IP Library › Granted Patent US 12,728,459
Granted Patent B2
US 12,728,459 · App. 18/548,160 · Granted Sep 8, 2026

Anomaly detection in additive manufacturing using meltpool monitoring, and related devices and systems

Inventors: Gunaranjan Chaudhry (Bangalore, IN); Jayesh Rameshlal Jain (The Woodlands, TX); Thomas Dobrowolski (The Woodlands, TX); Chad Yates (Houston, TX); Aaron Avagliano (Tomball, TX)
Assignee: Baker Hughes Oilfield Operations LLC
B22F10/85B22F10/31B22F10/34B22F10/366B33Y50/02
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 12,728,459
App. No.
18/548,160
Filed
Aug 28, 2023
Granted
Sep 8, 2026
Kind
B2
Art Unit
2117
USPC
700/98
Abstract

Methods for anomaly detection in additive manufacture using meltpool monitoring are disclosed. A method includes obtaining a process model representative of an object to be generated through additive manufacture. The method also includes generating, based on the process model and using a hybrid machine-learning model, an instruction for generating the object through additive manufacture. Another method includes generating a layer of an object, and taking a reading relative to the generation of the layer. The other method also includes updating, based on the reading and using a hybrid machine-learning model, a process model, the process model representative of the object. The other method also includes generating, based on the updated process model and using the hybrid machine-learning model, an instruction for generating a subsequent layer of the object through additive manufacture. Related systems and devices are also disclosed.

Claims (36)

1 . A method comprising:

obtaining a process model representative of an object to be generated through additive manufacture;

generating, based on the process model and using a hybrid machine-learning model, an instruction for generating the object through additive manufacture, the instruction comprising a threshold for additive manufacturing based, at least in part, on a correlation between input area energy density and emitted power; and

generating the object through additive manufacture according to the instruction comprising:

generating a layer of the object;

taking a reading relative to the generation of the layer;

comparing the reading to the threshold of the instruction;

adjusting, based on the comparison of the reading to the threshold, and using the hybrid machine-learning model, the instruction, the adjustment, based at least in part, on the correlation between the input area energy density and the emitted power; and

generating a subsequent layer of the object according to the adjusted instruction.

2 . The method of claim 1 , wherein the hybrid machine-learning model was trained using simulated data and measured data.

3 . The method of claim 1 , further comprising training the hybrid machine-learning model using simulated data and measured data.

4 . The method of claim 1 , further comprising generating the process model based on a build file.

5 . The method of claim 1 , wherein the hybrid machine-learning model was trained using data exhibiting the correlation between the input area energy density and the emitted power.

6 . The method of claim 1 , wherein the instruction further comprises an adjustment for additive manufacture responsive to a crossing of the threshold, the adjustment, based at least in part, on the correlation between the input area energy density and the emitted power.

7 . The method of claim 1 , wherein the reading is indicative of a temperature at a location of the layer and the adjusted instruction includes information related to operation of an energy source configured to provide energy for additive manufacture.

8 . The method of claim 1 , wherein the reading is indicative of one or more of: emissive power, energy density, intensity, scaled temperature, powder-bed depth, powder-bed density, a degree of vibration of a recoater, acoustic emissions, a degree of humidity, and a strength of an electromagnetic field at one or more locations of the layer and the adjusted instruction includes information related to one or more of: gas-flow speed, recoating direction, laser power, laser focus, scan speed, scan pattern, scan strategy, scan interval time, layer thickness, hatch spacing, and hatch distance.

9 . The method of claim 1 , wherein the reading that does not satisfy the threshold is indicative of an anomaly and the adjusted instruction includes information related to the anomaly.

10 . The method of claim 9 , wherein the adjusted instruction, based at least in part on the correlation between the input area energy density and the emitted power, includes information for correcting the anomaly while generating the subsequent layer.

11 . The method of claim 10 , wherein generating the object through additive manufacture according to the instruction further comprises correcting the defect while generating the subsequent layer of the object according to the adjusted instruction.

12 . A method comprising:

generating a layer of an object according to an instruction generated based on a process model and using a hybrid machine-learning model, the instruction comprising a threshold for additive manufacturing based, at least in part, on a correlation between input area energy density and emitted power;

taking a reading relative to the generation of the layer;

comparing the reading to the threshold of the instruction;

adjusting, based the comparison of the reading to the, and using a hybrid machine-learning model, the instruction, the adjustment, based at least in part, on the correlation between the input area energy density and the emitted power; and

generating a subsequent layer of the object according to the adjusted instruction.

13 . The method of claim 12 , wherein the hybrid machine-learning model was trained using simulated data and measured data.

14 . The method of claim 12 , further comprising, adjusting the process model, prior to updating the process model, generating the process model based on a build file.

15 . A system for additive manufacture, the system comprising:

a simulator configured to generate a process model according to a build file, the process model representative of an object to be generated through additive manufacture;

a hybrid machine-learning model trained using simulated data and measured data, the hybrid machine-learning model configured to generate, based on the process model, an instruction for generating the object, the instruction comprising a threshold for additive manufacturing based, at least in part, on a correlation between input area energy density and emitted power; and

an object generator configured to generate an object through additive manufacture according to the build file and the instruction: comprising:

generating a layer of the object;

taking a reading relative to the generation of the layer;

comparing the reading to the threshold of the instruction;

adjusting, based on the comparison of the reading to the threshold, and using the hybrid machine-learning model, the instruction, the adjustment, based at least in part, on the correlation between the input area energy density and the emitted power; and

generating a subsequent layer of the object according to the adjusted instruction.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Aug 28, 2023
From: CHAUDHRY, GUNARANJAN; JAIN, JAYESH RAMESHLAL; DOBROWOLSKI, THOMAS; YATES, CHAD; AVAGLIANO, AARON
To: BAKER HUGHES OILFIELD OPERATIONS LLC
Reel/Frame 064723/0800 →
Priority Claims (1)
IN 202111008523 · Mar 1, 2021 · national
Continuity (2)
Related Publication 20240131591A1 · Apr 25, 2024
Related Publication 20240227019A9 · Jul 11, 2024
References Cited (29)
US 20150217517A1 · Karpas · 2015 [cited by examiner]
US 20150321422A1 · Boyer · 2015 [cited by examiner]
US 20150331402A1 · Lin · 2015 [cited by examiner]
US 20170144378A1 · Giera · 2017 [cited by applicant]
US 20180071987A1 · Tsumuraya · 2018 [cited by examiner]
US 20180079125A1 · Perez et al. · 2018 [cited by applicant]
US 20180243977A1 · Meinders · 2018 [cited by examiner]
US 20180341248A1 · Mehr et al. · 2018 [cited by applicant]
US 20190004496A1 · Blom et al. · 2019 [cited by applicant]
US 20190134709A1 · Dave et al. · 2019 [cited by applicant]
US 20190283333A1 · Hwang · 2019 [cited by examiner]
US 20190299536A1 · Putman et al. · 2019 [cited by applicant]
US 20200016883A1 · Reese · 2020 [cited by examiner]
US 20200096970A1 · Mehr et al. · 2020 [cited by applicant]
US 20200341452A1 · Bi et al. · 2020 [cited by applicant]
US 20200391447A1 · Davis et al. · 2020 [cited by applicant]
US 20210170676A1 · Kitchen · 2021 [cited by examiner]
US 20220062997A1 · Liu · 2022 [cited by examiner]
US 20230415415A1 · Milshtein · 2023 [cited by examiner]
CA 3064593A1 · 2018 [cited by applicant]
EP 3646968A1 · 2020 [cited by applicant]
WO 2015120200A1 · 2015 [cited by applicant]
WO 2019055576A1 · 2019 [cited by applicant]
WO 2020251550A1 · 2020 [cited by applicant]
Chinese Office Action and Search Report for Application No. 202180094827.1 dated Aug. 28, 2025, 20 pages with English translation. [cited by applicant]
International Search Report for International Application No. PCT/US2021/033906, mailed Dec. 1, 2021, 04 pages. [cited by applicant]
International Written Opinion for International Application No. PCT/US2021/033906, mailed Dec. 1, 2021, 05 pages. [cited by applicant]
Canadian Office Action for Application No. 3,209,921 date Mar. 6, 2026, 4 pages. [cited by applicant]
CN Office Action, Including Search report received for Chinese Patent Application No. 202180094827.1, mailed on Mar. 31, 2026, 18 pages (10 pages of English Translation and 8 pages of Original Document). [cited by applicant]