Patent · US Active

Metal loss determinations based on thermography machine learning approach for insulated structures

US11112349B2 · kind B2 · utility

1Cited by
27References
21Claims
0Family size

Assignee

Inventors

Key dates

Filing dateJul 16, 2019
Grant dateSep 7, 2021
Priority date
Expiry dateFeb 1, 2040

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06N3/0418
  • WIPO fieldMeasurement
  • WIPO sectorInstruments

Abstract

A method for identifying metal wall loss in an insulated metal structure is provided. The method includes receiving thermograms of the outer surface of the structure using an infrared camera, applying filters to the thermograms using a first machine learning (ML) system, determining wall loss classifications based on outputs from the filters, validating the wall loss classifications by inspecting the structure, and training the first ML system using the validation results. Outputs of the first ML system and additional structural and environmental data are input to a second ML system that incorporates information from earlier states into current states. The second ML system is trained to estimate wall loss according to changes in the outputs of the first ML system and the additional data over time until a wall loss classification accuracy is reached. The metal wall loss is thereafter estimated using the first and second ML systems in coordination.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.