MANUFACTURING POWDER PREDICTIONS
Examples of methods are described. In some examples, a method includes determining, using a variational autoencoder model, a latent space representation based on a three-dimensional (3D) input. In some examples, the 3D input represents a build of manufacturing powder. In some examples, the method includes predicting manufacturing powder degradation based on the latent space representation.
1 . A method, comprising:
determining, using a variational autoencoder model, a latent space representation based on a three-dimensional (3D) input representing a build of manufacturing powder; and
predicting manufacturing powder degradation based on the latent space representation.
2 . The method of claim 1 , further comprising:
determining voxels based on build data to produce the 3D input; and
inputting the voxels to the variational autoencoder model to determine the latent space representation.
3 . The method of claim 1 , wherein predicting the manufacturing powder degradation comprises predicting, using a first machine learning model, a predicted stress based on the latent space representation.
4 . The method of claim 3 , wherein the predicted stress is predicted based on the latent space representation concatenated with an attribute.
5 . The method of claim 4 , wherein the latent space representation is concatenated with an initial stress, an X location, a y location, a Z location, a build height, and a calculated stress.
6 . The method of claim 3 , wherein predicting the manufacturing powder degradation comprises predicting, using a second machine learning model, the powder degradation based on the predicted stress.
7 . The method of claim 1 , wherein the variational autoencoder model is trained with a decoder.
8 . The method of claim 7 , wherein the variational autoencoder model is to determine the latent space representation without the decoder at an inferencing stage.
9 . The method of claim 1 , wherein each dimension of the latent space representation is independent of each other dimension.
10 . An apparatus, comprising:
a memory; and
a processor coupled to the memory, wherein the processor is to:
determine voxels representing a build of manufacturing powder in three dimensions;
input the voxels to a variational autoencoder model to produce a latent space representation of the build; and
determine a powder quality metric based on the latent space representation.
11 . The apparatus of claim 10 , wherein the processor is to determine the powder quality metric by predicting, using a first machine learning model, a predicted stress based on the latent space representation.
12 . The apparatus of claim 11 , wherein the processor is to predict, using a second machine learning model, the powder quality metric as a b* component of a color space based on the predicted stress.
13 . A non-transitory tangible computer-readable medium comprising instructions when executed cause a processor of an electronic device to:
voxelize a manufacturing build to produce voxels;
determine, using a variational autoencoder model without a decoder, a latent space representation based on the voxels, wherein the variational autoencoder model is trained with the decoder; and
predict, using a machine learning model, manufacturing powder degradation based on the latent space representation.
14 . The non-transitory tangible computer-readable medium of claim 13 , wherein the instructions when executed cause the processor of the electronic device to:
train the variational autoencoder model using training voxels to produce reconstructed voxels at an output of the decoder; and
generate a visualization indicating a difference between the training voxels and the reconstructed voxels.
15 . The non-transitory tangible computer-readable medium of claim 13 , wherein the instructions when executed cause the processor of the electronic device to sample a dimension of the latent space representation while maintaining other dimensions of the latent space representation.