Industrial-Automation

Machine Learning for Hotspots Detection in Live Electrical Structures

Client Overview

US based leading energy company delivering electric, gas and steam service to 10M+ people.

Business Need

  • The client was looking for a technology partner to build an intelligent inspection solution for their 250K+ underground electric structures to reduce the cost involved and dependency on skilled workforce to perform the asset visual inspection

VOLANSYS Contribution

  • The solution consists of a thermal camera to capture installation images, FLIR Image Extractor to obtain temperature, location and other metadata, a cloud for images storage, a cloud compute instance with a GPU, ML model to detect cables, a module to make decision to raise the warning flag or send notification
  • Annotation of 25K+ images using makesense.ai tool providing polygonal annotation in PVOC annotation format
  • The images will be used to train the ML model (Mask RCNN model) that is based on ResNet 101 architecture
  • Mask is extracted based on the detection of electrical asset associated that helps to get the relevant temperature from the matrix
  • Hotspot is then detected based on the ingenious algorithm

Solution Diagram

EXT_300

Benefits Delivered

  • Client saved 60% on the physical inspection manpower cost
Similar Success Stories