Transforming Battery Manufacturing with AI Inspection & Data Automation
From Manual Inspections to Smart Factories: The Evolution of Battery Production
Chloe Woo | Content Strategist
With the rapid expansion of the electric vehicle (EV) market, the Lithium-Ion battery industry has emerged as a high-growth sector. However, South Korea’s journey in rechargeable lithium-ion battery manufacturing dates back to the 1990s. In such long-established industries, it’s common to see old and new technologies coexist—after all, replacing everything at once is rarely feasible.
Early Challenges: Manual Inspection in EV Battery Production
In the early days of EV battery manufacturing, quality inspection relied heavily on manual sampling and visual inspection, much like traditional small-scale battery production.
Later, machine vision inspection systems were introduced, automating quality checks and generating digital images of battery components. However, despite the adoption of machine vision technology, data collection remained largely manual.
In 2020, when AHHA Labs began working with the client, data collection was manual—operators used USB drives to extract data from equipment PCs and inspection machines, a labor-intensive and error-prone process. Image Credit: AHHA Labs
The Data Bottleneck: Manual USB Transfers in a Digital Age
By the time AHHA Labs first engaged with the client in 2020, digital data collection was still a manual process. Operators had to physically extract data from equipment PCs and inspection machines using USB drives—a labor-intensive and error-prone task.
Step 1: Automating Data Collection, Visualization & Storage
Automating Data Capture at Scale
With the global EV market on the brink of exponential growth, digital transformation was no longer an option—it was a necessity. AHHA Labs deployed Data CAMP, its web-based data integration and management solution, to automate data collection in key assembly processes.
Automated Data Collection with Data CAMP
Data CAMP enabled seamless, real-time data collection from all connected equipment and inspection machines, eliminating manual downloads. Image Credit: AHHA Labs
(1) A Web-Based Breakthrough
Unlike conventional on-premise data systems, Data CAMP operates on a web-based architecture, enabling remote monitoring and data access. In 2020, web-based data management was still a revolutionary concept in manufacturing, earning Data CAMP recognition as a game-changer from the outset.
With Data CAMP, managers no longer needed to physically retrieve data from factory equipment—all data became accessible from a single screen, in real time.
(2) No-Code Flexibility for Seamless Expansion
Designed with a no-code recipe-based structure, Data CAMP allows users to effortlessly expand data collection and processing functions without relying on external developers. Previously, data collection solutions often became obsolete due to vendor dependency and maintenance challenges.
By eliminating these inefficiencies, Data CAMP enabled the client to shift focus to higher-value tasks, boosting productivity across the factory floor.
Impact: 99.9% Automated Data Capture in High-Volume Production
Today, Data CAMP continues to operate stably in full-scale production. In a factory producing 400,000 batteries per day, the system has achieved an automated inspection image collection rate of 99.9%—a remarkable milestone in digital manufacturing.
Step 2: Connecting Data for Full Traceability
Enhancing Quality Control Through Production Traceability
As the EV market surged, battery-related safety incidents—including fires—became a critical concern. Stricter quality control became a top priority, driving the need for comprehensive production history tracking.
Initially, Data CAMP was deployed for select assembly processes and stored data in a standalone database. However, the client now needed a fully integrated system that could track data across all production stages and connect with higher-level systems.
The Solution: Scalable, End-to-End Data Integration
- Expanding Data Scope: Data CAMP extended data collection across the entire battery manufacturing process—from assembly to activation.
- Seamless Data Matching: The system automatically linked production barcodes (from manufacturing equipment) with inspection images—a task previously impossible due to disconnected data sources.
- Enterprise-Wide Data Integration: The system automatically preprocessed and transmitted collected data in formats compatible with Manufacturing Execution Systems (MES) and other enterprise platforms.
Seamless Production Traceability with Data CAMP
By linking product barcodes with inspection images, the system achieved 99.9% data matching accuracy, enabling full traceability and defect root-cause analysis. Image Credit: AHHA Labs
Impact: 99.9% Accuracy in Data Matching & Defect Root-Cause Tracking
With Data CAMP’s automated traceability system, the client can now instantly retrieve the production history of any battery. Defective units can be traced back to their exact production conditions, identifying root causes in real time.
Step 3: AI-Powered Insights & Process Optimization
Beyond Monitoring: Driving Intelligence with AI
With an integrated data infrastructure in place, the next step was to leverage AI-driven insights for smarter decision-making and real-time process control.
Advanced Analytics with Data CAMP
From Statistical Process Control (SPC) to false-positive (Type I error) detection, AI-powered analytics enhance quality monitoring and reduce manual interventions. Image Credit: AHHA Labs
Q. Can the system detect abnormal process trends and send warning alerts?
→ Statistical Process Control (SPC) is a management technique that monitors process variability using data, allowing preventive action before issues arise. Data CAMP includes a comprehensive set of SPC tools for real-time monitoring.
For instance, control charts (I Chart, MR Chart, I-MR Chart) visualize process data trends over time. These charts include a central line (CL) representing the mean, an upper control limit (UCL), and a lower control limit (LCL)—all determined based on historical data.
With this system, the client can immediately assess whether the process is stable or deviating into an uncontrolled state. If Data CAMP detects an unexpected trend in process variations, it automatically sends an alert, enabling proactive intervention before unplanned downtime occurs.
Q. Our rule-based inspection software is unreliable, so we conduct secondary visual inspections manually. If a product is misclassified as defective, we record it separately in an Excel file. Can this be automated?
→ We developed and integrated an anomaly detection AI model into Data CAMP to identify false positives (Type I errors). This significantly reduces re-inspection time while improving accuracy.
Additionally, false positive rates and actual defect statistics are visualized using pie charts and Pareto diagrams, allowing the real-time monitoring of overall quality status.
Q. Can we monitor the quality of inspection images in real-time?
→ We developed and integrated a Deep Learning-based Data Quality Index (DQI) model into Data CAMP to evaluate whether inspection images are captured correctly.
DQI analyzes multiple factors, such as image brightness and focus, then calculates a DQI score, which is displayed on the Data CAMP dashboard.
If the DQI score exceeds a predefined threshold, meaning the inspection image quality degrades compared to standard reference images, the system sends an alert.
This ensures rapid intervention before prolonged inspection failures occur. Additionally, early detection of aging or failing inspection equipment enables predictive maintenance, preventing unexpected downtime.
Q. Can we remotely control factory equipment based on data-driven decisions from the dashboard?
→ Yes, Data CAMP enables remote factory control based on real-time data monitoring and analysis.
-
When anomalies are detected, the system triggers warning alerts and gains control over PLCs (Programmable Logic Controllers).
-
Data CAMP is designed with a micro-architecture, allowing for seamless scalability and rapid expansion of data collection and control functionalities.
Compared to traditional software development, deploying similar features with Data CAMP is 75% faster.
Initially, we implemented remote control for Lithium-ion battery assembly equipment, achieving:
-
A 67% reduction in on-site monitoring personnel, improving operational efficiency.
-
A 300% increase in monitoring effectiveness
Looking ahead, we plan to standardize and record engineers’ on-site intervention strategies, allowing AI models to learn and execute fully autonomous control operations in the future.
Step 4: Predictive Maintenance with Sensor Data
Shifting from Reactive to Predictive Maintenance
AHHA Labs expanded Data CAMP’s capabilities to incorporate real-time sensor data collection for predictive maintenance.
Predictive Analytics with Data CAMP
By combining sensor data (e.g., vibration, temperature, humidity) with historical failure events, AI models can predict machine breakdowns before they occur. Image Credit: AHHA Labs
(1) What Does Measurement Data Tell Us?
Measurement data from sensors provides key indicators of equipment status and production conditions. For example, vibration data can reveal whether a machine is operating normally or experiencing anomalies. Similarly, temperature, humidity, and pressure data help determine whether the production environment is optimal for manufacturing.
However, raw measurement data alone only provides insight into the current state—it does not predict future issues.
[Reference] Key Measurement Data in Secondary Battery Production
- Voltage: Monitors cell charging/discharging status, including initial voltage and voltage variations.
- Current: Measures current flow during charging and discharging to assess cell efficiency and performance.
- Resistance: Detects internal resistance levels, indicating cell quality and potential defects.
- Temperature: Monitors internal and external battery temperatures to prevent overheating and maintain stability.
- Humidity: Controls environmental humidity in assembly processes to prevent moisture infiltration, which can degrade battery performance.
- Pressure: Ensures consistent pressure during cell assembly and other key manufacturing steps.
- Weight: Verifies charge levels and material loading accuracy.
- Charge/Discharge Time: Tracks charge/discharge duration to evaluate battery efficiency.
- Cell Thickness & Dimensions: Measures cell dimensions to ensure consistent product quality.
- Electrolyte Injection Volume: Ensures the correct amount of electrolyte is injected.
- Defect Inspection Data: Includes X-ray, CT scan results, and other quality inspection data for detecting electrode and insulation defects.
(2) The Need for ‘Event’ Data
To predict equipment failures, it is crucial to record past event data alongside measurement data.
For example:
- If vibration data exceeds a certain threshold before equipment failure, the failure event itself becomes a critical data point.
- If specific temperature, humidity, or pressure conditions lead to defects in the final product, those defect occurrences become event data.
By combining event data with measurement data, AI models can learn patterns and make predictions based on past occurrences.
However, many manufacturing sites still rely on manual event logging, which is time-consuming and prone to errors.
Because our battery client has already established systematic data collection with Data CAMP, they can now automate event data recording and seamlessly integrate various data sources for more reliable predictions.
(3) Expected Benefits
✔ Prevent Equipment Failures in Advance
By identifying patterns in vibration and sensor data before failures occur, potential breakdowns can be predicted. This minimizes unexpected downtime and maximizes production efficiency.
✔ Reduce Maintenance Costs
Predictive analytics allows necessary components to be prepared in advance. Instead of reactive maintenance, targeted preventive maintenance can be scheduled, significantly reducing costs. This approach eliminates unnecessary inspections while focusing on critical maintenance needs.
✔ Data-Driven Decision Making
Analyzing historical failure trends and key influencing factors enables strategic optimization of equipment operation. This leads to better resource allocation and improved production efficiency.
Step 5: AI-Driven Autonomous Manufacturing with MLOps
The Future of DX: AI-Driven Autonomous Manufacturing
The ultimate goal of digital transformation (DX) is not just data collection and management but the implementation of AI-powered autonomous manufacturing systems. To achieve this, end-to-end MLOps (Machine Learning Operations) is an essential tool.
Data CAMP serves as the foundation for this transformation.
(1) High-Quality Dataset Creation
The performance of AI models depends on the quality of the input data. With Data CAMP, our client systematically collects and integrates vast amounts of manufacturing data, enabling the creation of high-quality datasets that power AI-driven operations.
(2) AI Training, Deployment, and Inference
With these datasets, various AI models can be trained and deployed. Once the trained models are installed on factory PCs, they can perform a range of tasks.
For example, our client is already utilizing Aha Labs’ AI models for:
(3) Continuous Learning to Maintain AI Performance
Manufacturing data is constantly evolving, which means AI models degrade over time if they are not regularly updated.
To advance toward autonomous manufacturing, it is critical to continuously maintain AI model reliability.
With Data CAMP, both real-time process data and AI model inference results are continuously monitored. This allows for early detection of performance degradation, enabling immediate retraining using accumulated high-quality datasets, ensuring AI models remain highly accurate and effective.
(4) Full Automation with MLOps
Automating each of these stages is the core concept of MLOps.
Aha Labs is set to officially launch DAISY, an industrial AI development and operations platform, in 2025, which will integrate end-to-end MLOps capabilities, enabling seamless AI-driven automation across manufacturing environments.
MLOps Framework for AI-Driven Manufacturing
By automating data collection, model training, and deployment, Data CAMP will enable fully autonomous AI-driven production. Image Credit: AHHA Labs
AI-Powered Autonomous Manufacturing: The Future of Smart Factories
The Aha Labs team is set to further integrate and enhance Data CAMP with AI, paving the way for a fully autonomous manufacturing system. This evolution will enable real-time process data analysis and optimization, with immediate feedback applied directly to production operations.
For example, if equipment failure is predicted, the system can automatically halt production, notify managers, and suggest the optimal course of action. As AI capabilities advance, the system could simulate multiple production scenarios, identify the most efficient process conditions, and adjust operations in real time.
Ultimately, AI-driven autonomous manufacturing will:
✔ Boost productivity
✔ Reduce defect rates
✔ Optimize energy efficiency
This transformation will lead to a sustainable and intelligent manufacturing environment.
Data CAMP is already laying the foundation for this future, powering global manufacturing sites today. Join us in moving beyond data and stepping into the future of AI-driven autonomous manufacturing.