Summary of "ficonTEC - AI/ML-based Process Control"

Summary of "ficonTEC - AI/ML-based Process Control" Webinar

This webinar, part of ficonTEC’s mini-series, focuses on the implementation and application of Artificial Intelligence (AI) and Machine Learning (ML) in process control within photonics assembly, testing, and industrial manufacturing environments. The discussion covers research & development, collaborative projects, and real-world industrial use cases, highlighting how AI/ML enhances machine performance, process optimization, and predictive maintenance.

Key Technological Concepts and Applications:

AI and ML Definitions & Context
- AI broadly refers to intelligent agents acting to maximize goal achievement, but in industrial settings, ML (a subset of AI) is more relevant.
- ML involves data-driven algorithms that improve automatically through experience, focusing on building predictive models without explicit programming for each task.
- Deep Learning (a subset of ML) is used for complex tasks like image classification.
Applications at ficonTEC:
- Passive and Active Alignment of Optical Components: ML analyzes pre-alignment data (beam width, focus displacement) to predict outcomes and reduce active alignment time by about 50%, improving throughput and accuracy. ML also helps identify unexpected data clusters, enabling process refinement.
- Adaptive Motion Control: Motion parameters of assembly machines are autonomously optimized using ML to balance speed and accuracy, adapting to environmental changes, mechanical wear, or load variations. This self-optimization happens both during commissioning and throughout machine operation.
- Visual Inspection via Deep Learning: Neural networks classify images of components (optical/mechanical) to detect defects like cracks, debris, or misformed parts. The system provides confidence levels for classifications, allowing human intervention for uncertain cases.
- Predictive Maintenance: ML models monitor machine components (e.g., grippers) to predict failures before they occur, reducing unscheduled downtime and enabling scheduled repairs with precise diagnostics.
- Process Step Monitoring: ML tracks the duration of sub-process steps, identifying slowdowns that impact throughput (units per hour, UPH) and quality, allowing real-time adjustments to optimize production.
Data Acquisition and Edge Computing:
- Real-time data capture is critical; ficonTEC uses Adaptics’ Edge Ops platform for on-site (edge) data processing, avoiding cloud latency and security issues.
- The system integrates with ficonTEC’s Process Control Master (PCM) and external sensors (heat, vibration) for comprehensive monitoring.
- Dashboards provide customizable, real-time visualization of key performance indicators (yield, UPH, quality) accessible remotely via laptops or mobile devices.
Industrial Impact and Scalability:
- Proven deployment on over 75 pieces of equipment with multiple clients, transitioning from pilot projects to full production contracts.
- Applicable primarily to high-volume, low-cost production or high-cost, low-volume critical parts where quality and yield are paramount.
- Remote installation and rapid commissioning (half-day onsite setup, model training within weeks) facilitate easy adoption without disrupting production.

Reviews, Guides, and Tutorials:

The webinar serves as a guide and overview of AI/ML applications in photonics manufacturing rather than a detailed tutorial.
References to previous webinars (e.g., passive vs. active alignment) are provided for deeper understanding.
Viewers are encouraged to engage via chat/Q&A and to consult ficonTEC’s online resources (LinkedIn, blog, Vimeo, YouTube) for ongoing updates and related content.
A giveaway quiz encourages viewers to interact and learn about clean room classifications related to ficonTEC systems.

Main Speakers / Sources:

Greg (Host/Moderator) – Coordinates the webinar, introduces topics, and facilitates discussion.
Dr. Colin Dankvart (ficonTEC R&D Specialist) – Provides technical explanations on AI/ML concepts, passive/active alignment, adaptive motion control, and Deep Learning-based inspection.
Dave Pritz (Adaptics, Industrial Partner, USA) – Presents on real-time data acquisition, edge computing platform (Edge Ops), predictive maintenance, process monitoring, and industrial deployment experiences.

Conclusion:

The webinar highlights ficonTEC’s advanced use of AI and ML to improve photonics manufacturing efficiency, accuracy, and uptime through data-driven process control. The collaboration with Adaptics enables real-time, edge-based analytics and autonomous machine optimization. The technology is scalable, proven in industrial settings, and adaptable to diverse production challenges. Interested parties are invited to contact ficonTEC for tailored AI/ML solutions.