Giovanni Delibra

The integration of Artificial Intelligence (AI) in turbomachinery and fan systems is transforming traditional design, diagnostics, and operational strategies. Artificial Intelligence allows for the efficient exploration of wide design space, easy and fast prediction of fan performance and improving existing system operation and maintenance. Nevertheless, this AI-driven revolution still raises concerns and diffidence in the community, as highlighted by the results of a survey delivered to over 100 fan experts and discussed in this paper. This manuscript aims to provide an overview of Fan-AI applications through a comprehensive literature review of notable use cases. The applications target different stages of the life cycle of fans, from ML-assisted three-dimensional design/optimization to data-driven performance prediction, AI-driven fan control and fault analysis/prognosis. For each of these categories, the relevant application are discussed, highlighting trends, adopted algorithms and strategies, as well as limiting factors. This study also shares the views of experts on both fan design, optimization and operations and AI methods in the upcoming challenges for fan industry. Starting from the need of high-quality data, the improvement of model generalization and the embedding of Fan-AI in the standard engineering practices. This paper concludes with a discussion on the future role of AI in fans, suggesting pathways for research and industrial adoption that balance technological innovation with domain-specific constraints.

The increasing deployment of wind energy systems, particularly offshore wind farms, necessitates advanced monitoring and maintenance strategies to ensure optimal performance and minimize downtime. Supervisory Control And Data Acquisition (SCADA) systems have become indispensable tools for monitoring the operational health of wind turbines, generating vast quantities of time series data from various sensors. Anomaly detection techniques applied to this data offer the potential to proactively identify deviations from normal behavior, providing early warning signals of potential component failures. Traditional model-based approaches for fault detection often struggle to capture the complexity and non-linear dynamics of wind turbine systems. This has led to a growing interest in data-driven methods, particularly those leveraging machine learning and deep learning, to address anomaly detection in wind energy applications. This study focuses on the development and application of a semi-supervised, multivariate anomaly detection model for horizontal axis wind turbines. The core of this study lies in Bidirectional Long Short-Term Memory (BI-LSTM) networks, specifically a BI-LSTM autoencoder architecture, to analyze time series data from a SCADA system and automatically detect anomalous behavior that could indicate potential component failures. Moreover, the approach is reinforced by the integration of the Isolation Forest algorithm, which operates in an unsupervised manner to further refine normal behavior by identifying and excluding additional anomalous points in the training set, beyond those already labeled by the data provider. The research utilizes a real-world dataset provided by EDP Renewables, encompassing two years of comprehensive SCADA records collected from a single offshore wind turbine operating in the Gulf of Guinea. Furthermore, the dataset contains the logs of failure events and recorded alarms triggered by the SCADA system across a wide range of subsystems. The paper proposes a multi-modal anomaly detection framework orchestrating an unsupervised module (i.e., decision tree method) with a supervised one (i.e., BI-LSTM AE). The results highlight the efficacy of the BI-LSTM autoencoder in accurately identifying anomalies within the SCADA data that exhibit strong temporal correlation with logged warnings and the actual failure events. The model’s performance is rigorously evaluated using standard machine learning metrics, including precision, recall, F1 Score, and accuracy, all of which demonstrate favorable results. Further analysis is conducted using Cumulative Sum (CUSUM) control charts to gain a deeper understanding of the identified anomalies’ behavior, particularly their persistence and timing leading up to the failures.

The introduction of hydrogen–methane blends as fuel in gas turbines raises concerns on the capability of state-of-art ventilation systems to dilute possible fuel leaks in the enclosures. Traditional numerical methods to perform leak analysis are limited by the number of factors involved, i.e., location and direction of the leak, cross section area, gas pressure in the pipelines, gas composition, and location of external objects. Hence, this raises the need for novel and fast tools capable for the accurate prediction of fuel dispersion in leak scenarios. To this extent, we propose a novel machine learning approach to model gas leaks. The model is trained on a dataset of numerical simulations accounting for several hydrogen/methane concentrations in the fuel, different storage to ambient pressure ratios at the leak section, and a set of cross-flow ventilation velocities. The architecture of the machine learning model is based on graph neural networks, to solve a node-level regression task predicting fuel concentration in space for different high-pressure leak scenarios. The model shows a significant speed up in predicting fuel dispersion with respect to conventional methodology (0.1 s vs 3.5 h) but the GPU memory requirements proved to be a problem when dealing with 3D domains.

Floating offshore wind energy will play a key role in the clean energy transition. The number of large-scale wind farm projects is growing in regions like Northern Europe, the East Coast of the U.S., and the Mediterranean Sea. Offshore wind farms face fewer layout constraints, as they can be situated in vast, open sea areas. Turbines are often arranged in simple layouts, such as grid patterns, but this can cause significant annual energy production (AEP) losses due to wake–rotor interaction. Increasing spacing can mitigate this effect, but it may not always be feasible due to marine space limitations or higher costs for cabling and maintenance. This paper introduces a multi-objective wind farm optimization framework using a non sorted genetic algorithm (NSGA II) to minimize costs and maximize AEP. The method is applied to two case studies in the Mediterranean Sea, assuming 15 MW wind turbines. Case A is located off the coast of Civitavecchia, and case B in the Gulf of Squillace. AEP evaluation is performed with the open-source library FLOw Redirection and Induction in Steady-State (floris), while optimization is done using pymoo. In case A, the layout characterized by the lowest levelized cost of energy (LCOE) features 16 turbines, achieving an AEP of 709 GWh, an LCOE of 112.06 €/MWh, and wake losses of 2.6%. Meanwhile, in case B, the layout with the lowest LCOE consists of 19 turbines, achieving an AEP of 1140 GWh, an LCOE of 80.82 €/MWh, and wake losses of 4%.

The aim of this research study is to perform experimental tests on a new morphable mini Wells turbine designed to operate in presence of waves with limited energy content which presents a high frequency of occurrence along the Mediterranean coasts (Corsini et al., 2010). The strengths of these mini Wells turbines are: i) extremely light rotor; ii) no need of actuators to morph the rotor blades; iii) low construction and maintenance costs; iv) capability to produce energy starting from low wave heights in the order of a few decimetres. Furthermore, the small size and low investment costs of the turbine make it particularly suitable to be installed either in existing structures, such as anti-reflective perforated caissons (often used in the Mediterranean harbors), or in devices for coastal defense from erosive phenomena located on shallow water conditions. The need to carry out experimental tests derives from the difficulty of handling in a virtual Computational Fluid Dynamics environment the huge difference in the time scale of the turbine, which spins at 1500-3000 rpm, and the wave periods, in the order of some seconds (Corsini et al., 2012; Barnabei et al., 2020). Properly designed lab tests, executed in controlled conditions with irregular sea states, are therefore mandatory to fully characterize the behavior of morphable Wells turbines to be employed in the Mediterranean Sea, and in particular to assess their self-start capability.

Axial fans may be equipped with passive flow control devices to enhance rotor efficiency or minimize noise emissions. In this regard, blade designs influenced by biomimicry, such as rotors with sinusoidal leading edges (LEs), have gained popularity in recent years. However, their design is predominantly driven by a trial-and-error approach, with limited systematic studies on the influence of rotor performance. Furthermore, their effectiveness is typically evaluated under controlled conditions that may significantly differ from operations in real installation layouts. In this work, a systematic review of the design process for sinusoidal LE axial fan rotors is provided, aiming to summarize previous design experiences. Then, a modified sinusoidal LE is designed and fitted to a 7.3 m low-speed axial fan for air-cooled condensers (ACCs). These fans operate at environmental conditions, providing a quasi-zero static pressure rise, often with inflow non-uniformities. A series of RANS computations were run to simulate the performance of the baseline fan and that of the sinusoidal leading edge, considering a real installation setup at Stellenbosh University, where the ACC is constrained between buildings and has a channel running on the ground below the fan inlet. The aim is to explore the nonbalanced inflow condition effects in both rotor geometries and to test the effect of the installation layout on fan performance. The results show that the modification to the rotor allows for a more even distribution of flow in the blade-to-blade passages with respect to the baseline geometry.