Page 509 - AI for Good Innovate for Impact
P. 509
AI for Good Innovate for Impact
Use Case - 5: Explosion-Proof Intelligent Inspection Robot Solution 4.5: Manufacturing
Organization: Sevnce Robotics Co., Ltd.
Contact person: - Chengcheng Wang; wangchengcheng@ sevnce .com
Country: China
1 Use Case Summary Table
Item Details
Category Manufacturing
Problem Addressed In scenarios such as traditional manual inspection of power facili-
ties and industrial equipment, there are issues like low efficiency,
high safety risks, and incomplete data collection. Existing inspec-
tion robots have insufficient intelligence and struggle to adapt to
complex environments and handle abnormal situations autono-
mously.
Key Aspects of Solution Adopt embodied intelligence technology and combine multimodal
sensors (such as infrared thermal imagers, high - definition cameras,
gas detectors, etc.) to achieve environmental and equipment status
perception. Utilize deep learning and reinforcement learning algo-
rithms to endow robots with autonomous path planning, abnormal
identification, and emergency response capabilities. Through edge
computing and cloud - based data analysis, realize real - time moni-
toring of equipment status and predictive maintenance.
Technology Keywords Embodied intelligence, multimodal sensors, deep learning, rein-
forcement learning, edge computing, predictive maintenance
Data Availability Private. Data is sourced from sensor data collection in actual inspec-
tion scenarios, historical equipment operation data, etc.
Metadata (Type of Data) Visual data (equipment appearance images), thermal imaging data
(equipment temperature distribution), gas concentration data, vibra-
tion data, sound data
Model Training and Based on a large number of equipment images and operation data,
Fine-Tuning use deep learning models to train for abnormal feature identifica-
tion. Optimize the robot's inspection path planning and decision
- making model in complex environments through reinforcement
learning.
Testbeds or Pilot Deploy- Pilot projects were carried out in a large - scale substation and a
ments chemical plant. Link:
473
