Correlating upstream preform parameters with downstream draw-process yield, to optimize optical fiber manufacturing.
A Process-Data Analytics Platform for Optical Fiber & Preform Manufacturers. Aeologic built an integrated analytics platform that connects preform production parameters with downstream draw-process outcomes, enabling manufacturers to identify the root causes of yield loss, reduce scrap and rework, and optimize manufacturing efficiency.
In short
Aeologic built an integrated process-data analytics platform for optical fiber and preform manufacturing that connects upstream preform production data with downstream draw-process telemetry and quality outcomes. By correlating process variables with defect and yield rates, the platform helps engineers identify root causes faster, intervene before yield loss escalates, and create a scalable foundation for predictive optimization.
- Industry Telecom / Optical Fiber Manufacturing — Industrial Manufacturing
- Client Enterprise Manufacturer — Optical Fiber Preform & Draw Production Facility
- Solution Integrated analytics connecting preform parameters with draw-process yield and defects
- Deployment Edge-based plant data capture with cloud-hosted analytics and dashboards
Yield loss could start upstream, but preform and draw operations were analyzed separately.
Optical fiber manufacturing is a two-stage process: a glass preform is first produced through vapor deposition and sintering (MCVD, OVD, or VAD), and that preform is then drawn into kilometers of finished fiber on a draw tower. Yield loss can originate at either stage, but the two stages were typically run, monitored, and analyzed in isolation. Preform process data lived in one system, draw-tower telemetry in another, and quality outcomes were recorded separately in lab or inspection logs. When defects such as bubbles, inclusions, geometry deviation, attenuation spikes, or fiber breaks appeared downstream, engineers had no systematic way to trace them back to the specific preform batch or upstream process conditions that caused them. Root-cause analysis was manual, reactive, and time-consuming, while yield fluctuated from batch to batch without a clear, data-backed explanation. The client therefore needed an analytics capability that connected preform-stage parameters to draw-stage outcomes and turned that connection into actionable, real-time process intelligence.
-
01
Preform production data, draw-tower telemetry, and quality records stored in separate systems.
-
02
No systematic connection between upstream process conditions and downstream yield outcomes.
-
03
Defects such as bubbles, inclusions, geometry deviation, attenuation spikes, and fiber breaks required manual investigation.
-
04
Batch-to-batch yield fluctuation lacked a clear, data-backed explanation.
What the analytics platform had to achieve.
Build a unified data pipeline connecting preform production and draw-process telemetry.
Correlate preform-stage process variables with downstream draw yield and defect rates.
Identify the root causes of yield loss and enable early, preventive intervention.
Give process engineers and plant managers real-time visibility through role-based dashboards.
Introduce statistical process control (SPC) with automated alerts on parameter drift.
Establish a scalable analytics foundation for future predictive-maintenance and AI-driven process optimization.
A connected process-data analytics layer linking preform conditions to downstream fiber yield and quality.
Unified process data pipeline
Data from MCVD, OVD, VAD deposition systems, sintering furnaces, draw towers, sensors, and quality records is brought into a unified time-series data historian.
Cross-stage analytics engine
Statistical and ML-based models correlate upstream preform variables with draw-stage process conditions, defects, and yield outcomes.
Real-time process intelligence
Engineers and plant managers receive actionable dashboards, SPC alerts, and batch-level traceability to intervene earlier and improve manufacturing performance.
Defect and yield analytics
Automatically classifies bubbles, inclusions, geometry deviation, attenuation spikes, and fiber breaks, then traces each outcome to its originating preform batch and implicated process parameters.
Real-time monitoring dashboards
Live role-based views surface yield trends, defect Pareto charts, and parameter behavior for process engineers and plant managers.
SPC-based alerting
Automated control-limit alerts identify parameter drift before it results in yield loss, shifting intervention from reactive investigation toward preventive action.
Parameter-level correlation
The analytics engine evaluates deposition rate, core/clad ratio, dopant concentration, soot density, and sintering temperature against downstream draw conditions and quality outcomes.
End-to-end batch traceability
Every finished fiber spool can be traced back to its originating preform batch and the exact process conditions under which it was produced.
Scalable analytics foundation
Edge-based plant capture and cloud-hosted analytics provide a modular foundation that can extend across production lines and plants into predictive maintenance and AI-driven process optimization.
Five specific process-data problems, five connected fixes.
Disconnected data sources across preform and draw stages
Preform production data, draw-tower telemetry, and quality records were maintained in separate systems, making cross-stage analysis difficult.
Unified time-series data pipeline
We integrated PLC/SCADA data, sensor telemetry, and manual quality records into a single time-series historian.
No visibility into which upstream parameters cause downstream defects
Engineers lacked a systematic way to connect preform process conditions with downstream draw yield and defect outcomes.
Statistical and ML correlation engine
We developed correlation models linking upstream preform variables with draw-stage yield and defect outcomes.
Reactive, manual root-cause analysis
Yield issues were investigated after defects appeared, requiring time-consuming review of historical process records.
SPC-based preventive alerting
Automated control-limit alerts notify engineers when process parameters drift toward conditions associated with yield loss.
Lack of batch-level traceability
Quality events could not be consistently traced from finished fiber back through the draw process to the originating preform batch.
End-to-end batch tracking
We created traceability linking every finished fiber spool to its originating preform batch and production conditions.
Limited cross-functional visibility
Process engineers, plant managers, and quality teams needed different operational views to make timely decisions.
Role-based analytics dashboards
We delivered dashboards tailored to engineering, plant-management, and quality workflows while using the same underlying process-data foundation.
"The key shift was connecting the two production stages. Instead of treating preform and draw operations as independent processes, the platform created a continuous analytical relationship between upstream parameters, downstream outcomes, and finished-fiber quality."
From isolated production data to connected yield intelligence.
For process engineers. Root-cause analysis moves from hours of manual log review to minutes of automated correlation, backed by clear defect-to-parameter traceability.
For plant managers. Real-time visibility into yield performance across production lines, with early warning before defects escalate into scrapped batches.
For quality teams. Full batch traceability from preform to finished fiber, supporting audits, customer quality claims, and continuous improvement initiatives.
For business operations. Reduced material waste and scrap costs, improved overall equipment effectiveness (OEE), and a scalable data foundation for future AI-driven process optimization and predictive maintenance.
Connecting preform intelligence with draw-process performance.
Preform & Process Yield Analytics moves optical fiber manufacturers beyond reactive, batch-by-batch quality control toward a proactive, data-driven production process. By correlating preform-stage production parameters with downstream draw-process outcomes, the platform helps engineers identify the root causes of yield loss faster, reduce scrap and rework, and build the process-data foundation needed for the next stage of AI-driven predictive optimization. Its modular, scalable architecture is designed to extend across additional production lines and plants as the manufacturer scales.
Common questions about this analytics deployment.
Find quick answers to common questions about connecting preform production data with downstream optical fiber draw-process yield.
How does the platform connect preform and draw-process data?
The platform integrates data from preform deposition systems such as MCVD, OVD, and VAD, sintering furnaces, draw towers, sensors, PLC/SCADA systems, and quality records into a unified time-series data historian. This creates a common data foundation for correlating upstream process conditions with downstream yield and defect outcomes.
Which preform parameters can be correlated with draw-stage outcomes?
The correlation engine can analyze variables including deposition rate, core/clad ratio, dopant concentration, soot density, and sintering temperature against draw-stage factors such as draw speed, tension, furnace temperature, fiber diameter, coating concentricity, yield, and defect rates.
How does the system help identify the root cause of yield loss?
The statistical and ML-based correlation engine links downstream defects and yield changes to upstream process variables and originating preform batches. Engineers can move from manual log review toward automated defect-to-parameter correlation and investigate the process conditions most strongly associated with a given outcome.
What types of defects and quality outcomes can be tracked?
The platform tracks defect and quality outcomes including bubbles, inclusions, geometry deviation, attenuation spikes, and fiber breaks. These outcomes can be associated with their originating preform batch and the process parameters implicated in the observed yield loss.
Can the platform support multiple production lines and plants?
Yes. The deployment uses edge-based data capture at the plant with a cloud-hosted analytics, correlation, and dashboarding layer. Its modular architecture provides a scalable foundation for additional production lines, plants, predictive-maintenance workflows, and AI-driven process optimization.
Ready to connect upstream process data with downstream yield?
Our architects can help map your preform, draw-tower, quality, and plant-data landscape into a connected analytics platform — starting with a practical process-data foundation.
Book a Workshop → Explore Manufacturing Solutions →