.

In a training data step 206 each material unit cell 52 the corresponding material replacement parameters 64 assigned and as material training data 66 saved.

In one training step 208 becomes an artificial neural network, material network 68 with the material training data 66 trained and so a material replacement model 70 formed, which is indicative of the material behavior of the material to be modeled with different material defect densities and different material defect sizes.

In other words, any load case is applied to the material replacement model 70 the material replacement model 70 immediately the material behavior back.

The material replacement model determination process can be carried out in parallel to other processes, separately or completely independently to prepare a manufacturing process.

It will follow on 4th Reference, which shows an example of a component replacement model determination method.

In one step 300 become component data or component layer data 32 by the component sensor device 30th were entered. The component data / component layer data 32 provide information in particular about component defects 72 , in particular component porosity and / or component defect density or size. From this, component deviations from the component model are determined.

In one step 302 are based on the component defects 72 or component deviations component areas 74 determined and the respective component area 72 one of the component defects in it 72 corresponding material replacement model 70 assigned. From the material step 302 there is a component replacement structure 76 for further processing.

In one stress step 304 different component load cases 78 on the component replacement structure 76 applied. The component load cases 78 are from the respective material replacement model 70 processed immediately and the corresponding material behavior returned. This then becomes the component structure reaction 80 determined that is indicative of the component behavior of the component 20th under a certain component load case 78 is. In particular, a structural collapse or catastrophic failure occurs as a component structure reaction 80 of the component 20th simulated.

In a training data step 306 are the component data or component layer data 26 the corresponding component structure reaction 80 assigned and as component training data 82 saved.

In one training step 308 becomes an artificial neural network, component network 84 , with the component training data 82 trained and so a component replacement model 86 formed, which is indicative of the component behavior of the component to be manufactured 20th .

In other words, any load case is applied to the component replacement model 86 created, so the component replacement model 86 the component behavior immediately.

The component replacement model determination process can be carried out in parallel to other processes, separately or completely independently in preparation for a manufacturing process.

Reference list

10th

additive manufacturing device

12th

selective laser sintering device

14

Metal powder bed device

16

Laser sintering device

18th

Metal powder coating device

20th

Component

22

laser beam

24th

Metal powder

26

Component layer

28

Control device

30th

Component sensor device

32

Component layer data

34

camera

38

Expectation component property determination device

40

Expected component properties

42

Material replacement models

44

Component replacement models

46

Control device

48

Default component properties

50

artificial neural network

52

Material unit cell

54

Material defect

56

Load case

58

loaded material unit cell

60

Material rigidity

62

Material thickness

64

Material replacement parameters

66

Material training data

68

Material network

70

Material replacement model

72

Component defects

74

Component areas

76

Component replacement structure

78

Component load case

80

Component structure reaction

82

Component training data

84

Component network

86

Component replacement model

100

Manufacturing step

102

Acquisition step

104

Expectation component property determination step

106

Validation step

108

Continuation step

110

Interrupt step

112

Control step

114

Change step

116

Learning step

200

Input step

202

Modeling step

204

Load step

206

Training data step

208

Training step

300

Input step

302

Material step

304

Load step

306

Training data step

308

Training step

Claims (15)

Control method for controlling an additive manufacturing device (10) designed for layer-by-layer additive manufacturing of a component (20), comprising the steps: a) by means of a component sensor device (30): non-destructive detection of a component layer (26) of the component to be manufactured (20) for the purpose of obtaining component layer data (32) or the component (20) for the purpose of receiving component data, the component data and component layer data (32) in each case are indicative of component properties of the component (20); b) by means of an expected component property determination device (38): determining expected component properties (40) from the component layer data (26) / component data, from at least one material replacement model (42, 70) and from at least one component replacement model (44, 86); c) as a function of the expected component properties (40) by means of a control device (46): - Controlling the additive manufacturing device (10) to continue an additive manufacturing process with unchanged manufacturing process parameters; or - Changing the manufacturing process parameters based on the expected component properties (40) and default component properties (48) in order to approximate the expected component properties (40) to the default component properties (48) and controlling the additive manufacturing device (10) to continue the manufacturing process with the changed manufacturing process parameters. Control method according to one of the preceding claims, characterized in that the component sensor device (30) is selected from a group comprising an imaging device, a camera (34), a VIS camera, an IR camera, an X-ray camera, a CT Device, an ultrasound imaging device, a thermal imaging device, a thermometer device, an eddy current measuring device and a vibration analysis device. Control method according to one of the preceding claims, characterized in that in step b) the material replacement model (70) by a method according to one of the Claims 8 to 10th and / or the component replacement model (86) by a method according to one of the Claims 11 and 12th is determined. Control method according to one of the preceding claims, characterized in that in step c) the last component layer (26) produced is revised with the changed manufacturing process parameters. Control method according to one of the preceding claims, characterized in that in step c) the changed manufacturing process parameters are determined by an artificial neural network (50). Control procedure after Claim 5 , characterized in that the artificial neural network (50) is trained with production process parameters recorded during the production process and expected component properties (40) and / or component behavior determined during the production process. Control procedure after Claim 6 , characterized in that the component behavior by means of a component replacement model determination method according to one of the Claims 11 or 12th is determined. Material replacement model determination method for determining a material replacement model (70) from material parameters of the material to be modeled and material defects (54), in particular material defects and / or material cracks, for the purpose of using the material replacement model (70) in a component replacement model determination method, with the steps: a) providing a plurality of material unit cells (52) with different configurations of material defects (54) in the material unit cell (52); b) determining material replacement parameters (64) for each material unit cell (52) from the material parameters and different load cases (56); c) assigning each material unit cell (52) provided in step a) to the material replacement parameters (64) determined in step b); d) Training the material replacement model (70), in particular by machine learning, with the material unit cells (52) and the associated material replacement parameters (64), so that the material replacement model (70) is indicative of the material behavior of the material to be modeled with different material defect densities and different material defect sizes . Material replacement model determination procedure according to Claim 8 , characterized in that in step a) the provision is made by computer-generated data and / or by component layer data or component data recorded during an additive manufacturing process. Material replacement model determination method according to one of the Claims 8 or 9 , characterized in that in step b) each material replacement parameter (64) is selected from a group that contains material stiffness (60), material thickness (62), material fatigue, material fracture elongation, material elasticity module, material compression module, material thrust module. Component replacement model determination method for determining a component replacement model (86) from at least one material replacement model (42, 70) and a component model of a component to be manufactured for the purpose of using the component replacement model (86) in a control method for controlling an additive manufacturing device (10) designed for layer-by-layer additive manufacturing of a component, with the steps: a) providing component data which are indicative of component deviations from the component model and / or component layer data (32) which are indicative of component layer deviations from a component model layer of the component model; b) determining a material replacement model (70) suitable for a component area (74) for each component area (74) and assigning the material replacement model (70) to this component area (74) in order to obtain a component replacement structure (76); c) determining a component structure reaction (80) for different load cases (78) using the component replacement structure (76); d) assigning the component data / component layer data (32) provided in step a) to each component structure reaction (80) determined in step c); e) Training the component replacement model (86), in particular by machine learning, with the component data / component layer data (32) and the associated component structure reactions (80), so that the component replacement model (86) is indicative of the component behavior of the component to be manufactured (20). Component replacement model determination procedure according to Claim 11 , characterized in that in step a) the provision is made by computer-generated data and / or by component layer data (32) or component data acquired during an additive manufacturing process. Control device (28) for controlling an additive manufacturing device (10) designed for layer-by-layer additive manufacturing of a component (20), the control device (28) for performing one of the methods according to one of the Claims 1 to 12th The control device (28) is a component sensor device (30) which can be attached to the additive manufacturing device (10) and which is used for the non-destructive detection of a component layer (26) of the component to be manufactured (20) in order to obtain component layer data (32) indicative of Component properties of the component are designed, an expected component properties determination device (38), which is designed to determine expected component properties (40) from the component layer data (32), from a material replacement model (42, 70) and from a component replacement model (44, 86) and one comprises a control device (46) which can be connected to the additive manufacturing device (10) and which is designed to control the additive manufacturing device (10) as a function of the expected component properties (40) such that the additive manufacturing device (10) continues an additive manufacturing process with unchanged manufacturing process parameters or that di e additive Manufacturing device (10) changes the manufacturing process parameters based on the expected component properties (40) and default component properties (48) to approximate the expected component properties (40) to the default component properties (48), and the additive manufacturing device (10) continues the manufacturing process with the changed manufacturing process parameters. Control device (28) after Claim 13 , characterized in that the component sensor device (30) is selected from a group comprising an imaging device, a camera (34), a VIS camera, an IR camera, an X-ray camera, a CT device, an ultrasound image device, includes a thermal imaging device, a thermometer device, an eddy current measuring device and a vibration analysis device. Additive manufacturing device (10) which is designed for layer-by-layer additive manufacturing of a component (20) and which is designed, a method according to one of the Claims 1 to 12th perform, and / or the one control device (28) according to one of the Claims 13 or 14 includes.

Images