Neural Network Inspection for Unknown Fabrication Parameter Discovery

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing automated vision inspection systems only identify defective products but do not provide insights on how to improve product quality, as they do not account for unknown fabrication parameters that may affect quality.

Innovation Solution

A neural network is trained with known and unknown fabrication parameters, where unknown parameters are represented as neurons in the same layer, allowing the network to determine their influence on product quality through image input and quality evaluation results, enabling precise evaluation and identification of unknown parameters affecting quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If automated vision inspection systems are used to identify defective products, then product defect detection capability is improved, but insight into unknown fabrication parameters affecting quality is lost

Engineering Contradiction:
Improvedefect detection capabilityVSAvoidinsight into fabrication parameters
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

A neural network is introduced as an intermediary system that processes both image data and fabrication parameter data. The neural network acts as a mediator between the vision inspection system and the fabrication process, enabling the system to not only detect defects but also identify and quantify the influence of known and unknown fabrication parameters on product quality.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces traditional mechanical vision inspection systems with an intelligent system that uses neural networks to process and analyze data. This substitution enables the system to go beyond simple defect detection and perform complex analysis of fabrication parameter relationships with product quality.

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

2Ease of operation

If only known fabrication parameters are considered in quality inspection, then inspection process simplicity is maintained, but complete quality analysis is insufficient due to unknown parameters

Engineering Contradiction:
Improveinspection process simplicityVSAvoidquality analysis completeness
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The patent adds another dimension to the quality inspection process by introducing unknown fabrication parameters as additional features in the neural network input space. Instead of working only with known parameters, the system expands the parameter space to include unknown factors, allowing comprehensive quality analysis while maintaining operational simplicity through automated neural network processing.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The system dynamically adjusts and identifies fabrication parameters based on the data. The neural network automatically determines which parameters (known or unknown) are most influential on quality, allowing the inspection process to adapt to different manufacturing conditions without requiring manual configuration of all possible parameters.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4058934B1Method and apparatus for product quality inspection
Publication Date: 2026.01.28 SIEMENS AG
  • EP4058934B1 patent drawingFigure 1~2
  • EP4058934B1 patent drawingFigure 3~4
  • EP4058934B1 patent drawingFigure 5A~6

AI summary

A method (200), apparatus (300) for product quality inspection are proposed, to find number of unknown fabrication parameters affecting product quality. A method (200) for product quality inspection includes: getting (S201) of each product in the group of products: image (30), value for each known fabrication parameter affecting quality of the group of products, and quality evaluation result (50); training (S202) a neural network (40), wherein the layer M (401) of the neural network comprises at least one first neuron (401a) and at least one second neuron (401b), and each first neuron (401a) represents a known fabrication parameter affecting quality of the group of products and each second neuron (401b) represents an unknown fabrication parameter affecting quality of the group of products, and the images (30) of the group of products are input to the neural network (40), the quality evaluation results (50) are output of the neural network (40), and the value of each first neuron (401a) is set to the value for the known fabrication parameter the first neuron (401a) representing.