Methods and apparatus for machine learning predictions of manufacture processes
The subject technology is related to methods and apparatus for discretization and manufacturability analysis of computer assisted design models. In one embodiment, the subject technology implements a computer-based method for the reception of an electronic file with a digital model representative of a physical object. The computer-based method determines geometric and physical attributes from a discretized version of the digital model, a cloud point version of the digital model, and symbolic functions generated through evolutionary algorithms. A set of predictive machine learning models is utilized to infer predictions related to the manufacture process of the physical object.
1. An apparatus, comprising:
a memory; and
a processor operatively coupled to the memory, the processor configured to:
partition a point cloud data structure representing a physical object into a set of shape distributions, each shape distribution from the set of shape distributions selected from a group of predetermined shape distributions, each shape distribution from the group of predetermined shape distributions representing a unique shape;
predict, via a machine learning model and based at least in part on the set of shape distributions, a set of attributes associated with a manufacture of the physical object, the set of attributes including: (1) a time to prepare a machine for the manufacture of the physical object, (2) a time to manufacture the physical object, and (3) a number of computer numerical control operations of a manufacturing machine selected for the manufacture the physical object; and
display in a graphical user interface at least one attribute from the set of attributes, the graphical user interface enabled for a user to change the at least one attribute from the set of attributes and predict, based on a user input, a subsequent set of attributes associated with the manufacture of the physical object.
2. The apparatus of claim 1 , wherein each shape distribution from the set of shape distributions is defined as a set of histograms of distances between random samples extracted from the point cloud data structure broken down by at least one of Euclidean distances, interior distances, and exterior distances.
3. The apparatus of claim 1 , wherein the processor is further configured to:
receive, via the graphical user interface an electronic file with a digital model representative of the physical object and a set of prediction requests associated with the manufacture of the physical object.
4. The apparatus of claim 1 , wherein the machine learning model includes an extremely randomized machine learning model, and the processor is further configured to:
train the extremely randomized machine learning model with a dataset including recorded manufacture times and computer numerical control operations of physical objects.
5. The apparatus of claim 1 , wherein the machine learning model includes a logistic regression model, and the processor is further configured to:
determine whether the physical object can be manufactured with a cylindrical or rectangular blank.
6. A method, comprising:
receiving, at a compute device, an electronic file with a digital model representative of a physical object;
instantiating in a memory of the compute device, a mesh scheme data structure based on the digital model with data values corresponding to a discretized version of the digital model;
determining, based on the mesh scheme data structure, a first set of parameters corresponding to physical attributes of the physical object;
instantiating in the memory of the compute device, a point cloud data structure with data values corresponding to three-dimensional coordinates downsampled from the digital model;
determining, based on the point cloud data structure, a second set of parameters corresponding to physical attributes of the physical object the second set of parameters includes a set of shape distributions related to one or more shapes within the digital model;
inferring based on the first set of parameters, the second set of parameters, and at least one machine learning model, a set of axioms associated with a manufacture process of the physical object; and
determining a set of predicted outcomes associated with the manufacture process of the physical object based on the set of axioms.
7. The method of claim 6 , wherein the electronic file is received from a client compute device and the set of predicted outcomes is sent to the client compute device in near real-time.
8. The method of claim 6 , wherein the at least one machine learning model is selected from a group consisting of a logistic regression machine learning model, a random forest machine learning model, and an extremely randomized trees machine learning model.
9. The method of claim 6 , wherein the first set of parameters includes any of a data value corresponding to a volume of the physical object, a data value corresponding to a surface area of the physical object, a 3-tuple data structure with a length value, a width value, and a height value associated with a prism enclosing the digital model, a 3-tuple data structure with coordinate data values of a center of the prism, a data value corresponding to a volume of a convex hull enclosing the digital model, and a data value corresponding to a number of holes associated with the physical object.
10. The method of claim 6 , wherein the first set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on orientation of at least one surface of the physical object.
11. The method of claim 6 , wherein the first set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on symmetric attributes associated with the physical object.
12. The method of claim 6 , wherein the first set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on distances between surfaces of the physical object.
13. The method of claim 6 , wherein the second set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on orientation of at least one surface of the physical object.
14. The method of claim 6 , wherein the second set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on symmetric attributes associated with the physical object.
15. The method of claim 6 , wherein the second set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on distances between surfaces of the physical object.
16. The method of claim 6 , wherein the set of shape distributions is determined based on a set of distance data values corresponding to distances between random digital samples of shapes and one or more shapes within the digital model, the set of distance data values includes any of an Euclidean distance data value, a full histogram data value, an interior distance data value, an exterior distance data value, and any combination of distance data values thereof.
17. The method of claim 6 , wherein the set of axioms includes any of a data value corresponding to a manufacturing method, a data value corresponding to a number of computer numerical control machine operations to manufacture the physical object, a data value corresponding to a time to equip a machine to manufacture the physical object, a data value corresponding to a time to manufacture a first instance of the physical object, a data value corresponding to a time to manufacture subsequent instances of the physical object after a first instance of the physical object has been manufactured, an identifier of a tool bit blank associated with a computer numerical control machine to manufacture the physical object, an identifier of a fixture type corresponding to a support device used during the manufacture of a physical object, and any combination of axioms thereof.
18. The method of claim 6 , wherein the set of predicted outcomes includes a predicted cost associated with the manufacture process of the physical object.
19. The method of claim 6 , wherein the set of axioms includes an axiom associated with the manufacture process of the physical object not used to train the at least one machine learning model.
20. A method, comprising:
receiving, at a compute device, an electronic file with a digital model representative of a physical object;
instantiating in a memory of the compute device, a mesh scheme data structure based on the digital model with data values corresponding to a discretized version of the digital model;
determining, based on the mesh scheme data structure, a first set of parameters corresponding to physical attributes of the physical object;
retrieving from the memory of the compute device a set of symbolic functions derived from an evolutionary computation model;
determining, based on the first set of parameters and the set of symbolic functions, a second set of parameters corresponding to physical attributes of the physical object;
inferring based on the second set of parameters and at least on one machine learning model, a set of axioms associated with a manufacture process of the physical object; and
calculating a set of predicted outcomes associated with the manufacture process of the physical object based on the set of axioms.
21. The method of claim 20 , wherein the evolutionary computation model is based on an execution of a genetic algorithm process.
22. The method of claim 20 , wherein the physical object is a first physical object, the data model is a first data model, and the evolutionary computation model is trained with datasets including any of a data value corresponding to a volume of a second physical object, a data value corresponding to a surface area of the second physical object, a 3-tuple data structure with a length value, a width value, and a height value associated with a prism enclosing a second digital model, a 3-tuple data structure with coordinate data values of a center of the prism, a data value corresponding to a volume of a convex hull enclosing the second digital model, and a data value corresponding to a number of holes associated with the second physical object.
23. The method of claim 20 , wherein the first set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on, dimensional analysis of the discretized digital model, orientation of at least one surface of the physical object, any of a data value corresponding to a measure of the physical object calculated based on symmetric attributes associated with the physical object, and any of a data value corresponding to a measure of the physical object calculated based on distances between surfaces of the physical object.
24. The method of claim 20 , wherein the second set of parameters includes any of a data value corresponding to a measure of the physical object calculated based on orientation of at least one surface of the physical object, any of a data value corresponding to a measure of the physical object calculated based on symmetric attributes associated with the physical object, and any of a data value corresponding to a measure of the physical object calculated based on distances between surfaces of the physical object.
25. The method of claim 20 , wherein:
the set of predicted outcomes includes a predicted cost associated with the manufacture process of the physical object; and
wherein the electronic file is received from a client compute device and the set of predicted outcomes is sent to the client compute device in near real-time.