Search

EP-4478232-B1 - APPARATUS AND METHOD FOR RESTRICTING 3D PRINTING

EP4478232B1EP 4478232 B1EP4478232 B1EP 4478232B1EP-4478232-B1

Inventors

  • Solass, Johannes
  • HOSCHKE, Klaus

Dates

Publication Date
20260513
Application Date
20230615

Claims (12)

  1. An apparatus (100) for printing three-dimensional, 3D, objects by a 3D printer, the apparatus (100) comprising: interface circuitry (102) configured to receive a G-code file (130) comprising instructions for 3D printing of one or more 3D objects (140), wherein the G-code file (130) specifies coordinates (132-1; 132-2); and processing circuitry (104) configured to classify a 3D object (140) represented by the G-code file (130) based on spatial dimensions of a geometric feature (142-1; 142-2) of the 3D object (140) corresponding to the coordinates (132-1; 132-2), and restrict 3D printing of the 3D object (140) if the 3D object (140) is classified as at least a part of a dangerous/illegal object, wherein the processing circuitry (104) comprises a trained machine learning network (106) having an input for the G-code file (130) and an output for a classification signal (134-1; 134-2) that is indicative of whether or not the 3D object (140) represented by the G-code file (130) is dangerous/illegal, the apparatus characterized in that the trained machine learning network (106) is based on an attention-based convolutional neural network.
  2. The apparatus (100) of claim 1, wherein the G-code file (130) specifies a path of a tool of the 3D printer, and wherein the processing circuitry (104) is configured to classify the 3D object (140) based on spatial dimensions corresponding to the path.
  3. The apparatus (100) of claim 1 or claim 2, wherein the trained machine learning network (106) is configured to map the G-code file (130) to an object type (136-1; 136-2; 136-3) of dangerous/illegal 3D objects if it is determined that the G-code file (130) corresponds to a dangerous/illegal 3D object.
  4. The apparatus (100) of claim 3, wherein the object type (136-1; 136-2; 136-3) corresponds to at least one of a barrel, frame, receiver, slide, magazine, hammer, handle, cartridge, trigger, spring, connecting rod, and pin of a gun.
  5. The apparatus (100) of any one of the previous claims, wherein the trained machine learning network (106) is configured to further update model parameters based on incremental training data including one or more new G-code files corresponding to dangerous/illegal objects not included in previous training data.
  6. The apparatus (100) of any one of the previous claims, wherein the interface circuitry (102) is configured to receive a request for 3D printing of the 3D object (140) according to the G-code file (130), and wherein restricting 3D printing of the 3D object (140) by the processing circuitry (104) includes at least one of: requiring 3D printing to be done through a user profile; sending and/or storing a notification of the request and/or instance of 3D printing of the 3D object (140) associated with the user profile; performing 3D printing of a modified version of the 3D object (140), and blocking the request for 3D printing of the 3D object (140).
  7. A method (1000) for printing three-dimensional, 3D, objects by a 3D printer using a trained machine learning network (106) based on an attention-based convolutional neural network, the method (1000) comprising: receiving (1010) a G-code file (130) comprising instructions for 3D printing of one or more 3D objects (140), wherein the G-code file (130) specifies coordinates (132-1; 132-2); classifying (1020) a 3D object (140) represented by the G-code file (130) based on spatial dimensions of a geometric feature (142-1; 142-2) of the 3D object (140) corresponding to the coordinates (132-1; 132-2), outputting a classification signal (134-1; 134-2) that is indicative of whether or not the 3D object (140) represented by the G-code file (130) is dangerous/illegal, and restricting (1030) 3D printing of the 3D object (140) if the 3D object (140) is classified as at least a part of a dangerous/illegal object.
  8. The method (1000) of claim 7, wherein classifying (1020) comprises using a machine learning network (106) to determine whether or not the 3D object (140) represented by the G-code file (130) is dangerous/illegal and to map the G-code file (130) to an object type of dangerous/illegal objects if it is determined that the G-code file (130) corresponds to a dangerous/illegal 3D object.
  9. The method (1000) of claim 8, further comprising training the machine learning network (106) based on ground truth information including labeled G-code files (160-1) representing dangerous objects and labeled G-code files (160-2) representing non-dangerous objects.
  10. The method (1000) of claim 9, wherein the training of the machine learning network (106) is based on ground truth information including labeled G-code files (162-1) and labeled 3D models (162-2), wherein a labeled G-code file (162-1) is indicated to correspond to a labeled 3D model, and wherein a group of coordinates of the labeled G-code file (162-1) is indicated to correspond to a geometric feature of the 3D model (162-2).
  11. The method (1000) of claim 9 or claim 10, wherein the training of the machine learning network (106) is based on ground truth information including labeled G-code files (164-1; 164-2) indicating to be within or outside of a predefined degree of similarity of a G-code file determined to correspond to a dangerous/illegal 3D object (160-1).
  12. The method (1000) of any one of claims 9 to 11, wherein the ground truth information includes authentic and/or synthetic G-code files.

Description

Field The present disclosure relates to administering 3D printing. In particular, examples of the present disclosure relate to an apparatus and method for receiving a G-code file, classifying the 3D object represented by the G-code file, and restricting the 3D printing based on the classification. Background US 2022/0011743 A1 describes systems and methods for 3D printer management that can allow or reject printing of an object based on a model that is trained with machine learning. The model can classify according to object type and can be compared against a list to determine whether to block the object from printing. US 2014/0058959 A1 describes systems and methods of enforcing 3D restricted rights in rapid manufacturing and prototyping environments, which may include, in response to receiving a 3D object data representative of a 3D object, performing at least one function on the 3D object data to determine a parameter set for each respective function. At least one algorithm may be performed to determine whether at least a portion of the 3D object matches a rights-restricted 3D object. WO 2023/086137A1 describes an adaptive AI model for 3D object detection using synthetic training data. The ML model may be trained to detect certain items of interest based on a training set that is synthetically generated in real time during the training process. Recent incidents have shown that there is an increasing threat from self-manufactured or printed firearms, particularly in nations with restrictive gun laws. Furthermore, it is known that 3D printers enable the production of unregistered weapons and weapon parts. Such weapons, also known as "ghost guns", cannot be assigned to an owner when they are found, which poses enormous challenges in fighting crime. In addition, critical components of any kind could be produced in 3D printing (e.g. machine parts of centrifuges for uranium enrichment), the production of which is regulated and restricted. A printer manufacturer or manufacturer of the printing software could even be held liable. There is currently no widespread solution to address the problem described. This is especially the case for print files or 3D models that are created independently by a user. Since the spread of the plans on the Internet can hardly be prevented, solutions are urgently sought to counteract this threat. Thus, there is a demand for restricting the use of 3D printers to printing only 3D objects that are not dangerous or illegal. Summary This demand is met by an apparatus and method for administering 3D printing of 3D objects. Advantageous embodiments are addressed by the dependent claims. According to a first aspect, the present disclosure proposes an apparatus. The apparatus comprises interface circuitry configured to receive a G-code file comprising instructions for 3D printing of one or more 3D objects. The apparatus further comprises processing circuitry configured to classify a 3D object represented by the G-code file, and restrict 3D printing of the 3D object if the 3D object is classified as at least a part of a dangerous/illegal object. According to a second aspect, the present disclosure proposes a method. The method comprises receiving a G-code file comprising instructions for 3D printing of one or more 3D objects. The method further comprises classifying a 3D object represented by the G-code file, and restricting 3D printing of the 3D object if the 3D object is classified as at least a part of a dangerous/illegal object. Brief description of the Figures Some examples of apparatuses and/or methods will be described in the following by way of example only, and with reference to the accompanying figures, in which Fig. 1schematically illustrates an exemplary apparatus for administering 3D printing of 3D objects;Fig. 2depicts an exemplary G-code file and corresponding geometric features of an exemplary 3D object;Fig. 3depicts exemplary G-code file lines used in calculations of vector lengths, vector angles, and a displaced filament mass associated with the vectors;Fig. 4depicts an exemplary 3D object and a corresponding graph of vector lengths obtained within a corresponding G-code file.Fig. 5depicts another exemplary 3D object and a corresponding graph of vector lengths obtained within a corresponding G-code file;Fig. 6Aprovides a histogram depicting a percentage of all vectors that are associated with a respective vector length;Fig. 6Bprovides a histogram depicting a percentage of all vectors that are associated with a respective vector angle;Fig. 7schematically illustrates a further exemplary apparatus for administering 3D printing of 3D objects using a machine learning network;Fig. 8illustrates an exemplary neural network;Fig. 9illustrates exemplary ground truth information provided to the machine learning network for training; andFig. 10illustrates a flowchart of an exemplary method for administering 3D printing of 3D objects. Detailed Description Some examples are now described