Predictive maintenance in safety-critical systems like turbofan engines increasingly relies on machine learning (ML) models to estimate remaining useful life (RUL), but the ‘black box’ nature of these models hinders their adoption and trustworthiness. While traditional ex-ante prognostic metrics (e.g. monotonicity, trendability) are used to pre-screen sensor data, a systematic comparison against the post-hoc explanations of what a model actually learns is lacking. We explore the application of SHapley Additive exPlanations (SHAP) from explainable artificial intelligence (XAI) to investigate feature importance in engine failure prediction using the second dataset of the Commercial Modular Aero-Propulsion System Simulation (CMAPSS). The preprocessing pipeline includes z-score normalisation of sensor data and the calculation of a health index (HI) to quantify system degradation. A power-law fit is applied to the HI to capture the underlying trends of engine wear and failure progression. We use the normalisation data to calculate prognostic feature selection metrics: monotonicity, trendability and prognosability. Then, we train two machine learning models – random forest (RF) regressor and gradient boosting (GB) method – directly from the raw data to predict the RUL based on the actual sensor readings. The SHAP values generated for both models are analysed to identify the features with the most significant impact on RUL predictions. By comparing the SHAP value distributions across models and prognostic predictors, we highlight feature robustness and their relative influence on engine degradation and failure prediction. This work provides insights into the interpretability of machine learning models in prognostics and enhances the understanding of sensor contributions to engine health monitoring. The results demonstrate the effectiveness of SHAP in elucidating feature importance, supporting the development of more transparent and reliable prognostic systems.

NEURAL MATERIALS (2024) is a live AV show created by SONAMB (Vicky Clarke). The project represents a collaboration between Vicky Clarke, visual artist Sean Clarke, and industry partner Bela, a company specialising in hardware with interactive sensors for music-making. The AV show utilises a new performance system incorporating a hybrid set-up in combination with both a sound sculpture and the output of a machine learning model trained on a ‘post-industrial’ sonic dataset. The dataset renders in sound Manchester’s industrial past and present through field recordings of cotton mills, the canal network and the electromagnetic resonances of a newly gentrified city centre. This article analyses NEURAL MATERIALS as musical composition, live AV show and a demonstration of creative audio-generative AI, linking the work to scholarly and compositional legacies of Sonic Materialism and musique concrète. By combining documentation analysis and performance analysis, I interrogate how sound’s indexical properties are transformed via machine learning (ML) processes, questioning whether machines are able to evoke a sense of space or heritage. Ultimately, I contend that such audio-generative systems have the capacity to reshape our perception of industrial histories, technologies and future sonic realities, indexing sociohistorical cues that are reactivated at the point of listening.

Marine heat waves (MHWs) are prolonged periods of elevated ocean temperatures that can devastate marine ecosystems, fisheries, and coastal communities. Skillfully predicting these events with sufficient lead time is crucial for mitigating their adverse effects. This study presents a probabilistic subseasonal MHW forecast tool using a U-Net-based neural network architecture, with a focus on the Northern Indian Ocean and the Arabian Sea. The model was trained using sea surface temperature and sea surface height reanalysis data. The U-Net-based forecast tool demonstrated significant predictive skill up to 10 weeks in advance across various deterministic and probabilistic skill metrics. The model outperformed persistence and climatology-based benchmarks, especially in the tropical warm pool. Future applications of explainable artificial intelligence (XAI) methods have the potential to identify the sources of predictive skill, inform understanding of underlying dynamics, and improve dynamic subseasonal to seasonal forecast models.

Data-embedded instruments that couple sensing, modelling and sound production are increasingly used in electroacoustic practice, yet their ethical and cultural configurations remain under-analysed. This article develops an ethical-embodied framework for examining how particular data, sensing and mapping arrangements configure relations of care, listening and musical agency. Drawing on feminist and decolonial listening practices, disability and critical data studies and accounts of embodied instrumentality, it combines a selective genealogy of electroacoustic and globally situated practices with a mid-level comparative lens that treats its technical axes as heuristic rather than taxonomic. Case vignettes analyse works using gesture tracking, electromyography (EMG) and brain–computer interfaces (BCI), audience-sensing installations and machine-learning vocal systems, alongside the author’s own data-embedded instrument. Across these examples, the analysis shows how similar technologies can reproduce or contest institutional surveillance, extractivism and aesthetic normativity and outlines implications for the design, evaluation and teaching of data-mediated musical systems foregrounding situated listening and collective accountability.

This study assesses classification-based predictive maintenance (PdM) for aircraft engines on the NASA Commercial Modular Aero-Propulsion System Simulation dataset and addresses the lack of wide-scope, unified benchmarks. PdM is cast as a short-term binary task – predicting whether an engine will fail within the next 30 cycles – and a comparison is conducted across 10 machine-learning models (Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, k-Nearest Neighbor, Naïve Bayes, Extreme Gradient Boosting, LightGBM, CatBoost, and Gradient Boosting) and 3 deep-learning models (Multilayer Perceptron, Gated Recurrent Unit, and Long Short-Term Memory). A leakage-aware pipeline applies Min–Max scaling; class imbalance is handled with Synthetic Minority Over-sampling Technique where appropriate; hyperparameters are tuned via GridSearchCV/BayesSearchCV; and performance is reported with accuracy, precision, recall, F1-score, and receiver operating characteristic–area under the curve (ROC–AUC), complemented by Shapley Additive Explanations (SHAP) explainability and nonparametric significance tests. Sequence models delivered the strongest performance: LSTM achieved Accuracy = 0.981 (Macro-F1 = 0.92; ROC–AUC = 0.96), and GRU achieved ROC–AUC = 0.97 with Accuracy = 0.975. Among classical learners, LightGBM reached Accuracy = 0.972 (Macro-F1 = 0.86; ROC–AUC = 0.93). These gains over weaker baselines were statistically significant across folds. Framing PdM as near-term failure classification yields operationally interpretable alerts. Models that explicitly capture temporal dependencies (GRU/LSTM) best track short-horizon failure dynamics, while gradient-boosted trees offer competitive, lightweight alternatives. The benchmark and analysis (including SHAP) provide a reproducible reference for model selection in aviation PdM.

Reconstructing near-wall turbulence from wall-based measurements is a critical yet inherently ill-posed problem in wall-bounded flows, where limited sensing and spatially heterogeneous flow–wall coupling challenge deterministic estimation strategies. To address this, we introduce a novel generative modelling framework based on conditional flow matching for synthesising instantaneous velocity fluctuation fields from wall observations, with explicit quantification of predictive uncertainty. Our method integrates continuous-time flow matching with a probabilistic forward operator trained using stochastic weight-averaging Gaussian, enabling zero-shot conditional generation without model re-training. We demonstrate that the proposed approach not only recovers physically realistic, statistically consistent turbulence structures across the near-wall region but also effectively adapts to various sensor configurations, including sparse, incomplete and low-resolution wall measurements. The model achieves robust uncertainty-aware reconstruction, preserving flow intermittency and structure even under significantly degraded observability. Compared with classical linear stochastic estimation and deterministic convolutional neural network methods, our stochastic generative learning framework exhibits superior generalisation for unseen realisations under same flow conditions and resilience under measurement sparsity with quantified uncertainty. This work establishes a robust semi-supervised generative modelling paradigm for data-consistent flow reconstruction and lays the foundation for uncertainty-aware, sensor-driven modelling of wall-bounded turbulence.

This article describes in detail several explicit computational methods for approaching such questions in phonology as the vowel/consonant distinction, the nature of vowel harmony systems, and syllable structure, appealing solely to distributional information. Beginning with the vowel/consonant distinction, we consider a method for its discovery by the Russian linguist Boris Sukhotin, and compare it to two newer methods of more general interest, both computational and theoretical, today. The first is based on spectral decomposition of matrices, allowing for dimensionality reduction in a finely controlled way, and the second is based on finding parameters for maximum likelihood in a hidden Markov model. While all three methods work for discovering the fairly robust vowel/consonant distinction, we extend the newer ones to the discovery of vowel harmony, and in the case of the probabilistic model, to the discovery of some aspects of syllable structure.

Punching shear failure in slab-column connections is a brittle collapse mode that threatens the safety of flat reinforced concrete (RC) slabs. Conventional design provisions are generally conservative but exhibit inconsistencies across geometric and material variations. This study develops an eXtreme Gradient Boosting (XGBoost) model to predict the ultimate punching shear capacity of flat RC slabs, using a database of experimental results categorized by four different geometric domains, including square slab with square column, circular slab with circular column, square slab with circular column, and circular slab with square column, covering the geometric, materials strength, and reinforcement properties of input parameters. The model achieved high predictive accuracy across the domains with coefficient of determination (R2) values > 0.930 in unseen testing datasets with minimal bias (0.994–1.006) and reduced scatter. Model interpretability, addressed through the SHapley Additive exPlanations analysis, confirmed slab thickness and average effective depth as the most critical predictors of shear capacity, followed by concrete strength and reinforcement parameters, while boundary condition parameters showed negligible influence due to the predominance of interior column cases. These findings demonstrate that XGBoost provides accurate, reliable, and interpretable predictions of punching shear capacity, offering a data-driven alternative to code-based methods and supporting safer and more consistent design of flat RC slabs.

With the rapid development of artificial intelligence technology, robotics, as its core branch, has attracted extensive attention from researchers. This paper designs and develops a robotic arm learning system based on multi-source sensor information fusion, which investigates the autonomous learning capability of robotic arms by closely integrating deep reinforcement learning (DRL) as the core framework for skill acquisition. By incorporating imitation learning as a source of expert prior and leveraging DRL’s intrinsic ability for policy optimization through environmental exploration, the proposed system achieves both rapid learning and robust generalization. Specifically, we introduce the gradient penalty mechanism from Wasserstein generative adversarial networks (WGANs), a technique that improves the stability of adversarial training by penalizing gradients that deviate from a specified norm. This mechanism is incorporated into the soft actor-critic (SAC) algorithm, a widely used off-policy DRL method known for its sample efficiency and robust performance in continuous control tasks. The resulting SAC-GP (SAC-gradient penalty) algorithm benefits from both SAC’s stable policy learning and WGAN’s improved training regularization, leading to superior convergence speed and system stability. Furthermore, this paper proposes a hybrid learning framework by combining generative adversarial imitation learning (GAIL) with SAC-GP, enabling the agent to benefit from both demonstration-based policy initialization and continuous self-improvement via reinforcement learning. Finally, a door-opening experiment is designed to verify the learning and execution capabilities of the system in both virtual and real environments. Experimental results demonstrate that the proposed learning system possesses excellent learning and motion execution abilities in practical applications. This achievement not only provides new insights for research in robot learning but also lays a solid foundation for the future development of robotic technology.

This paper investigates a specific culture of interdisciplinarity that has gained traction at the intersection of applied AI and ethics. To address social and ethical harms of AI applications, scholars have suggested importing norms, methodologies and governance frameworks from established disciplines such as the social sciences or medicine. I show how this importation presupposes and endorses a framing of applied AI as a domain separate from established disciplines. Yet, such separation is what initially allows AI practitioners to operate outside those disciplinary norms that have evolved to prevent harms now associated with AI applications. Conversely, if AI applications were understood as situated firmly within these disciplines, practitioners would already be accountable to their norms and standards. Paradoxically, this culture of interdisciplinarity might thus reinforce a problematic disciplinary isolation of applied AI underlying the very ethical issues it seeks to mitigate – fighting symptoms while playing into their cause. In response, I outline three paths forward.