Laser Machining Error Detection Using Image and Height Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current laser processing quality assessment and error detection in laser welding and soldering are complex, requiring extensive expertise and time, with post-process inspections being inefficient and prone to errors due to the need for manual parameter adjustment and expert training, especially when switching materials or processes.

Innovation Solution

A system that uses a neural network to process both image and height data from a workpiece surface to automatically detect processing errors, eliminating the need for manual feature extraction and parameterization, allowing for quick adaptation to changing conditions and materials.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual feature extraction and parameterization are used for quality assessment, then processing errors can be detected, but the system becomes complex and requires extensive expert knowledge and time

Engineering Contradiction:
Improveerror detection reliabilityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces manual mechanical feature extraction and parameterization processes with an automated neural network system. The neural network automatically learns relevant features from raw image and height data, eliminating the need for expert-manual feature engineering and complex parameter setting, thus reducing system complexity while maintaining detection reliability

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

Solution Approach 2:

The neural network performs self-learning and automatic feature extraction without requiring external expert intervention. The system automatically adapts to different materials and processes through the learning algorithm, making the system self-sufficient and reducing dependency on expert knowledge for parameterization

Inventive Principle:
Principle #25Self-service

2Measurement precision

If expert parameterization is used for quality assessment, then accurate error detection is achieved, but production interruptions occur during parameter adjustment

Engineering Contradiction:
Improvequality assessment accuracyVSAvoidproduction interruption time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The neural network is pre-trained with extensive data before deployment, so that when the system is put into operation, it immediately possesses the capability for accurate quality assessment. This preliminary training action eliminates the need for time-consuming parameter adjustments during production, preventing production interruptions

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system transitions from fixed manual parameters to dynamic learned parameters. The neural network automatically adjusts its internal parameters based on the specific material and process conditions being analyzed, eliminating the need for manual parameter changes and associated production interruptions

Inventive Principle:
Principle #35Parameter changes

3Ease of manufacture

If traditional image processing with multiple parameters is used, then quality assessment can be performed, but the process becomes time-consuming and requires frequent re-adjustment when materials change

Engineering Contradiction:
Improvequality assessment easeVSAvoidmaterial change adaptability
Core Design Contradiction:
Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The system transitions from static manual parameter settings to dynamic adaptive learning. The neural network continuously adapts its parameters based on the specific material and process conditions, making the system both easy to operate and highly adaptable to material changes without requiring manual re-adjustment

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The neural network is designed to handle multiple materials and processing conditions with a single unified system. Through its learning capability, it automatically adapts to different materials and processes, eliminating the need for separate parameterization procedures for each material type and simplifying the overall manufacturing process

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

Data Source

PatentEP3883716B1Detection of machining errors of a laser machining system using deep convolutional neural net
Publication Date: 2022.08.17 PRECITEC GMBH
  • EP3883716B1 patent drawingFigure 1
  • EP3883716B1 patent drawingFigure 2
  • EP3883716B1 patent drawingFigure 3A~3B

AI summary

The invention relates to a system for detecting machining errors for a laser machining system for machining a workpiece. The system comprises: a detection unit for detecting image data and height data of a machined workpiece surface; and a computing unit, wherein the computing unit is designed to generate an input tensor based on the detected image data and height data and to determine an output tensor on the basis of the input tensor using a transfer function, said output tensor containing information on a machining error.