[Paper Review] Integrating AI in NDE: Techniques, Trends, and Further Directions

Eduardo Pérez, Cemil Emre Ardıç+10 more|arXiv (Cornell University)|Apr 4, 2024

Impact of AI and Big Data on Business and SocietyDecision Sciences3 citations

TL;DR

This paper presents a comprehensive survey on integrating Artificial Intelligence (AI) into Nondestructive Evaluation (NDE), focusing on magnetic methods, ultrasound, thermography, and optical inspection. It identifies common mathematical problems across NDE tasks, maps AI techniques to sub-tasks, and outlines trends and challenges in advancing NDE 4.0 toward semi-automated, intelligent inspection systems.

ABSTRACT

The digital transformation is fundamentally changing our industries, affecting planning, execution as well as monitoring of production processes in a wide range of application fields. With product line-ups becoming more and more versatile and diverse, the necessary inspection and monitoring sparks significant novel requirements on the corresponding Nondestructive Evaluation (NDE) systems. The establishment of increasingly powerful approaches to incorporate Artificial Intelligence (AI) may provide just the needed innovation to solve some of these challenges. In this paper we provide a comprehensive survey about the usage of AI methods in NDE in light of the recent innovations towards NDE 4.0. Since we cannot discuss each NDE modality in one paper, we limit our attention to magnetic methods, ultrasound, thermography, as well as optical inspection. In addition to reviewing recent AI developments in each field, we draw common connections by pointing out NDE-related tasks that have a common underlying mathematical problem and categorizing the state of the art according to the corresponding sub-tasks. In so doing, interdisciplinary connections are drawn that provide a more complete overall picture.

Motivation & Objective

• To analyze the current state of AI integration in Nondestructive Evaluation (NDE) across key modalities including eddy current testing, ultrasound, thermography, and optical inspection.
• To identify common underlying mathematical problems across diverse NDE tasks to enable cross-domain AI method transfer and synergy.
• To evaluate the extent to which AI enables automation in NDE, particularly at higher operator levels (Level 2 and 3), and to assess current limitations.
• To map AI techniques to specific sub-tasks such as defect detection, signal reconstruction, and parameter estimation in NDE workflows.
• To provide a structured overview of AI applications in NDE 4.0, highlighting opportunities for interdisciplinary innovation and future research directions.

Proposed method

• The study conducts a systematic review of recent AI applications in selected NDE modalities, focusing on techniques like deep learning, compressed sensing, and physics-informed neural networks.
• It categorizes NDE tasks into sub-tasks such as signal denoising, defect detection, image reconstruction, and parameter estimation, linking them to shared mathematical formulations.
• The paper employs a taxonomy of AI methods—e.g., CNNs, RNNs, GANs, and sparse optimization algorithms—applied to NDE problems such as ultrasonic beamforming and thermographic signal reconstruction.
• Common mathematical problems such as inverse problems (e.g., inverse heat conduction, inverse scattering) are identified and linked to AI solutions like ISTA, FISTA, and ADMM.
• The authors use a task-based decomposition of NDE automation levels (aligned with DIN EN ISO 9712) to assess AI’s role in enabling higher levels of autonomy.
• Interdisciplinary connections are drawn by mapping AI techniques across modalities, revealing shared challenges and opportunities in signal processing, reconstruction, and decision-making.

Figure 1 : 20MnMoNi5-5 stress-dependent scatterplot of feature DZmax (maximum value of incremental permeability) which is not sufficient to differentiate the material variants considered in [ 28 ]

Experimental results

Research questions

• RQ1 Which AI techniques are most effective for solving core NDE tasks such as defect detection, signal denoising, and image reconstruction across different modalities?
• RQ2 What common mathematical problems underlie diverse NDE tasks, and how can AI-based solutions be shared across modalities to improve efficiency and performance?
• RQ3 To what extent can AI enable automation in NDE, particularly at higher operator levels (Level 2 and 3), and what are the current technical and methodological barriers?
• RQ4 How do physics-informed and hybrid AI models (e.g., PINNs, SNGANs) improve the accuracy and robustness of NDE results compared to purely data-driven approaches?
• RQ5 What are the key trends and future research directions in AI-powered NDE, especially in the context of NDE 4.0 and digital twin integration?

Key findings

• AI has become indispensable in advancing NDE, particularly in automating complex tasks such as defect detection and signal reconstruction across ultrasound, thermography, and magnetic methods.
• Common mathematical problems—such as inverse problems and sparse signal recovery—underlie diverse NDE tasks, enabling cross-modality transfer of AI solutions like ISTA and ADMM.
• Physics-informed neural networks (PINNs) and generative models (e.g., SNGANs) show promise in improving reconstruction accuracy and robustness in thermography and ultrasound imaging.
• Deep learning models such as U-Net, FCN, and R-CNN achieve high performance in defect detection with reported improvements in Dice scores and F1-scores, though generalization remains a challenge.
• AI-driven approaches significantly enhance signal-to-noise ratio and reduce processing time in full matrix capture (FMC) and synthetic aperture focusing (SAFT) techniques.
• Despite progress, challenges remain in interpretability, generalization across materials and geometries, and integration into standardized NDE workflows and certification processes.

Figure 2 : First two canonical components transformed by LDA of the extended feature space [ 28 ] , demonstrating a clearer separation of individual material variants compared to fig. 1 . "M1" refers to the material 20MnMoNi5-5 and "M2" refers to the material 22NiMoCr3-7