In the energy and industrial process sectors, it is very important to be able to accurately and in real time estimate multiphase flowrates for safe and efficient operations. However, the advancement of data-driven soft sensors is hindered by the limited availability and significant expense of labelled field data. This study investigates transfer learning (TL) as a viable approach to address data constraints and assesses its efficacy within two deep learning architectures: Deep Neural Network (DNN) and Long Short-Term Memory (LSTM) network. The source domain for both architectures was a large, high-quality dataset collected from the T & Uuml;V-S & Uuml;D NEL wet gas flow facility. Two representative field datasets with very few samples were selected as target domains to create realistic deployment scenarios. Conditional Tabular GANs (CTGANs) are used to ensure data augmentation while preserving physical plausibility. A structured TL framework was established to examine the effects of different layerfreezing strategies on the system. The results show that TL, especially when combined with full fine-tuning, performed better than non-TL methods. TL models not only provided more precise predictions of gas and liquid flow rates but also exhibited improved generalizability to unobserved field conditions and greater conformity with the physical behaviour. In both datasets and model architectures, TL consistently lowered prediction errors (in most cases) compared to non-TL methods. It also made convergence more stable and improved robustness when there was not much data available. In contrast, using pre-trained models without any changes led to a substantial drop in performance, whereas models built from scratch were more affected by a lack of data and often had trouble converging effectively. This framework has great potential for use in fields beyond oil and gas. This also supports the digital transformation of flow diagnostics in changing multiphase conditions.

Accurate uncertainty quantification (UQ) is essential for deploying machine learning (ML) models in multiphase flow metering, where limited training data, incomplete feature representation, and distribution shifts across operating conditions introduce significant epistemic uncertainty beyond the inherent variability of the sensors and flow dynamics. Using experimental datasets under diverse multiphase conditions, this study evaluates predictive uncertainty across five ML models, including deep neural networks (DNN), long short-term memory networks (LSTM), random forests (RF), extreme gradient boosting (XGBoost), and Gaussian process regression (GPR). Conformal prediction (CP) is employed as a model-agnostic framework to generate calibrated prediction intervals (PIs), whereas GPR estimates the predictive variance through its kernel-based structure. The evaluation results show that the gas flow rate predictions exhibit high accuracy and well-calibrated intervals across the models, with the CP producing fixed PIs and the GPR achieving the narrowest, dynamically adjusted intervals. Among the CP-based models, RF demonstrated the best balance between the maximum prediction accuracy and minimal uncertainty. Liquid flowrate predictions exhibited improved epistemic uncertainty across all models. In fact, introducing mixture fluid density, a potential engineered feature derived from the gas volume fraction (GVF) and phase densities, decreased uncertainties. The analysis continued using Explainable AI platforms to highlight the importance of features based on predictive strength. The study findings emphasize the importance of both targeted feature engineering and uncertainty-aware modeling, highlighting practical advantages of CP for model-agnostic UQ and the necessity of XAI tools for transparency. Overall, these results support the applicability of explainable, uncertainty-calibrated ML systems for real-time multiphase flow monitoring, with direct implications for metering confidence and operational decision-making.

The rising cost of energy and the urgent need for sustainability have driven industries to adopt smarter solutions for monitoring and optimizing resource consumption. In this study, we present an Industrial Internet of Things (IIoT)-based approach for real-time energy and air consumption monitoring in manufacturing, focusing on a legacy Turret Punch Press (TPP) at Mitsubishi Electric Air Conditioning Systems Europe Ltd. (M-ACE). Due to its age and lack of modern monitoring capabilities, the machine was suspected to be inefficient, requiring a retrofitting strategy for improved transparency and optimization. To address these challenges, a structured IIoT-enabled monitoring system was deployed, integrating KEYENCE MP-F series sensors, an energy monitoring module, and Ethernet communication via Modbus TCP/IP. A comprehensive dashboarding system was developed for real-time visualization and analysis of energy consumption trends, identifying inefficiencies and optimizing machine usage. The data-driven approach revealed significant energy savings of up to 56% and uncovered hidden inefficiencies, including a persistent air leak. By implementing a smart shut-off valve triggered by real-time power consumption data, unnecessary air leakage was eliminated, reducing compressed air waste and overall energy costs. The results demonstrate the effectiveness of IIoT-based retrofitting for industrial energy efficiency, showcasing a scalable framework that can be applied across various machines and production environments. This study highlights the importance of data-driven decision-making in smart manufacturing, contributing to both cost reduction and sustainability goals in industrial settings.

When it comes to optimizing the efficiency of transportation pipelines, the accurate quantification of wet gas flows which are mainly comprised of gas with minor fractions of liquid - is still a relevant topic of interest. Historically, XLM (Lockhart-Martinelli parameter) has served as a key parameter in the development of overreading correlations for the accurate determination of the mass flowrate of gas with differential pressure meters. This investigation aims to offer a potential alternative to the traditional correlation-based techniques through direct prediction of the mass flowrate of gas and liquid and thus addressing the inherent complexities of wet gas metering. The present investigation suggests a novel Machine Learning (ML) modelling algorithms to improve the prediction accuracy of both the liquid and gas flowrates. In this study, four ML models are discussed in terms of their efficacy. The study results promise significant advancements in flow measurements through introducing this proposed advanced technique.

Digital twin (DT) technology has become a key enabler for prognostics and health management (PHM) in complex industrial systems, yet scaling predictive models for multi-component degradation (MCD) scenarios remains challenging, particularly when transferring insights from predictive models of smaller systems developed with limited data to larger systems. To address this, a physics-informed neural network (PINN) framework that integrates a standardized scaling methodology, enabling scalable DT analytics for MCD prognostics, was developed in this paper. Our approach employs a systematic DevOps workflow that features containerized PINN DT analytics deployed on a Kubernetes cluster for dynamic resource optimization, a real-time DT platform (PTC ThingWorx™), and a custom API for bidirectional data exchange that connects the cluster to the DT platform. A key contribution of this paper is the scalable DT model, which facilitates transfer learning of degradation patterns across heterogeneous hydraulic systems. Three (3) hydraulic system configurations were modeled, analyzing multi-component filter degradation under pump speeds of 700–900 RPM. Trained on limited data from a reference system, the scaled PINN model achieved 88.98% accuracy for initial degradation detection at 900 RPM—outperforming an unscaled baseline of 64.13%—with consistent improvements across various speeds and thresholds. This work advances PHM analytics by reducing costs and development time, providing a scalable framework for cross-system DT deployment.

Flow regimes of vertical upflow for slightly cohesive Geldart A powders at high solids mass flux (Gs≳ 500 kg/m2s) are not fully resolved. In particular, Dense Suspension Upflow (DSU) as a distinct flow regime and its transition boundaries are not broadly accepted. Furthermore, the locus of the pressure gradient minimum, which is the broadly accepted dense–dilute transition at low Gs, requires validation at high Gs. In our recent work, by adapting the phase map of Wirth and by Eulerian modeling, DSU was defined as a distinct flow regime with gross upflow of solids and with granular temperature at the wall greater than that in the bulk. This study has further validated the definition of DSU and its transition boundaries by extending the modeling to areas not fully explored in the earlier work. Furthermore, this study has identified (a) the possibility of a phase of DSU between fast fluidization and turbulent regime at all Gs; and (b) the need to review the suitability of the locus of the pressure gradient minimum as the dense–dilute transition at high Gs. Additionally, our work has demonstrated (a) a new provisional correlation that the upper transport velocity for Geldart A powders is significantly greater than hitherto predicted; and (b) the slip velocity in the transport regimes increases with Gs to peak within fast fluidization and falls thereafter to attain low multiples of the terminal settling velocity within DSU.

This study seeks to address the challenge of limited degradation data in developing Fault Detection and Isolation (FDI) models for multi-component degradation (MCD) scenarios. Utilizing a small fraction (0.05%) of a previously utilized water distribution testbed dataset in a previous publication, a weighted ensemble hybrid approach is proposed and evaluated against more established modelling approaches used in the previous publication. The proposed approach combines heuristic approximation and Physics-Informed Neural Network (PINN) methods with a recurrent neural network (RNN) model to enhance diagnostic performance for predicting MCD scenarios. The hybrid model generally outperformed other algorithms when tested on an MCD dataset, demonstrating improved diagnostic accuracy in such scenarios. Future research aims to optimize ensemble weights based on model uncertainty, further enhancing diagnostic capabilities.

Precise quantification of multiphase flow rates holds paramount significance in the context of process surveillance and enhancement in the energy sector. Conventional methodologies depend on the physical partitioning of phases prior to the use of single-phase meters, presenting a labor-intensive and economically demanding procedure. Recent developments in the field of machine learning present innovative data-driven methodologies for approximating multiphase flow rates by leveraging sensor-derived measurements. This research explores the examination of neural network architectures, specifically exploring deep neural networks (DNN) and convolutional neural networks (CNN), with the aim of predicting multiphase flow rates in Venturi tubes. Temporal data series and mean values pertaining to variables such as differential pressure, temperature, as well as throat and recovery differential pressure serve as inputs for the model. The primary objective of these data-centric methodologies is to ascertain gas and liquid flow rates directly, eliminating the need for the identification of flow patterns. Both instantaneous and time-averaged predictions are studied. Academic parlance entails subjecting models to training and testing processes using empirical datasets across diverse multiphase flow scenarios. The findings unequivocally establish the viability and efficacy of the suggested DNN and CNN architectures for addressing the complexities inherent in this demanding application. Accuracy is gauged using MSE, RMSE, MAE, and R-squared to assess the disparities between predictions and reference measurements. The enhancement of sensor inputs, customization of network architectures, and the implementation of field testing are integral aspects within the purview of Outlook. These measures are undertaken to bolster resilience across various facilities and operating conditions, thereby contributing to an augmented level of robustness.

Predictive maintenance is crucial in modern industrial settings, aiming to optimise performance, minimise downtime, and prevent costly equipment failures. The advancements in the use of machine learning techniques, sensor technologies, and data acquisition systems within Industry 4.0 has generated a lot of interest lately. In this paper, a novel predictive maintenance model is developed using machine learning approach for modelling vibration and temperature data collection for electrical motors of power press machines in Mitsubishi Electric Air Conditioning production (MACE) factory in the United Kingdom. Vibration and temperature data were collected using Bluetooth sensors installed on the motors, transmitted to a central data storage system for further analysis. To ensure the accuracy as well as the quality and reliability of the collected data sets for further analysis, pre-processing of the data was conducted, and the Isolation Forest (IF) outlier detection method was employed to filter out the anomalies. Machine learning algorithms including Auto-Regressive Integrated Moving Average (ARIMA), Random Forest (RF), and Long Short-Term Memory (LSTM) networks were employed for predicting vibration signals, with hyperparameter tuning conducted using Sequential Model-Based optimisation (SMBO) with the Tree Parzen Estimator (TPE).The results and concluding remarks presented in this paper show how the performance of the optimized ARIMA model can be used in predicting future vibration levels of the electrical motor. The residual analysis is also used to monitor discrepancies between predicted and observed vibration values, enabling proactive identification of emerging issues.

Venturi tubes are differential pressure meters widely used for wet-gas flow measurement and are covered by international standards. A Venturi meter will overread the quantity of gas flowing in a pipe if there is a small amount of liquid present in the flow and its output should be suitably corrected to give accurate gas flow rate measurements. Numerous correlations were developed in the past 40 years to correct the Venturi tube over-reading response; however, they all require the amount of liquid flowing into the pipe to be known as an input to the correlation. The Lockhart-Martinelli parameter is an a-dimensional parameter generally used to represent the liquid amount and is an input to several overreading correlations. Unfortunately, the Lockhart-Martinelli parameter is generally unknown in the field and should be measured by adding additional instrumentation to the Venturi tube or adding additional devices in series to the Venturi tube. In this regard, one relatively inexpensive and simple method is to obtain the Lockhart-Martinelli parameter (XLM) by measuring the pressure loss ratio (PLR) response, i.e., XLM = f(PLR). However, the PLR method currently works for a narrow range at low liquid loading. At high liquid loading, the PLR response becomes insensitive to changes in the amount of liquid. A potential way to address this issue and expand the applicability of the PLR method to higher liquid loading is by modifying the Venturi tube geometry design. This study presents a 2D Computational Fluid Dynamics (CFD) study, using ANSYS Fluent, of the wet-gas flow through a standard ISO-compliant Venturi tube and a nonstandard Venturi tube under multiple operating conditions, with a critical focus on a parametric study of the Venturi tube geometry. The simulated fluids are nitrogen-water and nitrogen-kerosene under numerous flow conditions. First, a 3-inch beta 0.6 Venturi tube with a divergent angle of 7° was simulated. The simulation results for the standard Venturi tube were compared against experimental results. Then the Venturi tube divergent section was modified to be a sudden expansion (90° divergent angle), and the simulations were repeated. Finally, additional simulations were performed simulating numerous divergent angles (3°, 5°, 15°, 30°, and 60°) to investigate the sensitivity of the PLR response to a change in divergent angle.

In the development of analytics for PHM applications, a lot of emphasis has been placed on data transformation for optimal model development without enough consideration for the repeatability of the measurement systems producing the data. This paper explores the relationship between data quality, defined as the measurement system analysis (MSA) process, and the performance of fault detection and isolation (FDI) algorithms within smart infrastructure systems. This research employs a comprehensive methodology, starting with an MSA process for data-quality evaluation and leading to the development and evaluation of fault detection and isolation (FDI) algorithms. During the MSA phase, the repeatability of a water distribution system’s measurement system is examined to characterise variations within the system. A data-quality process is defined to gauge data quality. Synthetic data are introduced with varying data-quality levels to investigate their impact on FDI algorithm development. Key findings reveal the complex relationship between data quality and FDI algorithm performance. Synthetic data, even with lower quality, can improve the performance of statistical process control (SPC) models, whereas data-driven approaches benefit from high-quality datasets. The study underscores the importance of customising FDI algorithms based on data quality. A framework for instantiating the MSA process for IIoT applications is also suggested. By bridging data-quality assessment with data-driven FDI, this research contributes to the design of digital twins for IIoT-enabled smart infrastructure systems. Further research on the practical implementation of the MSA process for edge analytics for PHM applications will be considered as part of our future research.

A cost-effective alternative for lowering carbon emissions from building heating is the use of flat-plate solar collectors (FPSCs). However, low thermal efficiency is a significant barrier to their effective implementation. Favorable nanofluids’ thermophysical properties have the potential to increase FPSCs’ effectiveness. Accordingly, this study evaluates the performance of an FPSC operating with Fe3O4-water nanofluid in terms of its thermo-hydraulic characteristics with operating parameters ranging from 303 to 333 K for the collector inlet temperature, 0.0167 to 0.05 kg/s for the mass flow rate, and 0.1 to 2% for nanoparticles’ volume fraction, respectively. The numerical findings demonstrated that under identical operating conditions, increasing the volume fraction up to 2% resulted in an improvement of 4.28% and 8.90% in energy and energy efficiency, respectively. However, a 13.51% and 7.93% rise in the friction factor and pressure drop, respectively, have also been observed. As a result, the performance index (PI) criteria were used to determine the optimal volume fraction (0.5%) of Fe3O4 nanoparticles, which enhanced the convective heat transfer, exergy efficiency, and energy efficiency by 12.90%, 4.33%, and 2.64%, respectively.

The purpose of this work is to deepen our understanding of natural convection with large Prandtl number fluids and to resolve some controversies in the previous publications. To achieve this purpose, a new thermal multiple-relaxation-time lattice Boltzmann model is proposed. Natural convection in a square cavity, a benchmark test case, is investigated numerically using the new model. The Prandtl number is up to 100. For the first time, it is numerically observed that there are two critical Prandtl numbers in the natural convection, which will affect the correlation between the Nusselt number and Prandtl number critically. Three heat transfer characteristic ranges of natural convection are defined in this work, according to the two critical Prandtl numbers. In each range, the dominant heat transfer mechanism is different, which can solve a long-standing issue in the discipline of heat and mass transfer: completely opposing statements on the correlation between the Nusselt number and Prandtl number for natural convection, were published in the open literature. For the first time, this work reveals cause behind the controversial reports and provides the guidance for the future research.

Pneumatic conveying is a well-established technology within a wide range of industrial sectors. This chapter outlines the fundamentals of gas–solids flow in pipelines and details three simulation methods: single and two fluid models and combined CFD–DEM. The principal theories and equations commonly available in commercial off-the-self and open-source software packages are given. Example results of pressure drop, volume fraction, etc. are provided.

Rapid development of smart manufacturing techniques in recent years is influencing production facilities. Factories must both keep up with smart technologies as well as upskill their workforce to remain competitive. One of the recent concerns is how businesses can contribute to environmental sustainability and how to reduce operating costs. In this article authors present a method of measuring gas waste using Industrial Internet of Things (IIoT) sensors and open-source solutions utilised on a brownfield production asset. The article provides a result of an applied research initiative in a live manufacturing facility. The design followed the Reference Architectural Model for Industry 4.0 (RAMI 4.0) model to provide a coherent smart factory system. The presented solution's goal is to provide factory supervisors with information about gas waste which is generated during the production process. To achieve this an operational technology (OT) network was installed and Key Performance Indicators (KPIs) dashboards were designed. Based on the information provided by the system, the business can be more aware of the production environment and can improve its efficiency.

Vertical Axis Wind Turbines (VAWTs) are omni-directional, low-cost, low-efficiency wind power extractors. A conventional drag-based VAWT consists of multiple thin rotor blades with a typical peak Tip Speed Ratio (λ) of < 1. Their lower cut-in speed and maintenance cost make them ideal for power generation in urban environments. Numerous studies have been carried out analysing steady operation of VAWTs and quantifying their performance characteristics, however, minimal attention has been paid to their start-up dynamics. There are a few recent studies in which start-up dynamics of lift-based VAWTs have been analysed but such studies for drag-based VAWTs are severely limited. In this study, start-up dynamics of a conventional multi-blade drag-based VAWT have been numerically investigated using a time-dependant Computational Fluid Dynamics (CFD) solver. In order to enhance the start-up characteristics of the drag-based VAWT, a stator has been integrated in the design assembly. The numerical results obtained in this study indicate that an appropriately designed stator can significantly enhance the start-up of a VAWT by directing the flow towards the rotor blades, leading to higher rotational velocity (ω) and λ. With the addition of a stator, the flow fields downstream the VAWT becomes more uniform.

Early fault detection in production is crucial for manufacturing facilities to prevent unplanned downtimes and maximise the operational life of equipment. The aim of this paper is to present a method of anomaly detection for an in-service motor using self-supervised learning. The authors have developed a condition monitoring system for a Smart Factory using deep autoencoders. The system was installed in a live production facility with the goal of improving site maintenance.

In recent times, the oil and Gas industry has faced many challenges resulting from a tightening climate policy environment on oil and gas exploration, as well as the increasing risk of oversupply due to new discoveries globally. This has given stakeholders in the industry an incentive to integrate new technologies to optimize the operational efficiency of their assets, leading to the optimal recovery of hydrocarbons especially in marginal fields. Various Original Equipment Manufacturers (OEM) now provide different service offerings using data driven methods to provide condition monitoring of assets for oil and gas operators. However, a significant part of the value proposition by OEMs in their service delivery focuses on value generated at the component level with a reduction in asset downtime. This limits the broad economic benefits that a condition-based approach can provide, at the enterprise level. Therefore, the purpose of this paper is to develop a cost benefit analysis framework for assessing the implementation of condition and performance monitoring of oil and gas assets used in surface applications. The framework utilizes a combined technical-economic approach to determine a minimum predictive requirement for the implementation of condition-based principles to maintenance of assets in a hydrocarbon project from first oil to abandonment. This financial analysis framework uses a condition monitoring approach based on prognostics as well as a regression approach for fault detection and system performance. The paper will present a case study to evaluate the costs and benefits associated with implementing a condition-based maintenance approach for a set of valves in a Christmas tree subsystem, as part of a typical onshore production system. The framework illustrated using the case study compares a constant failure rate Time-Based approach to the PHM enabled condition-based maintenance. The results demonstrate that a prognostic enabled system can provide commercial benefits at the component level for a Condition Based Maintenance strategy but not necessarily at the enterprise level for an oil and gas project. The cumulative reduction in downtime at the component level over a project lifecycle offsetting the present value of the total cost of integrating a PHM enabled system into the overall maintenance strategy creates the ideal situation for commercial viability. However, the commercial viability of the PHM integration would depend on the accuracy of predicting failure events and monitoring asset degradation by the PHM enabled system which ultimately defines the performance of the condition-based maintenance approach. The accuracy level of asset failure therefore provides OEMs with a benchmark for executing their condition-based maintenance services with a minimum performance threshold. Secondly, an enterprise level financial viability, as well as OEM profitability in the implementation of a condition-based maintenance approach, requires an Optimal Service Point (OSP) which is a function of the minimum predictive requirement of the PHM system. The utility that the OSP provides is that, it gives the minimum value of the framework’s decision criteria that an operator can use a basis for incorporating a condition-based approach in its maintenance strategy. It also provides the maximum Annual Service Fee (ASF) derived from the cumulative OEM NPV needed for structuring and pricing servitization agreements with operators. This OSP cost benefit analysis approach ultimately provides OEMs and operators with a practical guide in the provision as well as the adoption of condition-based maintenance strategies respectively. It balances the risk of PHM integration by operators with a minimum PHM system performance threshold required for commercial viability for project lifecycle.

The stability of flying of a hummingbird-like flapping-wing micro air vehicle (MAV) has been challenging. In this paper, experimental studies are reported on the tail shapes of hummingbird-like flapping-wing MAVs, since tails play an important role in-flight stability. Dynamics parameters of hummingbird tails are firstly studied and evaluated. Then man-made tails inspired by the natural hummingbirds are designed, manufactured and optimized for experimental tests. The results show that lift generated by the tail is independent of a fan angle, whereas the pitch moment is related to the fan angle. Further, the tail can be applied to stabilising hovering twin-wing flapping wing MAVs.

Molecular dynamics (MD) simulation is an advanced method in microscale modelling of material but it depends on the complexity of the model. The performance of MD simulation is poor once the model size is huge. To accelerate the computing of MD simulation, the Markov state model (MSM) can be applied because of the ability to predict a future state of a stochastic system. With the advantage of MSM and MD applied in material modelling, a good result could be expected where the time scale limitation of MD simulation is bridged by MSM method. In this research, an MSM method based on the MD microstates in which a nickel superalloy's atomic model arrangements and their microstructure evolution have been treated with the Markov properties is presented. This MSM is based and classified by a dislocation model which is a fundamental of the microstructural tessellation evolution. The results indicate that the microstructure evolution in a situation of energy minimisation favours the formation of new faults alongside existing ones. And dislocation accumulation on the grain boundary was observed during fatigue resolving. Some dislocations formatted and grown in the middle of coarse grain and penetrated through the grain.