About Swire Coca-Cola
Swire Coca-Cola is a key bottling partner in the global Coca-Cola system. Operating 41 modern production bases across China, Southeast Asia, and the U.S. Midwest, the company manufactures and distributes over 60 beverage brands — including Coca-Cola, Sprite, and Minute Maid — serving more than 940 million consumers.
Project Background: Moving Beyond Legacy Maintenance
In fully automated beverage bottling lines, pneumatic and electric valves serve as critical execution components controlling the delivery of water, syrup, carbon dioxide, and finished liquid products. The responsiveness, motion stability, and repeatability of these valves directly dictate line speed and product consistency.
Historically, maintenance relied on two conventional approaches:
- Scheduled Preventive Maintenance: Servicing equipment at fixed time intervals, which frequently led to either over-maintenance or under-maintenance.
- Run-to-Failure Repair: Addressing issues only after a failure occurred, resulting in unplanned downtime and product wastage.
To overcome these limitations, Swire Coca-Cola sought to build an intelligent valve predictive maintenance system capable of real-time state perception, anomaly trend identification, and data-driven maintenance decision support.
The Challenge: The Bottlenecks of Legacy Infrastructure
Swire Coca-Cola had previously deployed Kepware as its underlying industrial data collection and protocol conversion platform to report valve PLC data and provide basic visualization. However, during the implementation of the predictive maintenance system, critical operational and architectural hurdles emerged:
- Platform Stability: Windows vs. Linux Reliability
Kepware relies on a Windows-based architecture, bringing inherent operational risks such as mandatory OS patch reboots, and high system resource overhead. This made it ill-suited for 7×24 continuous line operation demands and difficult to deploy onto lightweight edge gateways. - Closed Architecture and Limited Extensibility
As a conventional protocol conversion gateway, Kepware relies entirely on vendor updates for custom interfaces and lacks support for plugin-based algorithm integration. It cannot execute edge intelligence logic (such as predictive modeling or local anomaly detection) at the data acquisition layer. Forcing all heavy computations to upper-level systems resulted in high latency and an outdated overall architecture. - Misalignment with Corporate Technology Strategy
While Swire Coca-Cola's development team had established containerization as a core technology stack priority, Kepware's native Windows framework could not support containerized deployment. Operating parallel modern and legacy tech stacks increased O&M complexity and constrained the company's broader IT/OT convergence strategy.
The Solution: Architecting Edge Intelligence with EMQX Neuron
To address these challenges, Swire Coca-Cola introduced EMQX Neuron, an industrial edge intelligence platform, as the core data ingestion and edge processing node for its valve predictive maintenance system.
Unified Multi-Protocol Access & Full-Link Ingestion
EMQX Neuron utilizes built-in industrial protocol drivers to seamlessly interface with PLCs and valve control systems. It standardizes and unifies data collection across disparate sources, gathering key operational metrics including valve open/close status, actuation commands and feedback signals, cylinder pressure and current curves, and cycle execution counts and timings, effectively solving data fragmentation.
Real-Time Edge Computing & Feature Extraction
Leveraging Neuron's streaming SQL engine, real-time data cleansing and feature extraction take place directly on edge devices without needing prior cloud transmission. It performs local valve response time profiling, actuation delay analysis, pressure/current curve feature extraction, and runtime fluctuation detection at the edge.
Edge Model Deployment & On-Device Inference
Lightweight valve health evaluation models — trained in the cloud — are deployed onto EMQX Neuron edge nodes. This enables real-time local health scoring, failure probability calculation, and anomaly trend identification right on the factory floor, extending predictive intelligence from centralized cloud processing to real-time shop-floor decision-making.
Standardized Cloud Synchronization & System Integration
Using standard MQTT protocols, EMQX Neuron reliably streams cleansed, high-value structured data from the edge to EMQX Enterprise From there, data is dispatched simultaneously to time-series databases and visual O&M platforms, creating a complete "Ingest – Compute – Infer – Visualize" data loop to power upper-level maintenance decisions and traceability.
Business Impact & Key Results
By deploying an edge-intelligent architecture powered by EMQX Neuron, Swire Coca-Cola achieved significant operational improvements:
- Enhanced Equipment Reliability: Continuous state monitoring and early anomaly alerts effectively reduced unplanned line shutdowns caused by sudden valve failures, safeguarding continuous, stable, and high-speed bottling operations.
- Upgraded O&M Strategy: The system replaced fixed-interval maintenance with data-driven predictive maintenance, eliminating unnecessary routine overhauls, reducing resource waste, and shifting O&M operations from experience-based to data-centric workflows.
- Improved Product Consistency: Fine-grained monitoring and control over valve actuation precision, response delay, and operational status mitigated quality fluctuation risks during the filling process.
- Faster Incident Response: Real-time diagnostics at the edge let the team pinpoint root causes and potential failure points. Maintenance shifted from post-failure firefighting to proactive intervention, drastically improving troubleshooting efficiency.
- A Foundation to Build On: With the approach validated on valves, Swire Coca-Cola has a proven foundation to extend predictive maintenance to other critical assets, such as pumps and motors.
Conclusion
Swire Coca-Cola now runs unified data ingestion, edge stream computing, and local AI inference directly on the factory floor. Working with EMQX Enterprise, time-series databases, and visualization platforms, the system delivers real-time monitoring, early anomaly warnings, and predictive maintenance across Swire Coca-Cola’s production line valves.