School of Engineering, Computer and Mathematical Sciences - Te Kura Mātai Pūhanga, Rorohiko, Pāngarau

Permanent link for this collectionhttps://hdl.handle.net/10292/553

AUT is home to a number of renowned research institutes in engineering, and computer and mathematical sciences. The School of Engineering, Computer and Mathematical Sciences strong industry partnerships and the unique combination of engineering, computer and mathematical sciences within one school stimulates interdisciplinary research beyond traditional boundaries. Current research interests include:

Artificial Intelligence; Astronomy and Space Research;

Biomedical Technologies;

Computer Engineering; Computer Vision; Construction Management;

Data Science;

Health Informatics and eHealth;

Industrial Optimisation, Modelling & Control;

Information Security;

Mathematical Sciences Research; Materials & Manufacturing Technologies;

Networking, Instrumentation and Telecommunications;

Parallel and Distributed Systems; Power and Energy Engineering;

Software Engineering; Signal Processing; STEM Education;

Wireless Engineering;

Browse

Now showing 1 - 20 of 1865

Oracle Upper Bounds on Clean-EEG Recoverability from Single-Channel Decompositions Under EOG/EMG Contamination
(MDPI AG, 2026-04-22) Shaikh, Usman Qamar; Kalra, Anubha Manju; Lowe, Andrew; Niazi, Imran Khan
Objective: Single-channel EEG artifact suppression often relies on signal decomposition; however, it is not always clear how much clean EEG is recoverable from a given decomposition when component weighting is ideal. We present an oracle-based benchmark that characterises this best-case recoverability across common 1-D decomposition families under controlled EOG, EMG, and mixed contamination. This work does not propose a new denoising algorithm; rather, it isolates representation capacity from component-selection heuristics by computing an upper bound on reconstruction quality. Approach: Using EEGdenoiseNet, we constructed a synthetic benchmark of 4500 single-channel 2 s segments (125 Hz; T = 250) by mixing clean EEG with ocular (EOG) and/or cranial EMG exemplars at noise-to-signal ratios (NSRs) spanning −10 to +10 dB (floor −10 dB denotes an absent modality). We evaluated variational mode decomposition (VMD), singular spectrum analysis (SSA), discrete wavelet transform (DWT), and CEEMDAN by decomposing each mixture and reconstructing the clean EEG using a bounded nonnegative linear combination of components obtained via constrained least squares (the oracle). Main results: Under this oracle benchmark, SSA achieved the lowest reconstruction error in most tested conditions, while DWT tended to rank best in milder ocular regimes; VMD performance improved, with an increased mode count at higher computational cost. CEEMDAN exhibited higher latency dominated by ensemble settings. Significance: These results should be interpreted as decomposition-level upper bounds under controlled mixtures, not field-ready denoising performance. The benchmark provides a tool with which to compare representational recoverability across decompositions and to inform the subsequent design of practical component-selection strategies.
A Comprehensive Survey of NO₂ Gas Sensors: Functional Materials, Recent Advances, Present Challenges, and Future Directions
(Elsevier BV, 2026-04-21) Akindoyo, John Olabode; Yuan, Xiaowen; Li, Xue Jun
Nitrogen dioxide (NO₂), a major atmospheric pollutant, poses significant risks to human health even at low concentrations and exerts similarly harmful effects on plants and terrestrial ecosystems. Therefore, the real-time, accurate and reliable monitoring of NO₂ concentrations in the atmosphere is very important. NO₂ gas sensors translate the concentration of NO₂ at any given time into analysable signals. So, different materials/technologies have been explored to develop NO₂ sensors. However, there is an increasing global awareness about the possible connections between technological advances, waste generation, and environmental pollution. Moreover, research and technological advancements have recently tilted towards the utilisation of artificial intelligence to develop smart devices. Therefore, this paper reviews the different functional materials in NO₂ sensors, including their advantages and limitations. Likewise, the techniques/technologies for the fabrication of NO₂ gas sensors are discussed. Moreover, recent developments that support sustainability, circularity, or end-of-life composability are presented. Furthermore, the contemporary trend of research towards fast and seamless integration of new generation NO₂ gas sensors with AI and the use of IoT for remote data collection is included in this review. Finally, future perspectives on the manufacture of smart, durable, mechanical, and environmentally robust NO₂ gas sensors are discussed.
Probabilistic Modeling of Available Transfer Capability With Dynamic Transmission Reliability Margin for Renewable Energy Export and Integration
(MDPI AG, 2026-04-10) Edeh, Uchenna Emmanuel; Lie, Tek Tjing; Mahmud, Md Apel
This paper develops a probabilistic Available Transfer Capability (ATC) framework that quantifies export headroom for renewables across transmission-distribution interfaces under time-varying uncertainty. Static transmission reliability margins can unnecessarily curtail exports. A dynamic transmission reliability margin (TRM) is embedded within ATC using rolling window statistics and adaptive confidence factor scheduling to release capacity in calm periods and tighten margins during volatile transitions. Uncertainty is modeled as net nodal power imbalance variability from load and renewable deviations, together with stochastic thermal limit fluctuations. Correlated multivariate scenarios are generated via Latin Hypercube Sampling with Iman-Conover correlation preservation and propagated through full AC power flow analysis. Validation on the IEEE 39-bus system and New Zealand’s HVDC inter-island corridor recovers 93.31 MW of usable transfer capacity on the IEEE system relative to the pooled Monte Carlo P95 constant-margin baseline, with 78.11 MW attributable to rolling window volatility tracking and 15.20 MW to adaptive confidence factor scheduling, and 59.51 MW (+7.6%) on the New Zealand corridor relative to the corresponding pooled Monte Carlo P95 baseline, with the gain arising primarily from rolling window volatility tracking. Relative to a 95% one-sided reliability target, achieved coverage is 93.9% for IEEE and 91.8% for New Zealand, translating into increased export headroom and reduced curtailment.
Characterization of Nano Composite Film Based on Plasticised Polyhydroxybutyrate and Polycaprolactone Blends Films for Packaging Applications
(American Society of Mechanical Engineers, 2026-02-23) Govindan, Srinivasan; Ramos, Maximiano; Al-Jumaily, Ahmed
This research demonstrates that improving the characterization of Polyhydroxy butyrate (PHB) with 20 wt% Polycaprolactone (PCL), 5 wt% mixed Plasticizer (i.e., monomeric plasticiser (Triacetin 99%) and Polymeric plasticiser (Ultramoll IV), and nanomaterials, i.e., 1 wt% nanocellulose and 6 wt% nano clay, can improve the PHB characteristics for packaging purpose. Biodegradable polymer films were developed using injection moulding and hot-pressing methods, and characterised by tensile properties, water vapor barrier properties, and biodegradation properties in compost and seawater medium. The properties of the nanocomposite films compared to plasticised PHB-PCL 20 films show improved tensile strength, water vapour barrier properties, and slightly higher biodegradation rate with nanomaterials, though with reduced elongation at break. The nanocomposite films were also compared to neat PHB, with 1% nanocellulose (nCell) and 6% nano clay (nClay) providing higher tensile elongation (168% & 146% higher), higher water vapor barrier properties / Lower water vapor transmission rate (46% & 58% lower), and higher biodegradation in home compost (13 % & 11% higher) and seawater medium (10% & 11% higher). The research outcome shows that the high-performance PHB nanocomposite blend with 1% nanocellulose and 6% nano clay.
Modelling the Signatures of Supercritical Geothermal Resources Under Thermally-inhibited Permeability Constraints
(Copernicus Publications, 2026-03-13) Banerjee, Dona; Dempsey, David; Kennedy, Ben; Cater, John; Hewett, James; Cusack, Dale
Geothermal systems in the Taupō Volcanic Zone (TVZ), New Zealand, are sustained by large-scale convection of groundwater that is mainly confined to the brittle upper crust. The depth and temperature of the brittle–ductile transition (BDT) is hypothesised to demarcate the lower boundary of fracture-hosted permeability and hence fluid circulation. Thus, we expect the thermal structure of the convection cells, both at their base and the geothermal resource at drillable depths, to be influenced by the BDT temperature. As direct observations below 3 km are difficult to obtain, the objective of this study is to test whether temperatures in the upper 2 km of hydrothermal systems are correlated with BDT temperature and could hence serve as a proxy constraint on deep thermal and permeability structure.This work uses numerical models of hydrothermal circulation that couple Darcy flow and heat transport in a 2D axisymmetric domain. The models assume a deep basal “hotplate” of 800 to 1300°C at 15 km depth and then allows the permeable domain to be dynamically determined as a function of temperature. Following the work of Hayba & Ingebritsen (1997) and Scott et al. (2016), we use a logistic/sigmoid model of rock permeability that decreases smoothly across a prescribed BDT temperature range and whose mid-point temperature was varied between 350 and 650°C. The models reproduce the expected dominance of fluid convection at shallow depths where temperatures are sufficiently low to not inhibit permeability. A convective-conductive boundary forms at a depth that is self-determined by the system balance between shallow convective and deep conductive heat transfer.Analysis across a range of model parameters and anisotropy conditions confirms a correspondence between the rock’s BDT temperature range and hydrothermal fluid temperatures at 2 km depth. Across a range of BDT temperatures and anisotropy assumptions, we recover a linear relationship between the hydrothermal upflow temperature at 2 km depth and the applied BDT temperature (e.g., 338°C at 2 km depth corresponding to a 400–500°C BDT range). Modelled convection cells range in power outputs from 60 to 285 MW, which is consistent with the range of estimates for TVZ geothermal fields. These findings suggest that shallow temperature observations can be used to infer rock rheology and permeability properties in hydrothermal provinces.
An On-Demand Solution for Scalable Reflective Tutoring Using Customised AI Agents
(Australasian Association for Engineering Education (AAEE), 2025-12-09) Chanane, Nawal; Kuo, Matthew
CONTEXT: Generative Artificial Intelligence (AI) has been increasingly explored in engineering education to support student learning and reflection. In a second-year Software Development Practice course at Auckland University of Technology (AUT), Auckland, New Zealand, we developed and deployed an AI agent called the Upskilling Log guidance agent using the Cogniti platform, to help students develop and refine their Individual Upskilling Logs. Students submitted individual upskilling logs with reflective content as part of their Sprint 0 project preparation, allowing educators to focus on more in-depth mentoring during laboratory sessions. PURPOSE OR GOAL: The purpose of developing the Upskilling Log guidance agent was to address students' uncertainty about structuring their logs and effectively incorporating technical content. The agent's prompts were specifically designed to provide targeted feedback and guidance that supported independent critical thinking, self-assessment, and reflective writing skills, while maintaining academic integrity and ensuring alignment with project deliverables. APPROACH OR METHODOLOGY/METHODS: This paper uses a design-based research approach to document the iterative development of an AI agent. The team, consisting of the course lead and a senior learning designer, tested and refined the agent over 12 weeks without student data. The study focused on designing effective prompts and analysing responses across iterations. The Upskilling Log aimed to balance technical and reflective writing, encouraging students to critically evaluate their learning independently outside class. ACTUAL OR ANTICIPATED OUTCOMES: Through iterative development and testing, the Upskilling Log agent was successfully refined to help students structure their reflections around five key areas: development environment learning, team collaboration dynamics, Sprint 1 User Story readiness, areas for continued development, and insights gained during upskilling. Based on the researchers' observations and prompt reviews, students who engaged meaningfully with the agent demonstrated improved log structure, enhanced alignment with project requirements, and reduced dependence on educators for basic feedback guidance. CONCLUSIONS/RECOMMENDATIONS/SUMMARY: This development showed AI tools like Cogniti can effectively design agents that enhance reflective practice and self-directed learning in engineering education through systematic, iterative refinement and clear pedagogical frameworks. It highlighted the importance of careful prompt design, defined boundaries, and balancing AI support with educator guidance. This researcher-led approach offers a transferable model for creating AI agents that foster critical reflection while maintaining academic integrity.
Integrating Energy Efficiency, Water Efficiency and Indoor Environmental Quality Towards Advancing Sustainable Building Designs and Practices in New Zealand: Stakeholders’ Perspectives
(Emerald, 2026-04-08) Poorisat, Tharaya; Aigwi, Itohan Esther; Doan, Dat Tien; GhaffarianHoseini, Ali
Purpose Sustainable building designs and practices (SBDPs) are increasingly recognised worldwide for their potential to enhance environmental sustainability, reduce resource consumption and improve occupant well-being. In New Zealand, however, the integration of energy efficiency (EE), water efficiency (WE) and indoor environmental quality (IEQ) remains a critical but underexplored area. This study aims to investigate the combined benefits, challenges and recommendations associated with integrating EE, WE and IEQ in advancing SBDPs within the New Zealand context. Design/methodology/approach Semi-structured interviews were undertaken with 43 experts engaged in sustainable building initiatives. Thematic analysis was used to explore stakeholder perspectives on the benefits, challenges and recommendations associated with EE, WE and IEQ integration. Findings The findings reveal that EE is widely regarded as a fundamental driver of sustainable buildings, but cost concerns, weak regulations, and market-driven priorities constrain its implementation. WE is frequently overlooked, despite its potential to support water conservation, owing to perceptions of resource abundance, limited policy support and financial barriers. IEQ is increasingly valued for its role in promoting occupant health, comfort and productivity, but regulatory and awareness gaps remain significant. Originality/value This study provides one of the first holistic assessments of EE, WE and IEQ integration in SBDPs in New Zealand. By framing New Zealand as a testbed for countries with high renewable potential, abundant water resources and weak regulatory enforcement, the research advances academic discourse while delivering actionable insights for policymakers, practitioners and investors to promote meaningful sustainability transitions in the built environment.
Recurrent Prompt Learning for Spatio-Temporal Forecasting
(Institute of Electrical and Electronics Engineers (IEEE), 2026-03-18) Chen, C; Liu, Y; Niu, C; Shi, K; Chen, L; Zhu, T
Spatio-temporal forecasting holds great significance for various applications in the intelligent transportation system. Foundation models are revolutionizing spatio-temporal forecasting models due to their one-fits-all generalization capabilities. To reprogram the foundation models for the targeted downstream tasks, prompt learning has emerged as an effective approach by optimizing a small set of learnable input tokens while keeping the backbone intact. However, current prompt learning in the spatio-temporal domain typically suffers from two critical limitations: (1) time-agnostic, which cannot capture the temporal evolution during the prompt learning to deal with temporally dependent or sensitive scenarios. (2) input-agnostic, which remain static upon the end of the training, thus failing to fit different distributions at inference. To address these challenges, we propose a recurrent prompt learning framework named RePro to repurpose foundation models to downstream ST forecasting tasks. For the first challenge, we design a recurrent prompt network that is dynamically conditioned on the time-evolving prompts and recurrently optimized based on the historical context. This design injects time-awareness into the prompt to achieve progressive recalibration of the intermediate representations in the foundation model under varying temporal contexts. For the second challenge, we incorporate input data of the current time step into the update of each recurrent prompt state, leading to the input-conditioned prompt learning. This design effectively encapsulates distributional shifts into the prompt dynamics, improving generalization and robustness. Furthermore, two complementary modules are introduced to facilitate the effective application of the recurrent prompt to the foundation model, i.e., cross-prompt aggregation and layer-conditioned prompt adaptation. Specifically, the first module aims to unify the prompt representation and reduce the redundancy, while the second module distributes the recurrent prompts into diverse layers of the foundation model for hierarchical prompting. Extensive experiments on multiple spatio-temporal forecasting benchmarks demonstrate that RePro consistently outperforms strong state-of-the-art baselines across MAE, RMSE, and MAPE, achieving up to 8.3% reduction in MAE, with ablation studies validating the contribution of each proposed component.
Wavelet Packet Entropy Analysis for Detecting Humidity-induced Interference in Smart Gas Sensors
(Institute of Electrical and Electronics Engineers (IEEE), 2026-03-18) Xiong, Pan; Yuan, Xiaowen; Lu, Zhixing; Wang, Yinglin; Li, Xue Jun
Humidity significantly affects the accuracy and stability of smart gas sensors by altering response characteristics, introducing signal drift, and masking target gas signals, which poses challenges for reliable gas detection under variable environmental conditions. Consequently, humidity interference remains a key bottleneck for dependable gas sensing in resource-constrained embedded and wearable Internet of Things (IoT) systems. Although this study employs simulated synthetic datasets, all simulation parameters were carefully benchmarked against empirical sensor characteristics reported in existing literature, ensuring high fidelity to real-world behavior. This paper proposes a physics-informed computational framework based on Wavelet Packet Modal Entropy (WPME) to detect humidity-induced signal degradation without hardware augmentation and establishes a quadratic entropy-humidity model (WPME = 0.059A² − 0.119A + 0.094) derived from simulated NH₃ signals under 30%~86% RH. This model identifies a critical transition at 62.0% RH, which marks the shift from monolayer adsorption to capillary condensation. Below this threshold, entropy decreases by about 0.05 bits per 10% RH, while above it, the rate accelerates to 0.12 bits per 10% RH. The method achieves 98.26% detection accuracy (with only 1.47% degradation under extreme humidity) with a 0.16 ms response time, which demonstrates the viability of WPME as a fast, hardware-free solution for real-time humidity compensation in smart gas sensors.
Adaptive Learning-driven Contention Window Selection for Efficient Channel Access in Vehicular Networks
(Institute of Electrical and Electronics Engineers (IEEE), 2026-03-24) Hota, Lopamudra; Kumar, Arun; Chong, Peter Han Joo
In Vehicular Ad-hoc Networks (VANETs) and major transportation systems, efficient communication protocol is vital for timely data transmission to vehicles. The dense vehicular network poses challenges to efficient channel-sharing. For the proper utilization of the available bandwidth, optimization of channel mechanisms is crucial. The proposed approach enables vehicles to dynamically tune their Contention Windows (CWs) using locally observable MAC-layer information, with the objective of jointly maximizing throughput, minimizing delay, and maintaining fair channel access. Comprehensive simulations and analysis show notable improvement in the overall network efficiency in terms of throughput, collision, and delay. The adaptiveness of the proposed algorithm guarantees flexibility to changing traffic conditions and is well-suited to the evolving Intelligent Transportation Systems (ITS). With an emphasis on high throughput, low latency, and fair channel allocation, the proposed model contributes to the advanced communication protocols for VANETs. The proposed model also highlights the significance of intelligent adaptive techniques in obtaining enhanced network performance.
Tertiary Student Experiences Connecting Context and Mathematics While Mathematically Modelling
(Springer Nature Switzerland, 2025-08-10) Spooner, Kerri
Connecting mathematics and contexts is fundamental when mathematically modelling. However, connecting mathematics and contexts appears unfamiliar for mathematics students. To inform teaching practices, this study looked at: What are the student experiences for connecting mathematics and contexts while participating in mathematical modelling? Tertiary student mathematical modelling experiences with modelling activities from three different New Zealand university modelling courses were looked at. Data was collected through semi-structured interviews with the students. Reflective thematic analysis was used to identify themes in students’ experiences. The results showed connecting mathematics and contexts was: new for students; can make mathematics more relatable; a different experience for different students depending on their backgrounds.
Fully Automated, Deep Learning, Cardiac CT-based Multimodal Network for Cardiovascular Risk Stratification in High-Risk Perioperative Patients
(Oxford University Press, 2026-03-04) Lu, Juan; Huangfu, Gavin; Ihdayhid, Abdul; Bennamoun, Mohammed; Konstantopoulos, John; Kwok, Simon; Niu, Kai; Liu, Yanbin; Figtree, Gemma A; Chan, Matthew TV; Butler, Craig R; Tandon, Vikas; Nagele, Peter; Woodard, Pamela K; Mrkobrada, Marko; Szczeklik, Wojciech; Abdul Aziz, Yang Faridah; Biccard, Bruce M; Devereaux, Philip James; Sheth, Tej; Williams, Michelle C; Newby, David E; Chow, Benjamin JW; Dwivedi, Girish
AIMS: Major adverse cardiac events (MACE) significantly impact perioperative morbidity and mortality. We aimed to develop a fully automated multimodal deep learning (DL) system integrating patient demographics, comorbidities, and coronary computed tomography angiography (CCTA) findings to optimize risk prediction. METHODS AND RESULTS: We included 639 patients undergoing CCTA as part of perioperative risk assessment for elective non-cardiac surgery. Convolutional neural networks automatically identified coronary artery disease reporting and data system (CAD-RADS) scores and segmented the left ventricle, aorta, and heart. These imaging features were combined with patient demographics and comorbidities to predict MACE risk. We evaluated the performance of our multimodal model against the revised cardiac risk index (RCRI) using gradient boosting decision tree modelling and area under the receiver operating characteristic (AUROC) curves. Among 639 patients (mean age 70 ± 9 years, 56% males, median RCRI 1), 61% underwent orthopaedic surgery, 27% vascular surgery and the rest abdominal/pelvic or spine surgery. 45 patients experienced MACE within 30 days. Automated CAD-RADS (AUROC = 0.69) demonstrated comparable performance to human analysis (AUROC = 0.67, P = 0.77). The multimodal DL system (AUROC = 0.82) outperformed CAD-RADS (delta-AUROC = 0.13, CI: 0.02, 0.26, P = 0.02), and RCRI (delta-AUROC =0.22, CI: 0.05, 0.46; P = 0.001) in predicting MACE and demonstrated robust sensitivity (83%) and specificity (79%). CONCLUSION: Our multimodal system built using automated CAD-RADS, anatomical segmentations and patient demographics outperforms both human expert and automated CAD-RADS for MACE prediction. This approach has the potential to enhance patient outcomes by leveraging the synergy between automated imaging and clinical data.
Scrum Master Competencies: A Content Analysis of Job Postings
(Elsevier, 2026-03-24) Mehta, Nitin; Tan, Felix B
This study explores the competencies required for Scrum Masters by analyzing 51 job postings from the United States, the United Kingdom, New Zealand, and Australia. The findings reveal a growing demand for Scrum Masters with a diverse skill set, including leadership, coaching, communication, and problem-solving, alongside Agile expertise. The results align with frameworks such as the Scrum Guide and SAFe, emphasizing the importance of practical experience and the Scrum Master’s role in driving organizational transformation and promoting Agile maturity. These insights highlight the evolving expectations for Scrum Masters in modern Agile environments.
AVAESA: Adaptive VAE With Self-attention and Learnable Signal Processing for Robust Radar-based Heart Rate Estimation
(Elsevier BV, 2026-03-26) Shirazi, Mohammad Hossein; Yongchareon, Sira; Singh, Anuradha; Ma, Jing
Non-contact heart rate monitoring using radar sensors offers significant advantages for healthcare and automotive applications by preserving privacy while enabling continuous physiological assessment. Current Variational Autoencoder (VAE) approaches for radar-based vital sign monitoring, while superior to traditional neural networks, suffer from fixed preprocessing assumptions and inadequate temporal modeling that limit their generalization across diverse measurement conditions. This study introduces AVAESA (Adaptive VAE with Self-Attention and Learnable Signal Processing), a novel architecture that addresses these limitations through three key innovations: dual-stream in-phase/quadrature signal processing that preserves critical phase relationships, multi-head self-attention mechanisms for enhanced temporal dependency modeling, and adaptive signal preprocessing with learnable parameters that derive frequency bands and processing weights directly from input signal characteristics. The framework was evaluated on 1920 measurements from 10 participants across 48 measurement scenarios (4 distances  ×  3 angles  ×  4 orientations), assessing cross-scenario robustness under varied measurement conditions, with Polar H10 chest strap ground truth validation. Comprehensive comparison against multiple architectures (CNN, LSTM, Bi-LSTM, TCN, VAE) with statistical significance testing demonstrates substantial performance improvements, with mean absolute error reductions ranging from 17.3% under optimal conditions to 62.6% under challenging cross-scenario generalization scenarios. AVAESA maintains high accuracy (correlation coefficient  >  0.86, R² >  0.84) even under challenging measurement conditions where baseline approaches exhibit degraded performance, demonstrating potential for contactless cardiac monitoring systems across diverse measurement environments through improved cross-scenario robustness.
Computational Investigation of the Cooling and Heating Effect in Vortex Tube
(ASME International, 2026-03-31) Duan, Dingli; Guan, Shuaibing; Qiao, Yanfeng; Wang, Jay
Vortex tubes exhibit a unique temperature separation phenomenon, leading to their widespread application in refrigeration and heating. Using the exact flow and temperature field simulation, this study examined the cooling effect and heating effect under different cold mass fractions, inlet pressure and diameter ratios. The cooling effect shows a trend of first increasing and then decreasing with the increase of the cold mass fractions while the heating effect gradually increases. Actually, the cooling and heating effect are suppressed by backflow, it is difficult for the working fluid to flow out of the cold end leading to the backflow at the cold end is severe at low cold mass fractions, and the reverse flow boundary exceeds the control valve causing the backflow at the hot exit is enhanced at high cold mass fractions. In order to get excellent cooling effect and heating effect of the vortex tube, the inlet pressure should be increased while reducing the diameter ratio of the vortex tube, and the cold mass fractions should be controlled within a reasonable range (CMF=0.3∼0.7).
eXCube2: Explainable Brain-inspired Spiking Neural Network Framework for Emotion Recognition From Audio, Visual and Multimodal Audio–Visual Data
(MDPI AG, 2026-03-14) Kasabov, NK; Yang, A; Wang, Z; Abouhassan, I; Kassabova, A; Lappas, T
This paper introduces a biomimetic framework and novel brain-inspired AI (BIAI) models based on spiking neural networks (SNNs) for emotional state recognition from audio (speech), visual (face), and integrated multimodal audio–visual data. The developed framework, named eXCube2, uses a three-dimensional SNN architecture NeuCube that is spatially structured according to a human brain template. The BIAI models developed in eXCube2 are trainable on spatio- and spectro-temporal data using brain-inspired learning rules. Such models are explainable in terms of revealing patterns in data and are adaptable to new data. The eXCube2 models are implemented as software systems and tested on speech and video data of subjects expressing emotional states. The use of a brain template for the SNN structure enables brain-inspired tonotopic and stereo mapping of audio inputs, topographic mapping of visual data, and the combined use of both modalities. This novel approach brings AI-based emotional state recognition closer to human perception, provides a better explainability and adaptability than existing AI systems. It also results in a higher or competitive accuracy, even though this was not the main goal here. This is demonstrated through experiments on benchmark datasets, achieving classification accuracy above 80% on single-modality data and 88.9% when multimodal audio–visual data are used, and a “don’t know” output is introduced. The paper further discusses possible applications of the proposed eXCube2 framework to other audio, visual, and audio–visual data for solving challenging problems, such as recognizing emotional states of people from different origins; brain state diagnosis (e.g., Parkinson’s disease, Alzheimer’s disease, ADHD, dementia); measuring response to treatment over time; evaluating satisfaction responses from online clients; cognitive robotics; human–robot interaction; chatbots; and interactive computer games. The SNN-based implementation of BIAI also enables the use of neuromorphic chips and platforms, leading to reduced power consumption, smaller device size, higher performance accuracy, and improved adaptability and explainability. This research shows a step toward building brain-inspired AI systems.
Data-Driven Parameter Identification of Synchronous Generators: A Three-Stage Framework with State Consistency and Grid Decoupling
(MDPI AG, 2026-03-24) Peykarporsan, Rasool; Govinda Waduge, Tharuka; Lie, Tek Tjing; Stommel, Martin
As modern power systems grow increasingly complex, there is a pressing need for stability analysis methods capable of handling nonlinear dynamics while providing physically meaningful and reliable stability indices. Port-Hamiltonian (PH) frameworks have emerged as strong candidates in this regard, offering inherently stable formulations, energy-consistent representations, and modular plug-and-play scalability. However, the practical deployment of PH-based stability analysis remains hindered by the absence of reliable, high-fidelity parameter identification methods that rely on sensor measurements to capture system dynamics while remaining compatible with PH model structures. This paper addresses that gap by proposing a comprehensive three-stage data-driven identification framework for PH modeling of synchronous generators—the central dynamic component of any power system. While the IEEE Standard 115 provides established procedures for transient parameter identification, it exhibits fundamental limitations when applied to PH modeling, including single-scenario identifiability constraints, noise-sensitive derivative-based formulations that amplify sensor measurement errors, and the inability to decouple generator-internal damping from grid contributions. The proposed framework resolves these limitations through multi-scenario excitation using sensor-acquired voltage and current signals, derivative-free state consistency optimization, and physics-based regularization that enforces PH structure preservation. Complete identification of eight key parameters (H, D, Xd, Xq, Xd′, Xq′, Tdo′, Tqo′) is achieved with errors ranging from 1.26% to 9.10%, and validation confirms RMS rotor angle errors below 1.2° and speed errors below 0.15%, demonstrating suitability for transient stability analysis, passivity-based control design, and oscillation damping assessment.
Cyber-Physical Anomaly Detection a Deep Adversarial Fusion of Sensor and Network Data
(Springer Science and Business Media LLC, 2026-03-30) Pinto, Andrea; Herrera, Luis-Carlos; Donoso, Yezid; Gutierrez, Jairo A
In critical infrastructure, the convergence of physical systems with digital networks forms complex Cyber-Physical Systems (CPS), that are vulnerable to threats compromising both data and physical operations. Traditional security systems, often focused solely on network traffic, create a significant security gap by neglecting the rich contextual data provided by physical sensors. To address this issue, the paper introduces a novel unsupervised multimodal framework that synthesizes data from these dual sources for holistic anomaly detection. The proposed architecture combines pre-trained Variational Autoencoder-Long Short-Term Memory (VAE-LSTM) networks to model temporal dependencies with a dual cross-attention mechanism for deep fusion of latent representations. To enhance the detection of subtle, low-observability threats, the model is further regularized through adversarial training using a discriminator that distinguishes between original and reconstructed data. Evaluated on the comprehensive SWAT dataset, the model successfully identifies 24 out of 26 relevant attack scenarios using 10-second time sequences and achieves an Area Under the Curve (AUC) of 0.87, outperforming unimodal benchmarks. This work validates the critical importance of deep data fusion and presents a more resilient, context-aware defense mechanism for modern CPS.
Benchmarking Nonlinear Controllers for a Quad Active Bridge Resonant Converter Supplying Pulsed Power Loads in DC Microgrids
(Elsevier BV, 2026-05) Nduwamungu, Aphrodis; Lie, Tek Tjing; Nair, Nirmal KC; Lestas, Ionis
This paper investigates advanced nonlinear control strategies for a Quad Active Bridge (QAB) resonant converter supplying pulsed power loads (PPLs) in DC microgrids, where fast load transients and inter-port coupling challenge voltage stability. Three controllers are designed and evaluated for regulating the 48 V ports 3 and 4 under a PPL step from 100 W to 200 W : ASSOSMC, IHOSMC, and CMPC–ISMC. In terms of voltage regulation, ASSOSMC achieves the smallest peak deviations ( Δ V 3 = 0 . 5 V , Δ V 4 = 0 . 3 V ), whereas IHOSMC and CMPC–ISMC exhibit larger voltage dips of 3 . 0 / 2 . 8 V and 3 . 5 / 4 . 2 V at ports 3/4, respectively. Overshoot is kept below 1.531% (ASSOSMC) and limited to 0.667% (IHOSMC and CMPC–ISMC), with fast settling times of 3 . 6 μ s (ASSOSMC and CMPC–ISMC) and 4 . 7 μ s (IHOSMC). Energy efficiency is evaluated from steady-state averaged port powers over a common interval ( t = 2 . 1166 – 2 . 8610 s ): CMPC–ISMC achieves the highest overall efficiency ( η tot ≈ 90 . 775 % ), followed by ASSOSMC ( η tot ≈ 90 . 724 % ) and IHOSMC ( η tot ≈ 89 . 055 % ). Hardware-in-the-loop validation on an OPAL-RT platform confirms the effectiveness of the proposed controllers for stabilizing QAB port voltages under PPL disturbances, and highlights CMPC–ISMC as a balanced solution combining high efficiency with reduced current stress on the regulated ports.
LIGO Core-collapse Supernova Detection Using Convolutional Neural Networks
(MDPI AG, 2026-03-10) Pan, Zhicheng; Zahraoui, El Mehdi; Maturana-Russel, Patricio; Cabrera-Guerrero, Guillermo
Core-collapse supernovae (CCSNe) remain a critical focus in the search for gravitational waves in modern astronomy. Their detection and subsequent analysis will enhance our understanding of the explosion mechanisms in massive stars. This paper investigates the use of convolutional neural networks (CNN) to enhance the detection of gravitational waves originating from CCSNe. We employ two time–frequency analysis techniques to generate spectrograms (training data): short-time Fourier transform (STFT) and Q-transform (QT). Two CNNs were trained independently on sets of spectrogram images of simulated CCSNe signals and advanced LIGO noise. The CNNs detect CCSNe signals based on their time–frequency representation. Both CNNs achieve a near 100% true positive rate for CCSNe GW events with a signal-to-noise ratio greater than 0.5 in our test set. Nevertheless, the CNN trained on the STFT spectrograms outperforms the one based on the Q-transform for SNRs below 0.5.

Browse

Recent Submissions