welding-2819144_1920 1.png

MONITECH

AI-based welding monitoring system for welding quality of EV batteries

Development time

7 Month

Manpower

4 Professionals

The Brief

The demand for quality information of various welding methods is increasing in line with the trend of new material welding, the increase of micro and ultra-precision products, and the increasing demand for eco-friendly and energy-saving welding processes. In particular, the MONITECH team wanted to implement AI technology, to establish a high-precision welding monitoring system for guaranteed laser welding quality of lithium-ion batteries, which are exploding in demand along with the rapid growth of the electric vehicle market.

Service

Welding Quality monitoring

AI algorithm development

Dashboard / Visualization

흰육각형.png

One

흰육각형.png

2

흰육각형.png

3

흰육각형.png

4

흰육각형.png

5

Data Collection & Modelling

API based integration system

Maintenance

Technological Challenges

The laser/Ultrasonic welding monitoring system should deliver accurate and constant quality analysis from various power sources.

Reliable Quality Monitoring

Application of deep learning modeling by combining Monitech’s knowhow on exiting quality monitoring methods and Ellexi’s Confusion Matrix analysis on various sensor data, such as UV, IR, light, high-speed thermal image, to secure high defect detecting precision and minimize false alarms.

Group 111.png

Multi-sensor compatibility

Overcome limitations of statistical and wavelength data analysis on existing multi-sensor (optical rapid + thermal image, vision + rapid thermal image etc.) configuration-based quality monitoring system by applying deep learning technology.

Platform with Scalability

Provide modulation of AI models by its welding method (arc, spot, ultrasonic, laser etc.) and RestAPIs for interoperability.

Road Map

M+1 / M+2

  • Time series Data Modeling

M+5 / M+6

  • Platform system configuration

small3.png
small1.png
small2.png
small4.png

M

  • Philo-AD Installation / Data Collection

M+3 / M+4

  • Model Verification

  • Abnormal behavior visualization

  • Security Risk behavior adaptive learning

Key Features

  • Pattern Monitoring (weld quality monitoring from UV/IR signals)

  • Detection of weld defects based on changes in process parameter.

  • Process Visualization

  • AI weld quality judgement algorithm (reference adjustments)

  • Real-time laser welding quality management in mass production line

Group 134.png

The Result

Welding Result Recall above 90%

Achieved 99% recall when applied to the lead welding process of BMA (Battery Module Assembly) line

Judgement within 3 seconds after receiving welding results

It took less than a second from data measurement to derivation of quality monitoring result

API integration

AI modules fully integrated with EV battery laser welding quality monitoring system

Fast Grasp of welding quality judgement though visualization

Group 123.png
Group 124.png