Distortion prediction for additive manufacturing using image analysis

1 . A method comprising:

performing an image analysis on an image of a current layer of an object being manufactured by an additive manufacturing system to identify an actual exposed surface in the image of the current layer;

performing a build simulation using build data to generate a simulated distortion for the current layer of the object, at least one previous layer of the object and a next layer of the object;

performing a spreading simulation to predict a simulated exposed surface at the current layer from the simulated distortion, wherein the spreading simulation models a spreading of powder on the current layer based on a damage caused to a re-coater during application of the at least one previous layer;

evaluating the build data to determine a value of an influencing factor of the build data affecting the actual exposed surface for the current layer;

predicting, using a recurrent neural network, a reduction of the actual exposed surface within the next layer based on the image of the current layer and the simulated exposed surface at the current layer for which a re-coater interference does not occur in the next layer; and

implementing an action to continue fabrication of the next layer based at least in part on the predicted reduction.

2 . The method of claim 1 , wherein predicting the predicted distortion for the next layer is further based at least in part on a result of the spreading simulation.

3 . The method of claim 1 , wherein the spreading simulation models powder spreading on the layer further based on at least one selected from a group consisting of a chamber condition, a re-coater type, a re-coater speed, or a layer thickness.

4 . The method of claim 1 , further comprising training the recurrent neural network.

5 . The method of claim 4 , wherein training the recurrent neural network comprises inputting training data into the recurrent neural network to predict a re-coater interference in the current layer.

6 . The method of claim 5 , wherein training the recurrent neural network further comprises comparing the predicted re-coater interference in the current layer with an actual re-coater interference in the current layer.

7 . The method of claim 6 , wherein training the recurrent neural network further comprises, responsive to determining that an accuracy of the comparison is acceptable, generating a trained recurrent neural network.

8 . The method of claim 6 , wherein training the recurrent neural network further comprises, responsive to determining that an accuracy of the comparison is unacceptable, performing at least one additional training iteration, wherein at least one recurrent neural network parameter is adjusted during each iteration.

9 . A system comprising:

a processing system comprising a memory and a processing device, the processing system for executing computer readable instructions, the computer readable instructions controlling the processing device to perform operations comprising:

performing an image analysis on an image of a current layer of an object being manufactured by an additive manufacturing system to identify an actual exposed surface in the image of the current layer;

performing a build simulation using build data to generate a simulated distortion for the current layer of the object, at least one previous layer of the object and a next layer of the object;

performing a spreading simulation to predict a simulated exposed surface at the current layer from the simulated distortion, wherein the spreading simulation models a spreading of powder on the current layer based on a damage caused to a re-coater during application of the at least one previous layer;

evaluating the build data to determine a value of an influencing factor of the build data affecting the actual exposed surface for the current layer;

predicting, using a recurrent neural network, a reduction of the actual exposed surface within the next layer based on the image of the current layer and the simulated exposed surface at the current layer for which a re-coater interference does not occur in the next layer; and

implementing an action to continue fabrication of the next layer based at least in part on the predicted reduction.

10 . The system of claim 9 , further comprising the additive manufacturing system.

11 . The system of claim 10 , the instructions further comprising training the recurrent neural network, wherein training the recurrent neural network comprises inputting training data into the recurrent neural network to generate the predicted re-coater interference.

12 . The system of claim 11 , wherein training the recurrent neural network further comprises comparing the at least one of the predicted exposed region and the predicted re-coater interference with an actual exposed region.

13 . The system of claim 12 , wherein training the recurrent neural network further comprises, responsive to determining that an accuracy of the comparison is acceptable, generating a trained recurrent neural network.

14 . The system of claim 12 , wherein training the recurrent neural network further comprises, responsive to determining that an accuracy of the comparison is unacceptable, performing at least one additional training iteration, wherein at least one machine learning model parameter is adjusted during each iteration.

15 . The system of claim 9 , further comprising a camera, wherein the operations further comprise capturing the image of the layer of the object being manufactured by the additive manufacturing system during the manufacturing.

16 . The system of claim 9 , wherein implementing the action comprises at least one of altering a laser power of the additive manufacturing system or skipping manufacturing of at least a portion of the next layer.