Extended fabrication-aware convolution learning framework for predicting 3D shape deformation in additive manufacturing
An apparatus, system, and method is provided for predicting 3D shape deformation in additive manufacturing. A convolution learning framework for shape deviation modeling provides joint learning for a wide class of 3D shapes including both spherical and polyhedral shapes. A 3D cookie-cutter function can effectively capture the unique pattern of the shape deformation for polyhedral shapes. Since 3D freeform shapes can be approximated as a combination of spherical and polyhedral patches, the extended convolution learning framework builds a foundation for modeling and predicting the quality of 3D freeform shapes. By changing the kernel function and considering new distance measures for points from different shapes, the spatial correlations among different shapes can be correctly incorporated. The predicted deformation may be used to offset the machine instructions to an additive manufacturing machine to ameliorate deformation.
1 . A method for predicting a shape deformation of a 3D printed product comprising a 3D model made up from primitive shapes and having a plurality of individual layers, the method comprising:
determining, using a processor, an input function for each individual layer, the input function describing a layer-wise shape deformation of the each individual layer with 2D freeform shapes;
determining, using the processor, a transfer function for accumulating the individual layers together using dome shapes;
calculating, using the processor, a convolution using the input function and the transfer function to obtain a shape deformation of the dome shapes;
determining, using the processor, a cookie-cutter function for connecting the primitive shapes comprising the 3D model; and
calculating, using the processor, a summation of the convolution and the cookie-cutter function to generate a deformation function that predicts the shape deformation of the 3D printed product.
2 . The method of claim 1 , further comprising generating machine control instructions for an additive manufacturing machine corresponding to the input function offset according to the deformation function to cancel at least a portion of the shape deformation of the 3D printed product.
3 . The method of claim 1 , wherein the primitive shapes comprise sphere shapes and polyhedron shapes.
4 . The method of claim 1 , further comprising determining the transfer function using machine learning techniques.
5 . The method of claim 1 , further comprising determining the cookie-cutter function using machine learning techniques.
6 . The method of claim 1 , wherein the transfer function comprises a plurality of functional bases including Fourier bases.
7 . The method of claim 1 , wherein determining the deformation function also includes executing a gaussian process regression (GPR).
8 . A system for predicting a shape deformation of a 3D printed product comprising a 3D model made up from primitive shapes and having a plurality of individual layers, the system comprising:
a non-transitory computer-readable memory configured to store instructions; and
one or more processor configured to:
determine an input function for each individual layer, the input function describing a layer-wise shape deformation of the each individual layer with 2D freeform shapes;
determine a transfer function for accumulating the individual layers together using dome shapes;
calculate a convolution using the input function and the transfer function to obtain a shape deformation of the dome shapes;
determine a cookie-cutter function for connecting the primitive shapes comprising the 3D model; and
calculate a summation of the convolution and the cookie-cutter function to generate a deformation function that predicts the shape deformation of the 3D printed product.
9 . The system of claim 8 , wherein the processor is further configured to generate machine control instructions for an additive manufacturing machine corresponding to the input function offset according to the deformation function to cancel at least a portion of the shape deformation of the 3D printed product.
10 . The system of claim 8 , wherein the primitive shapes comprise sphere shapes and polyhedron shapes.
11 . The system of claim 8 , wherein the processor determines the transfer function using machine learning techniques.
12 . The system of claim 8 , wherein the processor determines the cookie-cutter function using machine learning techniques.
13 . The system of claim 8 , wherein the transfer function comprises a plurality of functional bases including Fourier bases.
14 . The system of claim 8 , wherein the processor also executes a gaussian process regression (GPR) when determining the deformation function.
15 . A non-transitory computer-readable medium comprising computer readable instructions, which when executed by a processor, cause the processor to perform operations for predicting a shape deformation of a 3D printed product comprising a 3D model made up from primitive shapes and having a plurality of individual layers, the operations comprising:
determining, using the processor, an input function for each individual layer, the input function describing a layer-wise shape deformation of the each individual layer with 2D freeform shapes;
determining, using the processor, a transfer function for accumulating the individual layers together using dome shapes;
calculating, using the processor, a convolution using the input function and the transfer function to obtain a shape deformation of the dome shapes;
determining, using the processor, a cookie-cutter function for connecting the primitive shapes comprising the 3D model; and
calculating, using the processor, a summation of the convolution and the cookie-cutter function to generate a deformation function that predicts the shape deformation of the 3D printed product.
16 . The non-transitory computer-readable medium of claim 15 , wherein the operations further comprise generating machine control instructions for an additive manufacturing machine corresponding to the input function offset according to the deformation function to cancel at least a portion of the shape deformation of the 3D printed product.
17 . The non-transitory computer-readable medium of claim 15 , wherein the primitive shapes comprise sphere shapes and polyhedron shapes.
18 . The non-transitory computer-readable medium of claim 15 , wherein the operations further comprise determining the transfer function using machine learning techniques.
19 . The non-transitory computer-readable medium of claim 15 , wherein the operations further comprise determining the cookie-cutter function using machine learning techniques.
20 . The non-transitory computer-readable medium of claim 15 , wherein the transfer function comprises a plurality of functional bases including Fourier bases.