AI Platform
Reducing Battery Cathode Material Defects by 12% with AI Qualiy Prediction
■ The Challenge of High-Nickel Cathode Production: Manual Weighing and Late Defect Detection
High-nickel cathode materials are widely used in electric vehicle (EV) batteries for their high power output. However, they require tightly controlled coating and doping processes to manage the risk of thermal runaway. Even minor manufacturing deviations can lead to large volumes of defective finished products.
Much of the production workflow still depended on manual work. Operators transported and weighed samples by hand, creating significant data variations depending on their experience. Quality issues could only be identified after the entire manufacturing process was complete, allowing small errors during production to result in substantial wasted time and failure costs.
■ Connecting Sample Preparation Robots and Equipment Data in a Unified AI Diagnostics Platform
SYSTEM R&D and COSMO AM&T adopted MonPlant, Mondrian AI’s multidimensional data analytics platform, to move from reactive quality management to proactive process control. Mondrian AI designed a multimodal data integration environment that connected seven-axis robotic sample preparation and quality analysis equipment with manufacturing equipment through a single interface.
The platform extended automation into process decision-making. By modeling large volumes of process data in MonPlant, Mondrian AI developed a feedback control algorithm that automatically adjusted variables such as temperature and pressure.
For critical equipment prone to breakdowns, the company also applied a zero-shot learning model that analyzed subtle changes in machine behavior to predict abnormal conditions without historical failure records. This strengthened predictive maintenance and enabled earlier intervention.
■ Improving Process Consistency and Reducing Quality Costs with Multidimensional Data
MonPlant enabled proactive adjustments to maintain process consistency, reducing data variability and establishing an autonomous control environment that kept operations in an optimal state. The combined application of real-time feedback control and zero-shot predictive maintenance delivered measurable manufacturing improvements.
- Quality data collected across all process stages: 5,284 records
- Equipment-based quality prediction accuracy: 90.5%
- Final product defect rate: Reduced by 12% from the previous baseline
- Company-wide quality costs: Reduced by 20% from the previous baseline
This intelligent cathode material manufacturing environment strengthens competitiveness in the EV and energy storage system (ESS) markets through consistent, high-quality production. The multimodal quality data accumulated in MonPlant also provides a foundation for R&D, helping researchers explore next-generation battery material combinations more quickly.
