Advancing a Non-Contact Structural and Prognostic Health Assessment of Large Critical Structures.

This paper presents an overview of integrating new research outcomes into the development of a structural health monitoring strategy for the floating cover at the Western Treatment Plant (WTP) in Melbourne, Australia. The size of this floating cover, which covers an area of approximately 470 m × 200 m, combined with the hazardous environment and its exposure to extreme weather conditions, only allows for monitoring techniques based on remote sensing. The floating cover is deformed by the accumulation of sewage matter beneath it. Our research has shown that the only reliable data for constructing a predictive model to support the structural health monitoring of this critical asset is obtained directly from the actual floating cover at the sewage treatment plant. Our recent research outcomes lead us towards conceptualising an advanced engineering analysis tool designed to support the future creation of a digital twin for the floating cover at the WTP. Foundational work demonstrates the effectiveness of an unmanned aerial vehicle (UAV)-based photogrammetry methodology in generating a digital elevation model of the large floating cover. A substantial set of data has been acquired through regular UAV flights, presenting opportunities to leverage this information for a deeper understanding of the interactions between operational conditions and the structural response of the floating cover. This paper discusses the current findings and their implications, clarifying how these outcomes contribute to the ongoing development of an advanced digital twin for the floating cover.

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon.

This study aims to enhance diagnostic capabilities for optimising the performance of the anaerobic sewage treatment lagoon at Melbourne Water's Western Treatment Plant (WTP) through a novel machine learning (ML)-based monitoring strategy. This strategy employs ML to make accurate probabilistic predictions of biogas performance by leveraging diverse real-life operational and inspection sensor and other measurement data for asset management, decision making, and structural health monitoring (SHM). The paper commences with data analysis and preprocessing of complex irregular datasets to facilitate efficient learning in an artificial neural network. Subsequently, a Bayesian mixture density neural network model incorporating an attention-based mechanism in bidirectional long short-term memory (BiLSTM) was developed. This probabilistic approach uses a distribution output layer based on the Gaussian mixture model and Monte Carlo (MC) dropout technique in estimating data and model uncertainties, respectively. Furthermore, systematic hyperparameter optimisation revealed that the optimised model achieved a negative log-likelihood (NLL) of 0.074, significantly outperforming other configurations. It achieved an accuracy approximately 9 times greater than the average model performance (NLL = 0.753) and 22 times greater than the worst performing model (NLL = 1.677). Key factors influencing the model's accuracy, such as the input window size and the number of hidden units in the BiLSTM layer, were identified, while the number of neurons in the fully connected layer was found to have no significant impact on accuracy. Moreover, model calibration using the expected calibration error was performed to correct the model's predictive uncertainty. The findings suggest that the inherent data significantly contribute to the overall uncertainty of the model, highlighting the need for more high-quality data to enhance learning. This study lays the groundwork for applying ML in transforming high-value assets into intelligent structures and has broader implications for ML in asset management, SHM applications, and renewable energy sectors.

Experimental and Computational Studies on the Scattering of an Edge-Guided Wave by a Hidden Crack on a Racecourse Shaped Hole.

Reliable and quantitative non-destructive evaluation for small fatigue cracks, in particular those in hard-to-inspect locations, is a challenging problem. Guided waves are advantageous for structural health monitoring due to their slow geometrical decay of amplitude with propagating distance, which is ideal for rapid wide-area inspection. This paper presents a 3D laser vibrometry experimental and finite element analysis of the interaction between an edge-guided wave and a small through-thickness hidden edge crack on a racecourse shaped hole that occurs, in practice, as a fuel vent hole. A piezoelectric transducer is bonded on the straight edge of the hole to generate the incident wave. The excitation signal consists of a 5.5 cycle Hann-windowed tone burst of centre frequency 220 kHz, which is below the cut-off frequency for the first order Lamb wave modes (SH1). Two-dimensional fast Fourier transformation (2D FFT) is applied to the incident and scattered wave field along radial lines emanating from the crack mouth, so as to identify the wave modes and determine their angular variation and amplitude. It is shown experimentally and computationally that mid-plane symmetric edge waves can travel around the hole's edge to detect a hidden crack. Furthermore, the scattered wave field due to a small crack length, a, (compared to the wavelength λ of the incident wave) is shown to be equivalent to a point source consisting of a particular combination of body-force doublets. It is found that the amplitude of the scattered field increases quadratically as a function of a/λ, whereas the scattered wave pattern is independent of crack length for small cracks a << λ. This study of the forward scattering problem from a known crack size provides a useful guide for the inverse problem of hidden crack detection and sizing.