IP Library Granted Patent US 12669796
Granted Patent B2
US 12669796 · App. 17/395,119 · Granted Jun 30, 2026

Methods and systems for generating an instantaneous quote of any part without toolpathing

Inventor: Stefan Emilov Atev (Bethesda, MD)
Assignee: Protolabs, Inc.
G05B19/406G06F30/20G06N20/00G05B2219/35212
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12669796
App. No.
17/395,119
Granted
Jun 30, 2026
Kind
B2
Abstract

Methods and In an aspect a method of generating an instantaneous quote of any part without toolpathing, the method includes receiving, using a computing device, a geometric model of a part, constructing, using the computing device, at least a rotation-invariant feature as a function of the geometric model, predicting, using the computing device, a manufacturing time as a function of the at least a rotation-invariant feature and a manufacturing time machine learning model, selecting, using the computing device, a stock as a function of the at least a rotation-invariant and a stock selection machine learning model feature, and estimating, using the computing device, a quote as a function of the manufacturing time and the stock.

Claims (38)

1 . A method of generating an instantaneous quote of a computer numeric control (CNC) part without toolpathing, the method comprising:

receiving, using a computing device, a geometric model of a part to be manufactured using subtractive manufacturing comprising CNC manufacturing;

constructing, using the computing device, at least a rotation-invariant feature as a function of the geometric model, wherein:

the at least a rotation-invariant feature is in a form adapted for use in a machine learning process and comprises a co-occurrence histogram representative of redundant or co-occurring data in the geometric model of the part to be manufactured; and

the co-occurrence histogram comprises a co-occurrence matrix including a plurality of elements comprising at least a distance to occluding geometry which is evaluated over a surface of the geometric model of the part to be manufactured;

predicting, by the computing device, a manufacturing time for the part by:

receiving, from a client device, a computer-aided design (CAD) file storing the geometric model of the part;

generating a tessellated surface representation of the geometric model;

constructing the at least one rotation-invariant feature by generating the co-occurrence histogram in matrix form that represents at least one of surface orientation, part thickness, and the distance to occluding geometry; and

providing the co-occurrence matrix to a manufacturing time machine-learning model stored in memory and comprising a first artificial neural network (ANN) to output the manufacturing time, wherein the manufacturing time comprises a predicted time to perform at least a milling operation;

selecting, using the computing device, a stock as a function of the at least a rotation-invariant feature, including the co-occurrence matrix, and a stock selection machine learning model comprising a second ANN; and

estimating, using the computing device, a quote as a function of the manufacturing time, including the time to perform the at least a milling operation, and the stock for CNC manufacturing of the part to be manufactured without generating a toolpath for a CNC machine tool during the prediction of the manufacturing time, the selection of the stock and the estimation of the quote.

2 . The method of claim 1 further comprising classifying, using the computing device, a quotation state as a function of the at least a rotation-invariant feature and a go/no-go classifier machine learning model.

3 . The method of claim 2 , further comprising training, using the computing device, the go/no-go classifier machine learning model according to an F-measure function and a desired false positivity rate.

4 . The method of claim 1 , further comprising training, using the computing device, the manufacturing time machine learning model as a function of manufacturing time training data comprising at least an element of data comprising a logarithmic representation of manufacturing time.

5 . The method of claim 1 , wherein the at least a rotation-invariant feature comprises a surface curvature feature.

6 . The method of claim 1 , wherein the at least a rotation-invariant feature comprises a convex hull feature.

7 . The method of claim 1 , wherein the at least a rotation-invariant feature comprises a decomposition of patches.

8 . The method of claim 1 , wherein predicting the manufacturing time further comprises predicting the manufacturing time as a function of the stock, the at least a rotation-invariant feature and the manufacturing time machine learning model.

9 . A system for generating an instantaneous quote of a computer numeric control (CNC) part without toolpathing, the system comprising a computing device configured to:

receive a geometric model of a part to be manufactured using subtractive manufacturing comprising CNC manufacturing;

construct at least a rotation-invariant feature as a function of the geometric model, wherein:

the at least a rotation-invariant feature is in a form adapted for use in a machine learning process and comprises a co-occurrence histogram representative of redundant or co-occurring data in the geometric model of the part to be manufactured; and

the co-occurrence histogram comprises a co-occurrence matrix including a plurality of elements comprising at least a distance to occluding geometry which is evaluated over a surface of the geometric model of the part to be manufactured;

predict a manufacturing time for the part by:

receiving, from a client device, a computer-aided design (CAD) file storing the geometric model of the part;

generating a tessellated surface representation of the geometric model;

constructing the at least one rotation-invariant feature by generating the co-occurrence histogram in matrix form that represents at least one of surface orientation, part thickness, and the distance to occluding geometry; and

providing the co-occurrence matrix to a manufacturing time machine-learning model stored in memory and comprising a first artificial neural network (ANN) to output the manufacturing time, wherein the manufacturing time comprises a predicted time to perform at least a milling operation;

select a stock as a function of the at least a rotation-invariant feature, including the co-occurrence matrix, and a stock selection machine learning model comprising a second ANN; and

estimate a quote as a function of the manufacturing time, including the time to perform the at least a milling operation, and the stock for CNC manufacturing of the part to be manufactured without generating a toolpath for a CNC machine tool during the prediction of the manufacturing time, the selection of the stock and the estimation of the quote.

10 . The system of claim 9 wherein the computing device is further configured to classify a quotation state as a function of the at least a rotation-invariant feature and a go/no-go classifier machine learning model.

11 . The system of claim 10 , wherein the computing device is further configured to train the go/no-go classifier machine learning model according to an F-measure function and a desired false positivity rate.

12 . The system of claim 9 , wherein the computing device is further configured to train the manufacturing time machine learning model as a function of manufacturing time training data comprising at least an element of data comprising a logarithmic representation of manufacturing time.

13 . The system of claim 9 , wherein the at least a rotation-invariant feature comprises a surface curvature feature.

14 . The system of claim 9 , wherein the at least a rotation-invariant feature comprises a convex hull feature.

15 . The system of claim 9 , wherein the at least a rotation-invariant feature comprises a decomposition of patches.

16 . The system of claim 9 , wherein predicting the manufacturing time further comprises predicting the manufacturing time as a function of the stock, the at least a rotation-invariant feature and the manufacturing time machine learning model.