System control based on acoustic and image signals

Patent metadata

Field	Value
Publication number	US-11092983-B2
Application number	US-201916441375-A
Country	US
Kind code	B2
Filing date	Jun 14, 2019
Priority date	Jun 18, 2018
Publication date	Aug 17, 2021
Grant date	Aug 17, 2021

Abstract

An example system includes at least one acoustic sensor and one optical sensor to monitor a thermal spray system controlled by a plurality of control parameters and performing a process associated with a plurality of process outputs. The system includes a computing device including a machine learning module and a control module. The machine learning module is configured to determine, based on at least the plurality of control parameters, an at least one time-dependent acoustic data signal, an at least one image data signal, and the plurality of process outputs, a relationship between the plurality of control parameters and the plurality of process outputs by machine learning. The control module is configured to control the thermal spray system to adjust the plurality of process outputs toward a plurality of respective operating ranges.

First claim

The invention claimed is:

1. A system comprising:
at least one acoustic sensor configured to generate at least one time-dependent acoustic data signal indicative of sound generated by a thermal spray system controlled by a plurality of control parameters and performing a process associated with a plurality of process outputs;
at least one optical sensor configured to generate at least one image data signal indicative of the thermal spray system performing the process, wherein the at least one image data signal is real-time, near-real-time, or continuous; and
a computing device comprising:
a transformation module configured to transform the at least one time-dependent acoustic data signal to a frequency-domain spectrum;
a machine learning module configured to determine, based on at least the plurality of control parameters, the frequency-domain spectrum, the at least one image data signal, and the plurality of process outputs, a relationship between the plurality of control parameters and the plurality of process outputs by machine learning; and
a control module configured to:
determine, based on the relationship determined by the machine learning module, respective values of the plurality of control parameters configured to cause the thermal spray system to generate predetermined values of the plurality of process outputs, and
control, based on the respective values of the plurality of control parameters, the thermal spray system to adjust the plurality of process outputs toward a plurality of respective operating ranges by sending a control signal to thermal spray system.

2. The system of claim 1 , wherein the control module is configured to:
select, based on the respective values of the plurality of control parameters, at least one component of the thermal spray system, wherein the at least one component is controlled by the plurality of control parameters, and
control, based on the respective values of the plurality of control parameters, the thermal spray system to adjust the plurality of process outputs toward the plurality of respective operating ranges by sending the control signal to the at least one component.

3. The system of claim 2 , wherein the system component comprises at least one of a thermal spray gun, a plasma electrode, a powder port, a gas inlet port, or a material inlet port.

4. The system of claim 1 , wherein the plurality of control parameters comprises at least one of a primary gas flow rate, a secondary gas flow rate, a gun current, a gun position, a part position, a carrier gas flow rate, a powder feed rate, a temperature, a pressure, a mass flow rate, a volumetric flow rate, a molecular flow rate, a molar flow rate, a composition, a velocity, or a concentration, or combinations thereof.

5. The system of claim 1 , wherein the plurality of process outputs comprises a coating microstructure, a coating hardness, a coating adhesion, a coating deposition rate, a coating deposition efficiency, a coating quality, a coating color, or a coating density, or combinations thereof.

6. The system of claim 1 , wherein the machine learning module is configured to determine, based on at least the plurality of control parameters, the frequency-domain spectrum, the at least one image, and the plurality of process outputs, at least one of a relationship between the plurality of control parameters and a plurality of measured process parameters or a relationship between the plurality of measured process parameters and the plurality of process outputs.

7. The system of claim 1 , wherein the machine learning module is configured to determine the relationship between the plurality of control parameters and the plurality of process outputs by at least one of comparing real-time or near-real time respective values of the plurality of process outputs with immediately prior respective values of the plurality of process outputs, comparing real-time or near-real time respective values of the plurality of process outputs with respective values at 5%, 25%, 50%, or 75% of a present run time, or comparing real-time or near-real time respective values of the plurality of process outputs with predetermined threshold values of the plurality of process outputs.

8. The system of claim 1 , wherein the computing device further comprises an output device configured to output a representation of at least one of the at least one time-dependent acoustic data signal, the frequency-domain spectrum, at least one image data signal, at least one process control parameter, at least one measured parameter, or at least one process output.

9. The system of claim 1 , wherein the process comprises at least one of spraying, gas combustion, electrical arcing, plasma generation, flow shock, powder transport, or mechanical motion.

10. The system of claim 1 , wherein the machine learning comprises at least one of linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), K-nearest neighbors, support vector machines (SVM), regression analysis, sensitivity analysis, optimization algorithm, basinhopping, Broyden-Fletcher-Goldfarb-Shanno (BFGS), fuzzy logic, artificial neural network (ANN), or gaussian processes (GP).

11. The system of claim 1 , wherein the computing system further comprises:
an acoustic data signal processing module configured to receive the at least one time-dependent acoustic data signal; and
an image data signal processing module configured to:
receive the at least one image data signal, and
transform the at least one image data signal into at least one image.

12. The system of claim 11 , wherein the acoustic data signal processing module is further configured to transform the at least one time-dependent acoustic data signal to a frequency-domain spectrum, and the machine learning module is configured to determine, based on at least the plurality of control parameters, the frequency-domain spectrum, the at least one image, and the plurality of process outputs, the relationship between the plurality of control parameters and the plurality of process outputs by machine learning.

13. A method comprising:
receiving, by a computing device, from at least one acoustic sensor, at least one time-dependent acoustic data signal indicative of sound generated by a thermal spray system controlled by a plurality of control parameters and performing a process associated with a plurality of process outputs;
receiving, by the computing device, from at least one optical sensor, at least one image data signal indicative of the thermal spray system performing the process, wherein the at least one image data signal is real-time, near-real-time or continuous;
transforming, by the computing device, the at least one time-dependent acoustic data signal to a frequency-domain spectrum;
determining, by the computing device, based on at least the plurality of control parameters, the frequency-domain spectrum, the at least one image data signal, and the plurality of process outputs, a relationship between the plurality of control parameters and the plurality of process outputs by machine learning;
determining, by the computing device, based on the relationship, respective values of the plurality of control parameters configured to cause the thermal spray system to generate predetermined values of the plurality of process outputs; and
controlling, by the computing device, based on the respective values of the plurality of control parameters, the thermal spray system to adjust the plurality of process outputs toward a plurality of respective operating ranges by sending a control signal to thermal spray system.

14. The method of claim 13 , further comprisin

Assignees

• Rolls Royce Corp

Inventors

• Cybulsky Michael
• Blair Taylor K
• Zimmerman Benjamin
• Sweet Marshall Louis
• Marcon Andrea

Classifications

• G05B2219/49065

  Execute learning mode first for determining adaptive control parameters · CPC title

• G05B13/0265

  the criterion being a learning criterion · CPC title

• G05B13/024

  in which a parameter or coefficient is automatically adjusted to optimise the performance · CPC title

• G05D23/19

  characterised by the use of electric means {(G05D23/1393 takes precedence)} · CPC title

• G05D7/0635 Primary

  by action on throttling means (G05D7/0688, G05D7/0694 take precedence) · CPC title