Anomaly detection in Skin Model Shapes using machine learning classifiers

Filmon Yacob,Erik Nordgren,Daniel Semere

doi:10.1007/s00170-019-03794-z

Abstract

The concept of Skin Model Shapes has been proposed as a method to generate digital twins of manufactured parts and is a new paradigm in the design and manufacturing industry. Skin Model Shapes use discrete surface representation schemes, such as meshes and point clouds, to represent surfaces, which makes them enablers to perform an accurate tolerance analysis and surface inspection. However, online inspection of manufactured parts through use of Skin Model Shapes has not been extensively studied. Moreover, the existing geometric variation inspection techniques do not detect unfamiliar changes within tolerance, which could be the precursors to the onset of the manufacturing of out of tolerance part. To detect the unfamiliar changes, as anomalies, and categorize them as systematic and random variations, some unique surface characteristics can be extracted and studied. Random surface deviations exhibit narrow normal distributions, and systematic deviations, on the other hand, exhibit wide, skewed, and multimodal distributions. Using those surface characteristics as key traits, machine learning classifiers can be used to classify deviations into systematic and random variations. To illustrate the method, multiple samples from a truck component manufacturing line were scanned and the collected 3D point cloud data was used to extract features. A prediction score of 97–100% can be achieved by decision tree, k-nearest neighbor, support vector machines, and ensemble classifiers. The purposed approach is expected to extend the existing online inspection approaches and applications of Skin Model Shapes in quality control.

Highlights

In the era of transition of manufacturing towards mass customization [1] and mass personalization [2], there is a need to inspect each manufactured product, preferably through automated online inspection systems
Surface deviations are often controlled by Geometric Dimensioning and Tolerancing (GD&T) schema that applies a set of primitive shapes
To represent acceptable normal variation, half of each data set was generated in such a way that they fall below a hypothetical threshold

Summary

Introduction

In the era of transition of manufacturing towards mass customization [1] and mass personalization [2], there is a need to inspect each manufactured product, preferably through automated online inspection systems. Application of automated inspection systems is one of the key goals of manufacturing industry in transition to Industry 4.0 [3, 4]. Automated surface inspection systems require inputs of detailed digital representation of physical products to classify as within or outside specifications. The concept of Skin Model, and its operationalization through Skin Model Shapes, has been proposed as representation of non-ideal surfaces [5] and digital twin of manufactured parts [6]. Skin Model Shapes (hereafter SMS) are obtained by discretizing nominal CAD model into finer meshes or from tactile and optical devices in the form of point clouds. The discrete representation scheme enables capturing a more detailed geometric characteristic of manufacturing parts. The classical methods of conformance to specification have been to fit primitive shapes that capture variation of simple shapes only

Methods

Findings

Discussion

Conclusion