System and method for minimizing deviations in 3D printed and sintered parts

1. A method for optimizing a three-dimensional (3D) printing process comprising:

receiving 3D data characterizing a target 3D part;

generating a 3D part mesh based on the received 3D data;

generating a Signed Distance Field (SDF) based on the 3D part mesh;

performing a predictive analysis using the SDF to generate a predicted error field;

calculating a correction field based on the predicted error field;

applying the correction field to the 3D part mesh to produce a corrected 3D part mesh;

creating a corrected 3D print profile using the corrected 3D part mesh;

outputting the corrected 3D print profile to a 3D printing device; and

printing a corrected 3D part using the 3D printing device and the corrected 3D print profile.

2. The method of claim 1 , wherein calculating the correction field includes executing a transformation function.

3. The method of claim 1 , further comprising executing, with respect to a second 3D part mesh based on second 3D data characterizing a second target 3D part, an iterative process including:

creating a 3D print profile using the second 3D part mesh;

outputting the 3D print profile to the 3D printing device;

printing a 3D part using the 3D printing device and the 3D print profile;

scanning the 3D part to generate a measured point-set; and

performing at least one regression analysis on the measured point-set.

4. The method of claim 3 , further comprising:

determining a coarse fit between the second 3D data and the second 3D part mesh;

determining, based on the coarse fit, a fine fit between the second 3D data and the second 3D part mesh; and

outputting, based on the coarse fit and the fine fit, an iterative error field.

5. The method of claim 4 , wherein determining the coarse fit includes executing a least squares regression analysis on the second 3D data and the second 3D part mesh to generate a seed value.

6. The method of claim 5 , wherein determining the fine fit includes executing a patchwise least squares analysis on the seed value.

7. The method of claim 4 , wherein the 3D print profile includes a plurality of scan points, wherein generating the iterative error field includes dividing each scan point of the plurality of scan points into a second plurality of scan points, and wherein a third plurality of scan points includes each respective second plurality of scan points.

8. The method of claim 7 , further comprising:

executing a fine fit analysis on the third plurality of scan points; and

outputting, based on the fine fit analysis, the iterative error field.

9. The method of claim 8 , wherein the predictive analysis comprises:

computing a distance field of the 3D part mesh over a 3D voxel grid to generate a 3D matrix of SDF values; and

processing the 3D matrix of SDF values using a neural network, wherein the neural network comprises a plurality of subnetworks arranged in a sequence from highest to lowest and processing comprises processing the 3D matrix of SDF values through each of the subnetworks in the sequence, and each subnetwork is configured to receive a preceding output representation generated by a preceding subnetwork in the sequence.

10. The method of claim 9 , further comprising:

training the neural network on a training data set obtained from printed and scanned 3D parts, the training data set including the iterative error field.

11. A non-transitory computer-readable medium storing sequences of computer-executable instructions for optimizing a three-dimensional (3D) printing process, the sequences of computer-executable instructions including instructions that instruct at least one processor to:

receive 3D data characterizing a target 3D part;

generate a 3D part mesh based on the received 3D data;

generate a Signed Distance Field (SDF) based on the 3D part mesh;

perform a predictive analysis using the SDF to generate a predicted error field;

calculate a correction field based on the predicted error field;

apply the correction field to the 3D part mesh to produce a corrected 3D part mesh;

create a corrected 3D print profile using the corrected 3D part mesh;

output the corrected 3D print profile to a 3D printing device; and

print a corrected 3D part using the 3D printing device and the corrected 3D print profile.

12. The computer-readable medium of claim 11 , wherein, in instructing the at least one processor to calculate the correction field, the sequences of computer-executable instructions instruct the at least one processor to execute a transformation function.

13. The computer-readable medium of claim 11 , wherein the sequences of computer-executable instructions further instruct the at least one processor to execute, with respect to a second 3D part mesh based on second 3D data characterizing a second target 3D part, an iterative process, and wherein, in instructing the at least one processor to perform the iterative process, the sequences of computer-executable instructions instruct the at least one processor to:

create a 3D print profile using the second 3D part mesh;

output the 3D print profile to the 3D printing device;

print a 3D part using the 3D printing device and the 3D print profile;

scan the 3D part to generate a measured point-set; and

perform at least one regression analysis on the measured point-set.

14. The computer-readable medium of claim 13 , wherein the sequences of computer-executable instructions instruct the at least one processor to:

determine a coarse fit between the second 3D data and the second 3D part mesh;

determine, based on the coarse fit, a fine fit between the second 3D data and the second 3D part mesh; and

output, based on the coarse fit and the fine fit, an iterative error field.

15. The computer-readable medium of claim 14 , wherein, in instructing the at least one processor to determine the coarse fit, the sequences of computer-executable instructions instruct the at least one processor to execute a least squares regression analysis on the 3D data and the 3D part mesh to generate a seed value.

16. The computer-readable medium of claim 15 , wherein, in instructing the at least one processor to determine the fine fit, the sequences of computer-executable instructions instruct the at least one processor to execute a patchwise least squares analysis on the seed value.

17. The computer-readable medium of claim 14 , wherein the 3D print profile includes a plurality of scan points, wherein, in instructing the at least one processor to generate the iterative error field, the sequences of computer-executable instructions instruct the at least one processor to divide each scan point of the plurality of scan points into a second plurality of scan points, and wherein a third plurality of scan points includes each respective second plurality of scan points.

18. The computer-readable medium of claim 17 , wherein the sequences of computer-executable instructions further include instructions that instruct at least one processor to:

execute a fine fit analysis on the third plurality of scan points; and

output, based on the fine fit analysis, the iterative error field.

19. The computer-readable medium of claim 18 , wherein, in instructing the at least one processor to perform the predictive process, the sequences of computer-executable instructions instruct the at least one processor to:

compute a distance field of the 3D part mesh over a 3D voxel grid to generate a 3D matrix of SDF values; and

process the 3D matrix of SDF values using a neural network, wherein the neural network comprises a plurality of subnetworks arranged in a sequence from highest to lowest and processing comprises processing the 3D matrix of SDF values through each of the subnetworks in the sequence, and each subnetwork is configured to receive a preceding output representation generated by a preceding subnetwork in the sequence.

20. The computer-readable medium of claim 19 , wherein the sequences of computer-executable instructions further include instructions that instruct at least one processor to:

train the neural network on a training data set obtained from printed and scanned 3D parts, the training data set including the iterative error field.