Intelligent additive manufacturing
AssigneeInkbit, LLC
InventorsWojciech Matusik, Desai Chen
Life of this patent
Estimating the term…
Abstract
An approach to intelligent additive manufacturing makes use of one or more of machine learning, feedback using machine vision, and determination of machine state. In some examples, a machine learning transformation receives data representing a partially fabricated object and a model of an additional part (e.g., layer) of the part, and produces a modified model that is provided to a printer. The machine learning predistorter can compensate for imperfections in the partially fabricated object as well as non-ideal characteristics of the printer, thereby achieving high accuracy.
Claims18
A method for iteratively fabricating a 3D object via a printer, the method comprising, in a first iteration:
- receiving a 3D specification characterizing a first part of the 3D object to be printed on a second part of the 3D object fabricated in a previous iteration;
- receiving scan data representative of the second part of the 3D object;
- using a predistorter to process the 3D specification of the first part using the scan data to produce a modified 3D specification of the first part of the 3D object, wherein the predistorter is configured with configuration data to compensate for at least some characteristics of a printing process;
- and
- causing the printer to print the first part of the object onto the second part according to the printing process based at least in part on the modified 3D specification, wherein the printed first part mates with the second part.
A method for fabricating a 3D object via a printer, the method comprising:
- receiving a 3D specification of a first part of the 3D object;
- using a predistorter to process the 3D specification to produce a modified 3D specification of the first part of the object, wherein the predistorter is configured with configuration data to compensate for at least some characteristics of a printing process;
- and
- causing the printer to print the first part of the object according to the printing process based at least in part on the modified 3D specification;
- wherein the 3D specification represents an additional part of the 3D object to be fabricated on a previously fabricated part of the 3D object.
A method for configuring a 3D printing system comprising:
- receiving training data comprising a plurality to training tuples, each training tuple comprising:
- a specification,
- scan data representing a part on which fabrication of the specification is performed, and
- scan data representing fabrication of the specification using a printer of the printing system;
- and
- processing the training data to produce configuration data for configuring a predistorter for use in printing using the printing system.
A method for configuring a predistorter to be used in fabricating a 3D object via a printer, the method comprising:
- estimating first values representing a printing process, wherein the printing process is configured to receive a specification of a first part of a 3D object to be fabricated on a second part of the 3D object fabricated in a previous iteration and cause the object to be fabricated;
- estimating second values representing a preprocess for the printing process, wherein the preprocess is configurable to accept the specification of the first part of the 3D object and produce a modified specification of said object;
- wherein estimating the first values and the second values comprises optimizing a loss function that depends at least in part on a combined characteristic of the preprocess and the printing process.
Dependencies are inferred from the claim text.