High-fidelity Simulation Model Research Articles

Optimal Tilt-Wing eVTOL Takeoff Trajectory Prediction Using Regression Generative Adversarial Networks

Electric vertical takeoff and landing (eVTOL) aircraft have attracted tremendous attention nowadays due to their flexible maneuverability, precise control, cost efficiency, and low noise. The optimal takeoff trajectory design is a key component of cost-effective and passenger-friendly eVTOL systems. However, conventional design optimization is typically computationally prohibitive due to the adoption of high-fidelity simulation models in an iterative manner. Machine learning (ML) allows rapid decision making; however, new ML surrogate modeling architectures and strategies are still desired to address large-scale problems. Therefore, we showcase a novel regression generative adversarial network (regGAN) surrogate for fast interactive optimal takeoff trajectory predictions of eVTOL aircraft. The regGAN leverages generative adversarial network architectures for regression tasks with a combined loss function of a mean squared error (MSE) loss and an adversarial binary cross-entropy (BC) loss. Moreover, we introduce a surrogate-based inverse mapping concept into eVTOL optimal trajectory designs for the first time. In particular, an inverse-mapping surrogate takes design requirements (including design constraints and flight condition parameters) as input and directly predicts optimal trajectory designs, with no need to run design optimizations once trained. We demonstrated the regGAN on optimal takeoff trajectory designs for the Airbus A3 Vahana. The results revealed that regGAN outperformed reference surrogate strategies, including multi-output Gaussian processes and conditional generative adversarial network surrogates, by matching simulation-based ground truth with 99.6% relative testing accuracy using 1000 training samples. A parametric study showed that a regGAN surrogate with an MSE weight of one and a BC weight of 0.01 consistently achieved over 99.5% accuracy (denoting negligible predictive errors) using 400 training samples, while other regGAN models require at least 800 samples.

Exploiting the potential of large eddy simulations (LES) for ducted fuel injection investigation in non-reacting conditions

The diesel combustion research is increasingly focused on ducted fuel injection (DFI), a promising concept to abate engine-out soot emissions in compression-ignition engines. A large set of experiments carried out in constant volume vessel and numerical simulations, at medium-low computational cost, showed that the duct adoption in front of the injector nozzle activates several soot mitigation mechanisms, leading to quasi-zero soot formation in several engine-like operating conditions. However, although the simplified CFD modelling so far played a crucial role for the preliminary understanding of DFI technology, a more accurate turbulence description approach, combined with a large set of numerical experiments for statistical purposes, is of paramount importance for a robust knowledge of the DFI physical behaviour.In this context, the present work exploits the potential of large eddy simulations (LES) to analyse the non-reacting spray of DFI configuration compared with the unconstrained spray. For this purpose, a previously developed spray model, calibrated and validated in the RANS framework against an extensive amount of experimental data related to both free spray and DFI, has been employed. The tests have been carried out considering a single-hole injector in an optical accessible constant volume vessel, properly replicated in the simulation environment. This high-fidelity simulation model has been adapted for LES, firstly selecting the best grid settings, and then carrying out several numerical experiments for both spray configurations until achieving a satisfying statistical convergence. With this aim, the number of independent samples for the averaging procedure has been increased exploiting the axial symmetry characteristics of the present case study.Thanks to this approach, a detailed description of the main DFI-enabled soot mitigation mechanisms has been achieved, shrinking the knowledge gap in the physical understanding of the impact of spray-duct interaction.

FAST Low-Energy Beamline Studies: Toward High-Peak 5D Brightness Beams for FAST-GREENS

The FAST beamline is the injector for the planned Gamma-Ray Electron ENhanced Source (GREENS) program, which aims to achieve the demonstration and first application of a high-efficiency, high-average-power free-electron laser at 515 nm. FAST-GREENS requires high 5D peak brightness; transverse normalized projected emittances of 3 mm-mrad and a peak current of 600 A are the minimum beam requirements for the FEL to reach the 10% efficiency goal. In this work, studies of the low-energy section of the FAST beamline are presented toward these ends, including preliminary measurements of beam compression and beam emittance. An effort toward developing a high-fidelity simulation model that could be later optimized for FAST-GREENS is presented.

Self-Optimizing Vapor Compression Cycles Online With Bayesian Optimization Under Local Search Region Constraints

Abstract Self-optimizing efficiency of vapor compression cycles (VCCs) involves assigning multiple decision variables simultaneously in order to minimize power consumption while maintaining safe operating conditions. Due to the modeling complexity associated with cycle dynamics (and other smart building energy systems), online self-optimization requires algorithms that can safely and efficiently explore the search space in a derivative-free and model-agnostic manner. This makes Bayesian optimization (BO) a strong candidate for self-optimization. Unfortunately, classical BO algorithms ignore the relationship between consecutive optimizer candidates, resulting in jumps in the search space that can lead to fail-safe mechanisms being triggered, or undesired transient dynamics that violate operational constraints. To this end, we propose safe LSR-BO, a global optimization methodology that builds on the BO framework while enforcing two types of safety constraints including black-box constraints on the output and local search region (LSR) constraints on the input. We provide theoretical guarantees that under standard assumptions on the performance and constraint functions, LSR-BO guarantees constraints will be satisfied at all iterations with high probability. Furthermore, in the presence of only input LSR constraints, we show the method will converge to the true (unknown) globally optimal solution. We demonstrate the potential of our proposed LSR-BO method on a high-fidelity simulation model of a commercial vapor compression system with both LSR constraints on expansion valve positions and fan speeds, in addition to other safety constraints on discharge and evaporator temperatures.

Model predictive control of heating in a low energy single-family house

This paper presents a control scheme that uses a Model Predictive Control (MPC) approach to manage the heating system of a low-energy single-family house. The house is equipped with an air-to-water Heat Pump (HP) and individually controlled hydronic underfloor heating circuits for each of its 11 heating zones. The MPC scheme is designed to maintain individual room comfort levels in each zone, while incorporating weather forecasts and following a heating reference to allow for load shifting for periods with low energy prices and high Photovoltaic (PV) production calculated by an upper level in a hierarchical control scheme. The focus of the design has been the model structure that allows for fast solutions to the MPC optimisation problem, while still capturing the high complexity, non-linear dynamics of the building. The controller is tested on a high-fidelity simulation model of the house, achieving disturbance rejection and system stabilisation. The rapid solving time makes repeated experiments and longer simulations feasible.

Surrogate model-based calibration of a flying Earth observation satellite

At the European Space Agency (ESA), Modeling and Simulation (M&S) plays a fundamental role during the lifetime of a spacecraft, being used from the design phase to the testing and during operations in space. M&S tools embed general physics-based models and disciplines characterized by configurable parameters which have to be calibrated in order to mimic the behavior of the actual flying spacecraft. However, their calibration requires a large number of simulations which are unfeasible to be obtained through computationally expensive high-fidelity simulation models. Thus, the inability to calibrate the high-fidelity simulation models poses limitations for the use of M&S tools during spacecraft operations.In this light, the present work proposes the use of a surrogate model-based approach for the calibration of simulation models of spacecraft. The approach integrates a computationally inexpensive deep-learning-based surrogate model, which mimics the high-fidelity simulation model without requiring the same computational burden, and a metaheuristic optimization algorithm, to identify the optimal values of the simulation model configurable parameters. This enhances the capabilities of M&S tools and allows their use in operations. The approach’s effectiveness is shown by its application to real flying Earth observation satellite data and simulation models.

A52Probing for Veins: A Cost-Effective, Reproducible Method for Teaching Ultrasound-Guided Peripheral Intravenous Access

Repeated attempts at peripheral intravenous (PIV) access cause increased discomfort and delays in treatment for patients in hospital. The use of ultrasound can improve success in PIV access, benefitting patient experience. Medical professionals are increasingly expected to use ultrasound when landmark techniques fail [1]. We sought to improve confidence using ultrasound for PIV cannulation with low-cost simulation models made from ADAMgel [2] or tofu in multiple teaching sessions. We organized teaching sessions with multiple specialties at St Richard’s Hospital. A total of 48 members of staff have been taught over four sessions. Prior to the teaching session learners were asked to complete a survey identifying previous ultrasound experience and rating confidence using ultrasound for PIV access. The teaching session consisted of an introductory multimedia presentation with a video demonstration, then practical experience. Models consisted of water-filled balloons placed in ADAMgel (Aqueous Dietary Fibre Antifreeze Mix gel) or tofu to simulate human tissue and veins, Models used to deliver teaching sessions. After ADAMgel became available this has been used exclusively due to it being more reusable compared to the tofu modelsA: tofu model with ultrasound image underneath.B: ADAMgel model with ultrasound image underneath The initial survey identified only 29% of staff had previously used ultrasound for PIV access, and half of these (15% overall) had only used it once or twice a year. Using a five-point Likert scale, 74% of respondents rated their confidence in using ultrasound as one or two out of five. All respondents thought they would benefit from further teaching in PIV access. Feedback after teaching sessions was favourable, with 95% of respondents finding the session very useful and 95% also believing it would increase their use of ultrasound in clinical practice. After the session, more than half rated their confidence in ultrasound cannulation as four or five out of five. We have demonstrated that there is a desire from medical staff to increase their competency in ultrasound-guided PIV cannulation. Using low-cost, high-fidelity simulation models with a blended learning method, we can deliver teaching sessions to a large number of medical staff. We hope to continue this teaching in collaboration with our sister sites throughout Sussex to increase confidence with ultrasound-guided PIV cannulation in this region. Authors confirm that all relevant ethical standards for research conduct and dissemination have been met. The submitting author confirms that relevant ethical approval was granted, if applicable.

Feasibility of High-Fidelity Simulator Models for Minimally Invasive Spine Surgery in a Resource-Limited Setting: Experience From East Africa.

Spine surgery is a rapidly evolving specialty with a continuous need to learn new skills. In resource-limited settings such as Africa, the need for training is greater. The use of simulation-based training is important in different stages of skill acquisition, especially for high-stake procedures such as spine surgery. Among the available methods of simulation, the use of synthetic models has gained popularity among trainers. Twenty participants of a neurosurgery training course, most of whom (65%) were neurosurgery residents and fellows, were recruited. They had hands-on training sessions using a high-fidelity lumbar degenerative spine simulation model and hands-on theater experience. After this, they completed a survey to compare their experience and assess the effectiveness of the lumbar spine model in stimulating real patient and surgery experiences. The participants were from four African countries, and the majority were neurosurgery residents. There were varying levels of experience among the participants in minimally invasive spine surgery, with the majority either having no experience or having only observed the procedure. All the participants said that the high-fidelity lumbar spine model effectively simulated real minimally invasive spine setup and real bone haptics and was effective in learning new techniques. Most of the participants agreed that the model effectively simulated real dura and nerve roots (95%), real muscle (90%), real bleeding from bones and muscles (95%), and real cerbrospinal fluid in the subarachnoid space. Among them, 95% agreed that the model is effective in lumbar minimally invasive spine training in resource-limited settings. With the development of new and better surgical techniques, the use of high-fidelity models provides a good opportunity for learning and training, especially in resource-poor settings where there is a paucity of training facilities and personnel.

Indicators of surgical stress and influence of clinical experience, simulation models, and cadaveric laboratory on the stress response of third-year veterinary students performing first elective surgery.

To evaluate the role of simulation models and previous surgical experience on subjective and objective stress levels of students performing their 1st elective surgery within the veterinary curriculum. 141 third-year veterinary students. Using a pre-post experimental design, salivary alpha-amylase, and cortisol were evaluated as markers of physiologic stress response before students' first elective surgery. Student self-reported State-Trait Anxiety Inventory (STAI) scores and quantitative measures of experience were correlated to biomarker results. No association was found for change in salivary biomarkers of stress, alpha-amylase, and cortisol, between baseline and presurgical samples accounting for gender, age, type of elective surgery performed, previous surgical experience, or simulation model use. Salivary cortisol levels were markedly elevated falling between the 66th and 99th percentile compared to an age and gender-matched population. Salivary alpha-amylase levels were also 2 to 3 times higher than those recorded by other health professionals. Veterinary student STAI scores were high falling between the 65th and 73rd percentile compared to working adults in the general population. Veterinary students' salivary cortisol, alpha-amylase, and STAI scores fell into the upper 2/3rds of the general population, demonstrating a high level of stress. Simulation models and previous surgical experience were not associated with decreased stress. Further evaluation of the implementation of high-fidelity simulation models and the role of stress on performance is indicated.

In-situ observation of powder spreading in powder bed fusion metal additive manufacturing process using particle image velocimetry

Providing a dense powder bed is necessary to ensure the rationality of the final product prepared by powder bed fusion-based additive manufacturing under broad building parameters. In this study, the powder spreading mechanism was elucidated using particle image velocimetry and discrete element simulations. The particle flow regimes in the spreading process were identified based on the particle displacement: alignment, rotation, and deposition. The alignment regime was dominant in gas-atomized stainless steel 304 powder piles, whereas the rotation regime was dominant in plasma rotating electrode processed stainless steel 304 powder piles. A high-fidelity spreading simulation model was developed to clarify the critical factors resulting in the different flow regimes of different stainless steel 304 powders. The dominant rotational regime of the plasma rotating electrode processed powder was substituted for alignment by increasing the cohesive force. The particle supply in the spreading process was further suppressed by increasing the cohesive force, owing to the formation of a strong force arch and agglomeration. The high cohesive force in the gas-atomized powder was mainly attributed to the electrostatic force caused by the thick oxide film. Therefore, it was proven that the oxide film thickness is a key factor in determining the powder spreading mechanism and powder bed quality in the powder bed fusion additive manufacturing process.

An ex vivo model for education and training of bilateral cleft lip surgery

BackgroundBilateral cleft lip surgery is very challenging and requires a high level of skill, knowledge and experience. Existing high-fidelity simulation models that can be used by novice cleft surgeons to gain experience and expand their knowledge are rare and expensive. In this study, we developed a bilateral cleft lip model using porcine snout discs, which are available anywhere and inexpensive.MethodsAnatomic reference points of a patient with a bilateral cleft lip were superimposed with landmarks of the porcine snout disc on a foil template. The template was used to construct an ex vivo bilateral cleft lip model. Surgery was performed on the model according to Millard and the surgical steps were photodocumented analogous to two clinical cases of bilateral cleft lip surgery. The suitability of the model was further tested by twelve participants and evaluated using self-assessment questionnaires.ResultsThe bilateral cleft lip ex vivo model made of a porcine snout disc proved to be a suitable model with very low cost and ease of fabrication, as the template is reusable on any snout disc. The Millard procedure was successfully performed and the surgical steps of the lip plasty were simulated close to the clinical situation. Regarding the nasal reconstruction, the model lacks three-dimensionality. As a training model, it enhanced the participants comprehension of cleft surgery as well as their surgical skills. All participants rated the model as valuable for teaching and training.ConclusionsThe porcine snout discs can be used as a useful ex vivo model for bilateral cleft lip surgery with limitations in the construction of the nose, which cannot be realistically performed with the model due to anatomical differences with humans. Benefits include a realistic tissue feel, the simulation of a multi-layered lip construction, a wide and rapid availability and low cost. This allows the model to be used by novice surgeons also in low-income countries. It is therefore useful as a training model for gaining experience, but also as a model for refining, testing and evaluating surgical techniques for bilateral lip plasty.

The Evaluation of a High-Fidelity Simulation Model and Video Instruction Used to Teach Canine Dental Skills to Pre-Clinical Veterinary Students.

In recent years, there has been an increased focus on the teaching of small-animal dentistry to veterinary students in order to address the recognized gap between dental skill training and the expectations of employers regarding the competencies of new graduates in this field. In this study, third-year veterinary students were trained in three canine dental core skills using either a high-fidelity model (Group A) or video instruction (Group B). An objective structured clinical examination was used to assess skill acquisition and questionnaires were distributed in order to assess student confidence and perceptions related to small-animal dentistry practice and related skills before and after the training. All results were compared between the two groups. Group A outperformed Group B in skill acquisition (p < 0.001) and there was greater improvement in skill confidence for Group A than Group B (p < 0.001). There was no statistical difference in perceptions related to small-animal dentistry between the two groups after the training (p ≥ 0.1). Group A rated their training experience more highly than Group B (p < 0.001). Although dental skill acquisition shows greater improvement when training is provided by models rather than video instruction, a blended approach to teaching dental skills is likely to be the best approach to optimizing dental skill acquisition.

Incremental control as an enhanced and robust implementation of gain scheduled controllers avoiding hidden coupling terms

This paper proposes a new and simple method to combine incremental control with the classical proportional-integral (PI) control in order to increase the robustness properties of the resulting control law. The proposed method avoids hidden coupling terms in a similar manner to the differential PI controller implementation (DPI) and hence allows for fast gain scheduling. The integral action is accomplished by feedback of actuator measurements and therefore avoids integrator windup without complicated anti-windup schemes. The control law obtains an increase in robust performance with a response less sensitive to variations in flight dynamic parameters when compared to the equivalent PI controller. Improvement in disturbance rejection is also obtained with a faster and more robust response. Increased robustness toward input time delay is observed by example. The trade-off between increased robustness and sensitivity to sensor noise and delay can be adjusted with an intuitive tuning parameter. The novel method, denoted incremental DPI (iDPI), is investigated in the context of fixed-wing flight control. The concept is verified as a retrofit for an industry-level PI baseline controller for the longitudinal motion of a fixed-wing aircraft using a high-fidelity simulation model.

Heat and mass transfer in electron beam additive manufacturing

The electron beam wire-feed additive manufacturing (EBAM) technology utilizes a high-energy electron beam as a heat source to bombard a metal wire in the vacuum environment to form a molten pool. The stability and quality of the deposition mainly depend on the transition behavior of the droplet. To conduct this research, a high-fidelity thermo-fluid dynamic simulation model based on the local reconstruction of the 3D model was developed and verified. The model used the Monte-Carlo algorithm to update the scattering trajectories and residual energies of high-energy electrons on the irradiated surface of the wire and molten pool at each time step in the numerical calculation. Quantitatively analyzed the “electron-wire” instantaneous energy coupling mechanism in the droplet transition mode, insertion transition mode, and liquid bridge transition mode, and finally established an energy-matching criterion for maintaining a stable deposited track morphology. The results show that during the initial stage of wire melting, the droplet will climb along the wire feeding direction, grow under gravity, and eventually drop under the combined effects of gravity and recoil pressure. Different forms of molten droplets will form various solidification morphology, with the liquid bridge transition mode being the most desirable. To maintain a stable liquid bridge transition, it is necessary to control the droplet generation height within the range of 0–2.7 mm. The collapse of the deposited track, caused by heat accumulation during the multilayer deposition, would also alter the droplet generation height. Therefore, to maintain the liquid bridge transition mode in the multilayer multi-channel forming process, it is also necessary to ensure that the energy density of the deposited track is slightly greater than 1.16×1010 J/m3 and the energy density of the wire is slightly greater than 2.63×1010 J/m3.

Surrogate model-based real-time gas path fault diagnosis for gas turbines under transient conditions

Advanced diagnostic algorithms and high-fidelity simulation models improve the accuracy of model-based gas path fault diagnosis for gas turbines (GTs). But simultaneously, it becomes difficult in real-time applications due to the increased calculation amount. To improve the diagnosis speed, this study adopts the surrogate method to realize the real-time gas path fault diagnosis of GTs under transient operating conditions. First, the component level model (CLM) is built and verified. Subsequently, the surrogate model is established by combining the artificial neural network (ANN) and the necessary physical model. The constructed surrogate model can almost entirely reproduce the simulation results of CLM under the whole operation conditions. Finally, the real-time fault diagnosis system combines the surrogate model and the unscented Kalman filter (UKF). The results show that the surrogate model-based fault diagnosis system has the same accuracy as the CLM-based system. At the same time, the calculation speed is increased by nearly 50 times, which meets the real-time requirements of fault diagnosis.

Reevaluation of failure criteria location and novel improvement of 1/4 PCCV high fidelity simulation model under material uncertainty quantifications

Reactor containment buildings serve as the last barrier to prevent radioactive leakage due to accidents and their safety is crucial in overpressurization conditions. Thus, the Regulatory Guide (RG) 1.216 has mentioned the global strain as one of failure criteria in the free-field for cylindrical prestressed concrete containment vessels (PCCV) subject to internal pressure. However, there is a limit that RG 1.216 shows the free-field without the specific locations of failure criteria and also the global strain corresponding to only azimuth 135° has been mentioned in NUREG/CR-6685, regardless of the elevations of the structure. Therefore, in order to reevaluate the failure criteria of the 1:4 scaled PCCV, the high fidelity simulation model based on the experimental test was significantly validated in this study, and it was interesting to find that the experimental and numerical result was very close to each other. In addition, for the consideration of the material uncertainties, the Latin hypercube method was used as a statistical approach. Consequently, it was revealed that the radial displacements of various azimuth area such as 120°, 135°, 150°, 180° and 210° at elevations 4680 mm and 6,200 mm can represent as the global deformation at the free-field, obtained from the statistical approach.

Blackstart and Fault Ride-Through Capability of DFIG-Based Wind Turbines

Grid-forming wind turbines must be able to re-energize transmission branches, motor loads, unbalanced loads, and power electronics loads following a blackout. Further, they must ride through faults that can arise during restoration. To tackle these challenges, this paper sets forth a grid-forming control strategy for wind turbines using doubly-fed induction generators (DFIGs). In particular, a special set of feedforward control loops are engineered to mitigate the impacts of restoration transients as well as symmetrical and asymmetrical faults on the regulation of DFIG rotor currents and stator voltages. This paper also leverages two-axis anti-windup regulators to bound two-axis voltage and current commands within ratings. The controller performance is demonstrated via a high-fidelity simulation model of a notional wind-dominant grid.

PD45-04 TEMPERATURE EFFECT OF MOSES TM 2.0 DURING FLEXIBLE URETEROSCOPY: AN IN VITRO ASSESSMENT

You have accessJournal of UrologyCME1 Apr 2023PD45-04 TEMPERATURE EFFECT OF MOSESTM 2.0 DURING FLEXIBLE URETEROSCOPY: AN IN VITRO ASSESSMENT José Villena, Vicente Elorrieta, and José Salvadó José VillenaJosé Villena More articles by this author , Vicente ElorrietaVicente Elorrieta More articles by this author , and José SalvadóJosé Salvadó More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003358.04AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: One of the main issues related to the use of high-power lasers is the associated rise in temperature. The aim of this study was to characterize temperature variations with activation of the Moses 2.0 laser. METHODS: An in vitro experimental study was designed using a high-fidelity uretero-nephroscope simulation model to assess changes in temperature during intracorporeal laser lithotripsy. Renal and ureteral temperature records were obtained from the treatment of BegoStones positioned in the renal pelvis. Different laser settings over three time periods and two possible irrigation flow speeds were evaluated. We considered 43°C as the threshold since it is associated with denaturation of proteins. The Wilcoxon–Mann–Whitney test was used to assess quantitative variables and the Kruskal–Wallis test for categorical variables. RESULTS: The highest increase in intrarenal temperature was reached with 30 s of laser activation at a laser setting of 0.5 J/100 Hz (50 W) and a flow of 10 mL/min. Only 15 s of activation was sufficient for most combinations of settings to exceed the threshold of 43°C. The ureteral temperature did not increase significantly, regardless of the combination of laser setting, time, or irrigation flow, except when 30 W was used for a 30 s period. Multivariate analysis showed that an irrigation flow of 20 mL/min produced an intrarenal temperature decrease of 4.7–9.2°C (p<0.001). CONCLUSIONS: The use of high-power lasers, both for the ureter and kidney, should involve consideration of the temperature increases evidenced in this study, due to the potential biological risk entailed. Source of Funding: None © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e1163 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information José Villena More articles by this author Vicente Elorrieta More articles by this author José Salvadó More articles by this author Expand All Advertisement PDF downloadLoading ...

Cleft lip repair simulation: an Australian trainee experience

**Background**: Cleft lip repair is technically challenging surgery that requires an appreciation of the complex anatomy of the lip and nose. Effective transference of skill has been demonstrated through the use of simulations. We established a cleft lip and palate simulation workshop to improve training. In this report we assessed the confidence of plastic surgery trainees to perform a cleft lip repair before and after the simulation exercise. **Design**: We purchased 3D high-fidelity simulation models for the workshop. Participants watched a short video depicting cleft lip markings, incisions and repair. Two senior cleft surgeons provided additional commentary and teaching throughout the workshop.**Method**: A 'confidence scale' questionnaire was distributed to 11 plastic surgery trainees. The questionnaire comprised of eight questions on a Likert scale of 1–5 (total score = 40) that prompted a self-assessment of confidence, surgical skill, knowledge and level of 'anxiety' in carrying out a cleft lip repair. Questions about the level of training and prior experience were added to the survey. Participants completed the survey before and after the simulation. **Results**: Comparison of pre and post-simulation surveys showed an average increase of 53.5 per cent in total scores. Improvements were found for all attendees regardless of their level of training or prior experience. **Conclusion**: Simulation training is a validated solution that aids in the acquisition of skills and knowledge. Using simulations in the surgical curriculum has the potential to improve operative learning.

Deep-learning-generalized data-space inversion and uncertainty quantification framework for accelerating geological CO2 plume migration monitoring

Efficient coupling of high-fidelity simulation models and history-monitored data to predict future migration behavior of CO2 plume is crucial for leakage risk management and assessment during the process of geological CO2 sequestration (GCS). However, the conventional model-based inverse modeling and forecasting procedure generally presents intensive computation-cost due to numerous high-fidelity model simulations and thus hinders its practical applications. We present a novel data-driven direct forecast and uncertainty quantification framework that hybridizes deep convolutional autoencoder and deep residual neural network to represent spatial distribution of subsurface fluid by latent space features and achieve predictions of future fluid dynamic behavior given history-monitored data, respectively. The linear-type data-space inversion approaches based on Bayesian theorem are generalized to non-linear model applications through deep-learning, i.e., referred to as DSI-DL. Compared to the conventional learning-based direct forecast approach (LDFA) that numerous training samples are required to guarantee predictability of the supervised machine-learning models, our proposed DSI-DL employs a data-augmentation strategy to automatically generate sufficient training samples from a relatively small ensemble of high-fidelity model simulations. The performance DSI-DL method is demonstrated on a synthetic 2D models and a practical 3D model for deep saline aquifer GCS applications. The deep-learning surrogate directly predicts the latent features of future spatial CO2 plumes given history-monitored CO2 plumes. Comparisons between LDFA and DSI-DL confirm that the proposed DSI-DL method obtains more robust and better results in predicting future CO2 plume migration, especially when solely a few number of large-scale high-fidelity model simulations is available at hand and large observation errors exist. The proposed deep-learning-based direct forecast approach contributes to an efficient computation framework to manage uncertainty of the long term process of GCS.