AI-Based Quality Inspection System

I designed and implemented a machine learning–based inspection system to detect missing or improperly placed overmolded inserts during an automated loading process. The solution integrated directly with existing PLC-controlled equipment and required no major changes to the underlying automation hardware. This was prior to Keyence's IV-3 sensor series.

When signaled by the robot, an image is captured of the part on a conveyor. Regions of interest corresponding to critical insert locations are extracted and processed individually. These images are evaluated using dedicated machine learning models (bushings, outer insert, inner insert), each trained to classify presence and correctness.

Based on the combined results of all inspections, the system determines whether the part is acceptable. Passing parts are conveyed forward for operator handling, while failed parts are automatically diverted to a scrap bin. Process data is captured and stored for traceability and analysis.

During deployment, an operational edge case was identified: parts remained stationary for image capture longer than in the previous process, leading to occasional operator interference within the inspection area. This was resolved by integrating light curtains to enforce safe and consistent operation, improving both classification reliability and system safety.

Outcomes:

• Implemented ML-based inspection without requiring major hardware changes

• Automated detection of missing or incorrect inserts across multiple critical features

• Reduced manual inspection requirements and improved process consistency

• Identified and resolved a real-world operator interaction issue impacting system reliability

• Established a foundation for integrating machine learning into existing PLC-controlled systems

ERP Integration

When I joined Eclipse Mold, I was tasked with designing and implementing a data pipeline between the shop floor and the ERP system (DelmiaWorks / IQMS).

The facility’s injection molding presses were already equipped with individual PLCs, which I leveraged to extract machine data without requiring additional hardware. I configured Modbus TCP servers on each PLC and worked with IT to establish a dedicated industrial network using managed Ethernet switches.

A centralized PC running Node-RED was used to poll machine registers in real time and forward the data to an OPC UA server (Prosys OPC UA Server). The ERP system was then configured to consume this data stream for production tracking and reporting.

The system has operated reliably for over 16 months, with downtime limited to extended power outages. Both Node-RED and the OPC UA server are configured for automatic startup, allowing the entire pipeline to recover immediately after power restoration.

Outcomes:

• Successfully delivered a stalled ERP integration project within weeks of joining

• Eliminated the need for additional third-party software or licensing costs

• Established a stable, real-time data pipeline between shop floor equipment and ERP systems 

Tooling Repair & Asset Tracking System

At Maple Mold Technologies, I designed and implemented a system to track tooling repairs and mold movement between two facilities. The initial approach relied on a combination of Excel, VBA, and Access, but compatibility issues between different Office versions made the solution unreliable across machines.

To address this, I developed a web-based application using Django (Python) and Bootstrap, with SQL Server Express as the backend database. This architecture eliminated dependency on local software configurations and allowed the system to be accessed from any device with a web browser.

The application was deployed on the company’s internal network using the Waitress web server and integrated into the existing domain infrastructure. This enabled seamless access from both the production facility and the tool shop, including mobile devices connected to the network.

The system provided centralized visibility into tooling status, repair history, and location tracking, improving coordination between facilities and reducing reliance on manual communication.

Outcomes:

• Replaced an unreliable Excel/VBA-based workflow with a stable web-based system

• Enabled cross-facility access from any device on the internal network

• Improved tracking and visibility of tooling repairs and asset movement

• Established a reusable web application framework for future internal tools

Predictive Maintenance (AI)

Capstone project - Computer Science, B.S. - Western Govenors Univerity

Completed March 2025.

More information coming soon!