Industrial AI
DJI x National Energy Industrial Drone Inspection
An early industrial drone-inspection project for high-altitude confined-space equipment checks in thermal power plants.
Problem
Manual boiler inspection required long cooling downtime and exposed workers to dangerous confined-space conditions. The project needed a way to convert field experience into machine-usable inspection rules.
Workflow
- 01Map manual inspection into targets, paths, data capture, abnormality judgment, human review, and acceptance nodes.
- 02Use a cost-game scheduling and exit strategy to compare high-value generation time against low-cost expendable drone hardware.
- 03Send drones into partially cooled boilers around 100°C and restart the unit without waiting for full cooling and physical retrieval.
- 04Work with domain experts to define visual thresholds for water-wall damage and related field indicators.
Evidence
Innovation gold award
China Electric Power Technology Market Association Electric Power Industry Innovation Application Gold, 2020.08.
Recommendation catalog
Selected into the Electric Power Industry Innovation and Creativity Excellent Achievements recommendation catalog, page 226.
QC third prize
National Energy Group QC group activity result third prize, 2019.06.
Patent disclosure
National invention patent technical disclosure submitted under P1926220CN-GN.
Boundary
- This is a field project and technical-disclosure experience, not a claim of personally building all DJI hardware.
- Public proof does not expose plant-sensitive operational details.
- The role signal is industrial AI workflow and knowledge engineering, not pure robotics research.
Role Mapping
- Industrial AI: converts field inspection knowledge into thresholds and workflow logic.
- AI product / solution product: ties technology choices to downtime economics and operational constraints.
- Agent / hardware intelligence: motivates context-aware physical-world decision systems.