PSI - Issue 80

Sadjad Naderi et al. / Procedia Structural Integrity 80 (2026) 77–92 Sadjad Naderi et al. / Structural Integrity Procedia 00 (2025) 000–000

80 4

Fig. 1. Overview of the DT framework.

The remainder of this paper is structured as follows. Section 2 summarises relevant prior work and provides key details on the experimental setup, data acquisition process, and the AI-based framework for crack monitoring and characterisation. Section 3 presents the development of the prognosis framework, detailing the implementation of the physics-informed DBN and its preliminary results. Section 4 extends the discussion by benchmarking the DBN against a purely data-driven GPR model, highlighting key differences and insights. Finally, Section 5 concludes the paper and outlines directions for future research. 2. Real-time crack detection framework Our recent study presented an in-service fatigue crack monitoring method, enabling real-time, baseline-free crack detection. The approach uses third harmonic components generated during fatigue-induced vibrations to detect cracks without relying on baseline signals or externally controlled excitation. In this configuration, the cyclic loading itself serves as the excitation source, eliminating the need for power-intensive high-frequency actuation and reducing sensitivity to environmental and operational variability. Multifunctional piezoelectric (PZT) sensors were employed as passive sensors, supporting continuous monitoring without external power sources. This configuration enables a lightweight, energy-efficient, and scalable system suitable for structural health monitoring in flight. For crack quantification, a Long Short-Term Memory (LSTM) neural network (Sherstinsky 2020) was integrated with Paris’ law. The framework was experimentally validated on six aluminium specimens with varied sensor layouts and loading conditions, demonstrating robustness and generalisability. A concise overview of this framework is provided below to support the present work; full methodological and experimental details can be found in (Pan, Khodaei, and Aliabadi 2025). 2.1. Experimental setup To replicate the operational loading conditions experienced by aircraft structures (e.g., take-off, landing, turbulence, and structural vibration) fatigue tests were conducted under laboratory-simulated low-frequency dynamic loading, ranging approximately from 0.01 to 100 Hz. The test specimens were aluminium 5251 plates, each with dimensions of 300 × 100 × 2 mm and featuring a central circular hole with a diameter of 10 mm to induce stress concentration. The material properties were experimentally characterised: Young’s modulus was determined as 64 GPa, and the yield strength was 143.45 MPa, based on the 0.2% offset method. Six identical specimens were subjected to cyclic loading with a maximum stress of 90 MPa (62% of the yield strength) and a minimum stress of 9 MPa, applied at a load ratio of 0.1. Fatigue cracks initiated and propagated