Surface Quality Classification Across Variable Measurement Setups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in comparing and ensuring the surface quality of workpieces across different production locations and setups with varying environmental conditions and measurement systems, leading to inconsistent measurement results and excessive tolerance utilization.
Innovation Solution
A computer-implemented method and system for surface quality inspection that uses a classification system based on surface parameter values derived from raw data, allowing for better comparability across setups and enabling the use of non-standard measurement methods, with intelligent evaluation algorithms that adapt to environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If different sensor technologies and measurement methods are used at different production locations, then measurement coverage and applicability are improved, but measurement result comparability deteriorates
Solution Approach 1:
The patent transforms measurement data from different sensor technologies into a unified parameter space by extracting common surface characteristics (roughness, waviness, form errors) that can be compared across setups. This involves converting raw measurement data into standardized surface parameter representations that are independent of the original measurement method, enabling comparability while maintaining versatility in using different sensors and measurement techniques at different locations.
2Reliability
If traditional measurement methods are used with strict tolerance criteria, then measurement reliability is improved, but tolerance utilization increases leading to excessive rejection rates
Solution Approach 1:
The patent performs preliminary classification of surface quality into distinct categories (defect-free, tolerable defects, unacceptable defects) before final quality decisions are made. By pre-processing measurement data through machine learning models that learn from historical data, the system prepares classification results in advance, enabling more reliable and nuanced quality decisions that reduce excessive rejections while maintaining high reliability.
3Adaptability or versatility
If multiple measurement setups are used in parallel, then measurement coverage is improved, but result comparability and evaluation effort deteriorate
Solution Approach 1:
The patent introduces an intermediary processing layer that standardizes and harmonizes data from multiple measurement setups. This intermediary layer includes data normalization routines, coordinate system transformations, and unified parameter extraction that act as a mediator between diverse measurement systems and the final evaluation process, reducing complexity while maintaining the flexibility to use multiple setups in parallel.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and a corresponding system for surface quality inspection of workpieces are presented. The method involves determining surface parameter values based on initial raw data measured by a surface sensor system from an initial setup, as well as storing these surface parameter values. Furthermore, the method classifies the surface parameter value matrix into quality classes using a trained classifier system from the associated setup. These quality classes comprise a first "Good" class, a first "Draw" class, and a first "Poor" class, with the quality classes being non-overlapping.Additionally, the procedure involves determining a second set of surface parameter values for a second setup based on second measured raw data, as well as classifying the surface parameter matrix into a second good class or a second bad class based on the determined second set of surface parameter values.