Wear Classification of Abrasive Flap Wheels using a Hierarchical Deep Learning Approach

Falko Kähler; Maxim Wille; Ole Schmedemann; Thorsten Schüppstuhl

cs.CVMar 13, 2026

Wear Classification of Abrasive Flap Wheels using a Hierarchical Deep Learning Approach

Authors: Falko Kähler, Maxim Wille, Ole Schmedemann, Thorsten Schüppstuhl

Organizations: Hamburg University of Applied Technology, Institute of Aircraft Production Technology, 21073 Hamburg, Germany.

Abstract

Abrasive flap wheels are common for finishing complex free-form surfaces due to their flexibility. However, this flexibility results in complex wear patterns such as concave/convex flap profiles or flap tears, which influence the grinding result. This paper proposes a novel, vision-based hierarchical classification framework to automate the wear condition monitoring of flap wheels. Unlike monolithic classification approaches, we decompose the problem into three logical levels: (1) state detection (new vs. worn), (2) wear type identification (rectangular, concave, convex) and flap tear detection, and (3) severity assessment (partial vs. complete deformation). A custom-built dataset of real flap wheel images was generated and a transfer learning approach with EfficientNetV2 architecture was used. The results demonstrate high robustness with classification accuracies ranging from 93.8% (flap tears) to 99.3% (concave severity). Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) is utilized to validate that the models learn physically relevant features and examine false classifications. The proposed hierarchical method provides a basis for adaptive process control and wear consideration in automated flap wheel grinding.

Figures & tables

Figure 2: Flap wheel wear categorization

Figure 3: Hierarchical wear classification design

Hierarchy level	1	2		3
No.	Use condition	Flap profile	Flap tear	concave	convex
1	new	rectangular	no tear	n.a.	n.a.
2	used	rectangular	no tear	n.a.	n.a.
3	used	rectangular	with tear	n.a.	n.a.
4	used	concave	no tear	partially	n.a.
5	used	concave	with tear	partially	n.a.

Table 1: Consistent results of the hierarchical wear classification

Hierarchy level	1	2
No.	Use condition	Flap profile	Flap tear
1	new	any	with tear
2	new	concave	any
3	new	convex	any

Table 2: Contradicting result paths between level 1 and 2

Figure 4: Schematic setup for image acquisition. Left: Radial perspective. Right: Axial perspective.

Level	Model (classes)	Training/Validation		Test
Level	Model (classes)	No. tools	Total images	No. tools	Total images
1	Usage condition (new/used)	5/43	3700	7/14	1500
2	Flap tear (with/without)	11/31	3020	12/17	1153
2	Flap profile (rect/concave/convex)	26/33/22	2720	13/16/24	3564
3	Concavity (fully/partially)	14/19	720	4/4	440
3	Convexity (fully/partially)	12/10	1400	4/4	380

Table 3: Dataset composition

Figure 5: Examples of captured images and after preprocessing

Panel of Table 4 — Table 4: Confusion matrix and performance metrics of usage condition classification

Figure 6: Receiver Operating Characteristic (ROC) for the usage condition classification model

Panel of Table 5 — Table 5: Confusion matrix and performance metrics of flap profile classification (level 2)

Figure 7: Grad CAM overlays of correct flap profile classifications

Figure 8: Grad CAM overlays of false flap profile classification

Panel of Table 6 — Table 6: Confusion matrix and performance metrics of concave profile severity classification (level 3)

Panel of Table 7 — Table 7: Confusion matrix and performance metrics of convex profile severity classification (level 3)

Figure 9: Receiver Operating Characteristic (ROC) for the shape classification models. a) Flap profile classification (level 2). b) Concave severity classification (level 3). c) Convex severity classification (level 3)

Panel of Table 8 — Table 8: Confusion matrix and performance metrics of flap tear classification

Figure 10: Receiver Operating Characteristic (ROC) for the tear classification model

Figure 11: Grad CAM overlays of false flap tear classifications

Hierarchical Classification

Hierarchical classification uses relationships among labels to predict categories at multiple levels, from broad groups to fine-grained classes.

See all 89 papers

Towards Successful Implementation of Automated Raveling Detection: Effects of Training Data Size, Illumination Difference, and Spatial Shift
Apr 14, 2026Xinan Zhang, Haolin Wang, Zhongyu Yang +2Distribution Shift Robustness Structural Health Monitoring
James · School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332 · School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332
FlexiGrad: Adaptive Gradient Modulation for Hierarchical Fine-Grained Classification
Jul 20, 2026Zilu Zhou, Dongliang Chang, Junhan Chen +1Hierarchical Classification Gradient Interference
Beijing University of Posts and Telecommunications, Beijing 100876, China · Beijing Key Laboratory of Multimodal Data Intelligent Perception and Governance, Beijing 100876, China
Morphology-Aware Ambiguity Learning for Wafer Defect Decision Support
Sep 18, 2026Seungjun Chu, Seokhyun ChungClassification Semiconductor Manufacturing
School of Industrial and Management Engineering, Korea University, Seoul 02841, Republic of Korea

Apr 14, 2026J/K move · Enter open · S save

Towards Successful Implementation of Automated Raveling Detection: Effects of Training Data Size, Illumination Difference, and Spatial Shift

Xinan Zhang, Haolin Wang, Zhongyu Yang, Yi-Chang+1

James · School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332 · School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332

Distribution Shift Robustness Structural Health Monitoring

Open paper AskPDF arXiv

Raveling, the loss of aggregates, is a major form of asphalt pavement surface distress, especially on highways. While research has shown that machine learning and deep learning-based methods yield promising results for raveling detection by classification on range images, their performance often degrades in large-scale deployments where more diverse inference data may originate from different runs, sensors, and environmental conditions. This degradation highlights the need of a more generalizable and robust solution for real-world implementation. Thus, the objectives of this study are to 1) identify and assess potential variations that impact model robustness, such as the quantity of training data, illumination difference, and spatial shift; and 2) leverage findings to enhance model robustness under real-world conditions. To this end, we propose RavelingArena, a benchmark designed to evaluate model robustness to variations in raveling detection. Instead of collecting extensive new data, it is built by augmenting an existing dataset with diverse, controlled variations, thereby enabling variation-controlled experiments to quantify the impact of each variation. Results demonstrate that both the quantity and diversity of training data are critical to the accuracy of models, achieving at least a 9.2% gain in accuracy under the most diverse conditions in experiments. Additionally, a case study applying these findings to a multi-year test section in Georgia, U.S., shows significant improvements in year-to-year consistency, laying foundations for future studies on temporal deterioration modeling. These insights provide guidance for more reliable model deployment in raveling detection and other real-world tasks that require adaptability to diverse conditions.

Abstract

Hierarchical Classification

Explore similar work

Towards Successful Implementation of Automated Raveling Detection: Effects of Training Data Size, Illumination Difference, and Spatial Shift

FlexiGrad: Adaptive Gradient Modulation for Hierarchical Fine-Grained Classification

Morphology-Aware Ambiguity Learning for Wafer Defect Decision Support