Deep Learning Stopper-Edge Inspection for False Reject Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Automated visual inspection systems for pharmaceutical products face challenges in accurately distinguishing particles from bubbles in high viscosity solutions, leading to high false reject rates and equipment redundancy due to poor performance at specific inspection stations.

Innovation Solution

Applying deep learning to an AVI station with a line scan camera to detect and distinguish objects on a vessel stopper edge, using a trained neural network to generate accurate two-dimensional images and reduce false rejects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional machine learning algorithms are used for defect detection, then the system is easier to implement, but the detection accuracy and reliability are insufficient

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional machine learning algorithms (mechanical/system A) with deep learning algorithms (system B) that utilize neural networks. This substitution enables the system to achieve higher detection accuracy and reliability by automatically extracting features from images, while the modular architecture manages the increased computational complexity through standardized processing pipelines.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If deep learning algorithms are implemented, then detection accuracy improves, but computational resources and processing time increase

Engineering Contradiction:
Improvedefect detection reliabilityVSAvoidcomputational resource consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary actions by pre-processing images (normalization, augmentation) and pre-training neural network models on large datasets before actual defect detection. This allows the model to learn general defect patterns in advance, reducing the computational burden during real-time inspection while maintaining high reliability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The deep learning system is divided into modular components: image acquisition module, pre-processing module, neural network analysis module, and result output module. Each module handles specific computational tasks independently, allowing for optimized resource allocation and parallel processing that reduces overall computational resource consumption while maintaining detection reliability.

Inventive Principle:
Principle #1Segmentation

3Productivity

If automated visual inspection is implemented, then productivity increases, but the ability to detect subtle or varying defect types decreases

Engineering Contradiction:
Improveinspection speedVSAvoiddefect type coverage
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The system employs dynamic adaptability through configurable neural network architectures and adjustable detection parameters. The inspection system can dynamically adapt to different defect types by modifying network configurations, threshold values, and processing parameters, enabling high-speed automated inspection across multiple defect categories without sacrificing detection capability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent implements a universal inspection platform that can detect multiple defect types (surface defects, dimensional deviations, contamination, etc.) using a single deep learning system. The neural network is trained on diverse defect patterns and can generalize to detect various defect types, providing both high productivity and broad adaptability through one multi-functional system.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4055559B1Targeted application of deep learning to automated visual inspection equipment
Publication Date: 2026.05.20 AMGEN INC
  • EP4055559B1 patent drawingFigure 1
  • EP4055559B1 patent drawingFigure 2
  • EP4055559B1 patent drawingFigure 3A

AI summary

In a method for enhancing accuracy and efficiency in automated visual inspection of vessels, a vessel containing a sample is oriented such that a line scan camera has a profile view of an edge of a stopper of the vessel. A plurality of images of the edge of the stopper is captured by the first line scan camera while spinning the vessel, where each image of the plurality of images corresponds to a different rotational position of the vessel. A two-dimensional image of the edge of the stopper is generated based on at least the plurality of images, and pixels of the two-dimensional image are processed, by one or more processors executing an inference model that includes a trained neural network, to generate output data indicative of a likelihood that the sample is defective.