Case Study

FRISA Uses Deep Learning to Automate Aerospace Quality Checks

FRISA transformed a manual inspection bottleneck into an AI-assisted quality workflow, increasing consistency and precision for aerospace forging operations.

Challenge

FRISA manually inspected over 150 complex rings each day at its plant.
The process was labor-intensive and costly.
Inspection outcomes depended on subjective inspector judgment.
Human error and inconsistent classification reduced reliability.
Manual review became a bottleneck for engineering teams and decision-making.
FRISA needed an AI-driven path to automate inspection and classification.

Solution

Ensitech built a custom deep learning system for FRISA quality control.
The team aligned and centered 3D point cloud scans of forged parts.
The system defined machining height, optimal centering zones, and layout classification logic.
A two-stage pipeline was implemented for robustness and accuracy.
Stage 1 used classical optimization and data cleaning for alignment.
Stage 2 used neural networks trained on labeled inspection data for layout.
Development ran on Microsoft Azure with deployment on AWS SageMaker.
Transfer learning with ResNet-34, ResNet-50, Inception, and DenseNet helped reduce false negatives.
Ensitech embedded mathematicians and engineers with FRISA teams for collaborative delivery.

Results

Achieved 95% precision in identifying parts with potential quality issues.
Eliminated false negatives and reduced false positives to 5%.
Cut inspection time significantly, freeing engineering capacity.
Improved accuracy and consistency in plant-level decisions.
Enabled reuse of AI models and insights across other FRISA initiatives.