Tech Plastics Case Study

Tech Plastics is a leading manufacturer of precision-molded plastic components for medical equipment and disposable products such as syringes, valves, and sensor housings.

As production volumes grew and certification standards became stricter, the company faced a serious challenge: manual visual inspection was slow, inconsistent, and heavily dependent on human factors.

Client Goal — to implement an automated AI-based visual inspection system capable of detecting even the smallest molding defects and geometric deviations in real time, while reducing inspection time and eliminating operator error.

Project Objectives

• Integrate an AI vision system into the existing injection-molding line without halting production.

• Achieve inspection accuracy up to 99% across multiple product types.

• Process up to three items per second with minimal latency.

• Collect and visualize detailed defect statistics for ongoing quality improvement.

• Ensure industrial-grade reliability under challenging factory conditions — vibration, reflections, dust, and variable lighting.

Development and Implementation

1. Production Process Analysis
Polygant’s engineering team conducted a full audit of the production environment — lighting, part geometry, cycle frequency, and workstation layout.
Based on these findings, the team designed a custom setup of cameras and LED lighting to eliminate glare and shadow interference.

2. AI Vision Model Development
A custom computer vision model was trained using the YOLOv8 architecture, optimized for transparent and semi-transparent surfaces.
The dataset included thousands of annotated images containing micro-cracks, air bubbles, deformations, and contamination.
The final model detects defects as small as 0.1 mm with extremely low false-positive rates.

3. Hardware Integration
The system was built around an NVIDIA Jetson module, integrated with Basler industrial cameras and synchronized with the molding machine’s production cycle.
Each part is scanned in under 0.3 seconds, and defective items are automatically rejected.
All inspection data is logged in a centralized database for traceability and reporting.

4. Visualization and Analytics
A web-based dashboard was developed for quality engineers and production managers.
It displays real-time statistics on defect types, production line performance, and automated shift reports.
This data helps optimize machine parameters and improve operator efficiency.

Results

• Defect detection accuracy: 98.7%

• Inspection speed: up to 3 parts per second

• Scrap reduction: 35%

• Overall line productivity: +18%

• System ROI: achieved in less than 8 months

Outcome

By deploying the AI-powered visual inspection system, MedTech Plastics achieved a new level of consistency and transparency in quality control.
The solution not only enhanced ISO and FDA certification readiness but also created a scalable framework for future production lines.

Polygant’s system transformed quality assurance from a manual process into a strategic advantage — turning precision and reliability into measurable business value.