Symmetry constrained neural networks for detection and localization of damage in metal plates

2024-09-09

APL Machine Learning

AI Summary

To address damage detection and precise localization in thin aluminum plates, this paper proposes a symmetry-constrained deep learning method incorporating physical priors. Leveraging time-series Lamb wave responses acquired from four piezoelectric transducers arranged in a symmetric square configuration, we design an SE(2)-equivariant-inspired neural network architecture. For the first time, the spatial symmetry of sensor placement and material homogeneity are explicitly encoded as structural constraints within the network—embedding square-topology and transducer equivalence priors—to enhance generalizability and physical consistency. Experimental results demonstrate that the method achieves 99.2% detection accuracy and a mean localization error of 2.58 ± 0.12 mm, outperforming existing data-driven approaches. This work establishes a new, interpretable, and robust intelligent sensing paradigm for structural health monitoring.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Deep Learning Algorithms

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

The present paper is concerned with deep learning techniques applied to detection and localization of damage in a thin aluminum plate. We used data collected on a tabletop apparatus by mounting to the plate four piezoelectric transducers, each of which took turns to generate a Lamb wave that then traversed the region of interest before being received by the remaining three sensors. Upon training a neural network to analyze time-series data of the material response, which displayed damage-reflective features whenever the plate-guided waves interacted with a contact load, we achieved a model that detected with >99% accuracy in addition to a model that localized with 2.58 ± 0.12 mm mean distance error. For each task, the best-performing model was designed according to the inductive bias that our transducers were both similar and arranged in a square pattern on a nearly uniform plate.

Problem

• Detect damage in thin aluminum plates using deep learning
• Localize damage with high accuracy using neural networks
• Analyze Lamb wave data from piezoelectric transducers for damage detection

Innovation

• Symmetry constrained neural networks design
• Piezoelectric transducers for Lamb wave generation
• Time-series analysis for damage detection