Machine Learning System and Data Fusion for Optimization of Deployment Conditions for Detection of Corrosion Under Insulation

• US 20200074616A1
• Filed: 01/23/2019
• Published: 03/05/2020
• Est. Priority Date: 08/30/2018
• Status: Active Grant

• Current Assignee: Avitas Systems, Inc. (GE Aerospace), Saudi Arabian Oil Company (Government of Saudi Arabia)
• Original Assignee: Avitas Systems, Inc. (GE Aerospace), Saudi Arabian Oil Company (Government of Saudi Arabia)
• Inventors: Al Shehri, Ali, Lim, Ser Nam, Amer, Ayman, Uzunbas, Mustafa, Aldabbagh, Ahmad, Ababtain, Muhammad, Cunningham, Vincent, Boot, John, Chan, Godine Kok Yan
• Granted Patent: US 10,643,324 B2
• Time in Patent Office: Days
• CPC Class Codes: G01N 17/00 Investigating resistance of materials to the weather, to corrosion, or to light
• CPC Class Codes: G01N 25/72 Investigating presence of flaws
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06T 2207/10048 Infrared image
• CPC Class Codes: G06T 2207/30136 Metal
• CPC Class Codes: G06T 7/0004 Industrial image inspection

Abstract

A system for predicting corrosion under insulation (CUI) in an infrastructure asset includes at least one infrared camera positioned to capture thermal images of the asset, at least one smart mount supporting and electrically coupled to the at least one infrared camera and including a wireless communication module, memory storage, a battery module operative to recharge the at least one infrared camera, an ambient sensor module adapted to obtain ambient condition data and a structural probe sensor to obtain CUI-related data from the asset. At least one computing device has a wireless communication module that communicates with the at least one smart mount and is configured with a machine learning algorithm that outputs a CUI prediction regarding the asset. A cloud computing platform receive and stores the received data and the prediction output and to receive verification data for updating the machine learning algorithm stored on the computing device.

Claims

1. A system for predicting and detecting of corrosion under insulation (CUI) in an infrastructure asset using machine learning and data fusion comprising:

at least one infrared camera positioned to capture thermal images of the asset;

at least one smart mount mechanically supporting and electrically coupled to the at least one infrared camera and having a position that is adjustable with respect to the asset, the at least one smart mount including;

a communication module;

memory storage adapted to store thermal image data received from the at least one camera;

a battery module operative to recharge the at least one infrared camera;

an ambient sensor module adapted to obtain ambient condition data; and

at least one additional sensor adapted to obtain CUI-related data from the asset;

at least one computing device having a processor, the computing device being configured with instructions for executing a machine learning algorithm in the processor which receives input from either the communication module or memory storage of the at least one smart mount;

a) thermal images of the asset;

b) the ambient condition data;

c) CUI-related data from the at least one additional sensor providing data fusion; and

d) result data obtained from conventional CUI tests, and which outputs a CUI prediction regarding the asset; and

a machine-learning platform adapted to train the machine learning algorithm based on accumulated input data received from the at least one computing device over time.

2. The system of claim 1, wherein the at least one computing device is adapted to test current deployment conditions of the at least one infrared camera and at least one smart mount by executing the machine learning algorithm using current infrared images, ambient sensor data and CUI-related data by determining whether performance of the machine learning algorithm under current deployment conditions meets a threshold level.

3. The system of claim 2, wherein the performance of the machine learning algorithm conditions is measured by corrosion prediction accuracy.

4. The system of claim 2, wherein the performance of the current deployment conditions is measured by a cost of deployment and testing.

5. The system of claim 2, wherein the performance of the current deployment conditions is measured by a speed of deployment and testing.

6. The system of claim 2, wherein the deployment conditions tested by the at least one computing device include a time of year in which the at least one infrared camera and at least one smart mount are deployed.

7. The system of claim 2, wherein the deployment conditions tested by the at least one computing device include a time of day in which the at least one infrared camera and at least one smart mount are deployed.

8. The system of claim 2, wherein the deployment conditions tested by the at least one computing device include a position, orientation and image capture duration of the at least one camera.

9. The system of claim 2, wherein the deployment conditions tested by the at least one computing device include a position, type and number of images captured per time period of the at least one camera.

10. A method of optimizing deployment conditions of at least one investigative kit for obtaining data from an infrastructure asset to enable prediction and detection of corrosion-under-insulation (CUI) comprising:

receiving thermal images, sensor data acquired from asset over time acquired by the at least one investigative kit, and deployment condition information regarding the at least one investigative kit;

obtaining result data of conventional CUI detection methods corresponding of the asset over time;

training a machine learning algorithm using the captured thermal images and other sensor data and the result data acquired from the asset; and

testing the deployment conditions of the at least one investigative kit by;

receiving current test conditions including ambient condition data and current deployment conditions at the at least one investigative kit to obtain ambient condition data over time;

executing the machine learning algorithm using the ambient conditions data and current deployment conditions of the at least one investigative kit;

determining whether a performance of the machine learning algorithm based on the test conditions exceeds a preset performance threshold;

if performance of the machine learning algorithm exceeds the threshold, optimizing the deployment conditions using the test conditions;

if performance of the machine learning algorithm does not exceed the threshold, changing the deployment conditions of the at least one investigative kit; and

iteratively repeating execution of the machine learning algorithm until performance exceeds the threshold or a set number of repetitions have been conducted.

11. The method of claim 10, further comprising measuring the performance of the machine learning algorithm based on corrosion prediction accuracy using current test conditions.

12. The method of claim 10, further comprising measuring the performance of the machine learning algorithm based on a cost of deployment and testing.

13. The method of claim 10, further comprising measuring the performance of the machine learning algorithm based on a speed of deployment and testing.

14. The method of claim 10, wherein the current deployment conditions include a time of year at which the at least one investigative kit is deployed.

15. The method of claim 10, wherein the current deployment conditions include a time of day at which the at least one investigative kit is deployed.

16. The method of claim 10, wherein the current deployment conditions include tested by the at least one computing device include a position, orientation and image capture duration of the at least one investigative kit.

17. The method of claim 10, wherein the current deployment conditions include tested by the at least one computing device include a position, type and number of images capture per time period of the at least one investigative kit.