RESUMEN

Machine learning (ML) methods present a valuable opportunity for modeling the non-linear behavior of the injection molding process. They have the potential to predict how various process and material parameters affect the quality of the resulting parts. However, the dynamic nature of the injection molding process and the challenges associated with collecting process data remain significant obstacles for the application of ML methods. To address this, within this study, hybrid approaches are compared that combine process data with additional process knowledge, such as constitutive equations and high-fidelity numerical simulations. The hybrid modeling approaches include feature learning, fine-tuning, delta-modeling, preprocessing, and using physical constraints, as well as combinations of the individual approaches. To train and validate the hybrid models, both the experimental and simulated shrinkage data of an injection-molded part are utilized. While all hybrid approaches outperform the purely data-based model, the fine-tuning approach yields the best result in the simulation setting. The combination of calibrating a physical model (feature learning) and incorporating it implicitly into the training process (physical constraints) outperforms the other approaches in the experimental setting.

RESUMEN

In directed energy deposition (DED), accurately controlling and predicting melt pool characteristics is essential for ensuring desired material qualities and geometric accuracies. This paper introduces a robust surrogate model based on recurrent neural network (RNN) architectures-Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Gated Recurrent Unit (GRU). Leveraging a time series dataset from multi-physics simulations and a three-factor, three-level experimental design, the model accurately predicts melt pool peak temperatures, lengths, widths, and depths under varying conditions. RNN algorithms, particularly Bi-LSTM, demonstrate high predictive accuracy, with an R-square of 0.983 for melt pool peak temperatures. For melt pool geometry, the GRU-based model excels, achieving R-square values above 0.88 and reducing computation time by at least 29%, showcasing its accuracy and efficiency. The RNN-based surrogate model built in this research enhances understanding of melt pool dynamics and supports precise DED system setups.

RESUMEN

Ebola disease is a lethal viral hemorrhagic fever caused by ebolaviruses within the Filoviridae family with mortality rates of up to 90%. Monoclonal antibody (mAb) based therapies have shown great potential for the treatment of EVD. However, the potential emerging ebolavirus isolates and the negative effect of decoy protein on the therapeutic efficacy of antibodies highlight the necessity of developing novel antibodies to counter the threat of Ebola. Here, 11 fully human mAbs were isolated from transgenic mice immunized with GP protein and recombinant vesicular stomatitis virus-bearing GP (rVSV-EBOV GP). These mAbs were divided into five groups according to their germline genes and exhibited differential binding activities and neutralization capabilities. In particular, mAbs 8G6, 2A4, and 5H4 were cross-reactive and bound at least three ebolavirus glycoproteins. mAb 4C1 not only exhibited neutralizing activity but no cross-reaction with sGP. mAb 7D8 exhibited the strongest neutralizing capacity. Further analysis on the critical residues for the bindings of 4C1 and 8G6 to GPs was conducted using antibodies complementarity-determining regions (CDRs) alanine scanning. It has been shown that light chain CDR3 played a crucial role in binding and neutralization and that any mutation in CDRs could not improve the binding of 4C1 to sGP. Importantly, mAbs 7D8, 8G6, and 4C1 provided complete protections against EBOV infection in a hamster lethal challenge model when administered 12 h post-infection. These results support mAbs 7D8, 8G6, and 4C1 as potent antibody candidates for further investigations and pave the way for further developments of therapies and vaccines.

Asunto(s)

RESUMEN

For identifying the optimal model for real-time conflict prediction, there is a necessity for proposing a quantitative analysis approach that adaptively selects the optimal prediction model from a large pool of task-suited models, while simultaneously considering the computational efficiency and prediction precision. Based on this line, this study developed an innovative approach termed surrogate model-based optimal prediction model selection (SM-OPMS). This approach aims to accelerate the optimal model selection while incorporating prediction precision considerations, under the precondition of comprehensively evaluating task-suited models. An analytical framework was proposed, further illustrated through a detailed case study. In the case study, real vehicle trajectory data from HighD were processed and applied, which can be aggregated to extract both traffic state variables and corresponding conflict data during a specific time interval. As for the conflict detection, Time-to-Collision (TTC) and Deceleration Rate to Avoid a Crash (DRAC) indicators were utilized to identify risky conditions. Based on the proposed approach, the selection for the optimal prediction model was conducted, and the variable importance in conflict prediction within the optimal models derived from the SM-OPMS was also investigated. Finally, a comparative analysis with the enumeration-based optimal prediction model selection (E-OPMS) approach was conducted to validate the superiority of the proposed approach. Results indicate that SM-OPMS outperforms E-OPMS in optimal model selection, notably enhancing computational efficiency by up to 94.03%, while maintaining prediction precision within a maximum reduction of only 7.91%. The significance of the SM-OPMS approach is revealed by its comprehensive selection of the optimal prediction models for specific traffic scenarios, taking into account both prediction efficiency and precision simultaneously. The proposed approach is expected to contribute to the development of real-time conflict prediction in the future.

Asunto(s)

RESUMEN

In space missions, heating films are crucial for uniformly heating onboard equipment for precise temperature control. This study develops an optimization method using surrogate models for lightweight anisotropic substrate thermal conductive heating films, meeting the requirements of uniform heating in thermal control for space applications. A feedforward neural network optimized by particle swarm optimization (PSO) was employed to create a surrogate model, mapping design parameters to the temperature uniformity of the heating film. This model served as the basis for applying the NSGA-II algorithm to quickly optimize both temperature uniformity and lightweight characteristics. In this study, the PSO-BP surrogate model was trained using heating film thermal simulation data, and the surrogate model demonstrated an accurate prediction of the mean square error (MSE) of the predicted temperature difference within 0.0168 s. The maximum temperature difference in the optimal model is 1.188 ℃, which is 30.5 times lower than before optimization, and the equivalent density is only increased by 3.9%. In summary, this optimization design method effectively captures the relationships among various parameters and optimization objectives. Its superior computational accuracy and design efficiency offer significant advantages in the design of devices such as heating films.

RESUMEN

An intelligent optimization technique has been presented to enhance the multiple structural performance of PA6-20CF carbon fiber-reinforced polymer (CFRP) plastic injection molding (PIM) products. This approach integrates a deep neural network (DNN), Non-dominated Sorting Genetic Algorithm II (NSGA-II), and Monte Carlo simulation (MCS), collectively referred to as the DNN-GA-MCS strategy. The main objective is to ascertain complex process parameters while elucidating the intrinsic relationships between processing methods and material properties. To realize this, a numerical study on the PIM structural performance of an automotive front engine hood panel was conducted, considering fiber orientation tensor (FOT), warpage, and equivalent plastic strain (PEEQ). The mold temperature, melt temperature, packing pressure, packing time, injection time, cooling temperature, and cooling time were employed as design variables. Subsequently, multiple objective optimizations of the molding process parameters were employed by GA. The utilization of Z-score normalization metrics provided a robust framework for evaluating the comprehensive objective function. The numerical target response in PIM is extremely intricate, but the stability offered by the DNN-GA-MCS strategy ensures precision for accurate results. The enhancement effect of global and local multi-objectives on the molded polymer-metal hybrid (PMH) front hood panel was verified, and the numerical results showed that this strategy can quickly and accurately select the optimal process parameter settings. Compared with the training set mean value, the objectives were increased by 8.63%, 6.61%, and 9.75%, respectively. Compared to the full AA 5083 hood panel scenario, our design reduces weight by 16.67%, and achievements of 92.54%, 93.75%, and 106.85% were obtained in lateral, longitudinal, and torsional strain energy, respectively. In summary, our proposed methodology demonstrates considerable potential in improving the, highlighting its significant impact on the optimization of structural performance.

RESUMEN

This paper investigates a novel approach to efficiently construct and improve surrogate models in problems with high-dimensional input and output. In this approach, the principal components and corresponding features of the high-dimensional output are first identified. For each feature, the active subspace technique is used to identify a corresponding low-dimensional subspace of the input domain; then a surrogate model is built for each feature in its corresponding active subspace. A low-dimensional adaptive learning strategy is proposed to identify training samples to improve the surrogate model. In contrast to existing adaptive learning methods that focus on a scalar output or a small number of outputs, this paper addresses adaptive learning with high-dimensional input and output, with a novel learning function that balances exploration and exploitation, i.e., considering unexplored regions and high-error regions, respectively. The adaptive learning is in terms of the active variables in the low-dimensional space, and the newly added training samples can be easily mapped back to the original space for running the expensive physics model. The proposed method is demonstrated for the numerical simulation of an additive manufacturing part, with a high-dimensional field output quantity of interest (residual stress) in the component that has spatial variability due to the stochastic nature of multiple input variables (including process variables and material properties). Various factors in the adaptive learning process are investigated, including the number of training samples, range and distribution of the adaptive training samples, contributions of various errors, and the importance of exploration versus exploitation in the learning function.

RESUMEN

Surrogate endpoints are becoming increasingly important in health technology assessment, where decisions are based on complex cost-effectiveness models (CEMs) that require numerous input parameters. Daniels and Hughes Surrogate Model was used to predict missing effect estimates in randomized controlled trials (RCTs) evaluating first-line treatments in metastatic castration-resistant prostate cancer (mCRPC) patients. Network meta-analyses (NMAs) were conducted to assess the comparative efficacy of these treatments. Databases were searched (inception to October 2022) using Ovid®. Several grey literature searches were also conducted (PROSPERO: CRD42021283512). Available trial data for radiographic progression-free survival (rPFS) and overall survival (OS) were used to predict the unreported effect of rPFS or OS for relevant comparator treatments. Bayesian NMAs were conducted using observed and predicted treatment effects. Effect estimates and 95% credible intervals were calculated for each comparison. Mean ranks and the probability of being best (p-best) were obtained. Twenty-five RCTs met the eligibility criteria and of these, 8 reported jointly rPFS and OS; while rPFS was predicted for 12 RCTs and 10 comparators, and OS was predicted for 5 RCTs and 6 comparators. A nonstandard dose of docetaxel (docetaxel 50 mg/m2 every 2 weeks) had the highest probability of being the most effective for rPFS (p-best: 59%) and OS (p-best: 48%), followed by talazoparib plus enzalutamide (13% and 19%, respectively). Advanced surrogate modelling techniques allowed obtaining relevant parameter and indirect estimates of previously unavailable data and may be used to populate future CEMs requiring rPFS and OS in first-line mCRPC.

Asunto(s)

RESUMEN

As magnetic field strength in Magnetic Resonance Imaging (MRI) technology increases, maintaining the specific absorption rate (SAR) within safe limits across human head tissues becomes challenging due to the formation of standing waves at a shortened wavelength. Compounding this challenge is the uncertainty in the dielectric properties of head tissues, which notably affects the SAR induced by the radiofrequency (RF) coils in an ultra-high-field (UHF) MRI system. To this end, this study introduces a computational framework to quantify the impacts of uncertainties in head tissues' dielectric properties on the induced SAR. The framework employs a surrogate model-assisted Monte Carlo (MC) technique, efficiently generating surrogate models of MRI observables (electric fields and SAR) and utilizing them to compute SAR statistics. Particularly, the framework leverages a high-dimensional model representation technique, which constructs the surrogate models of the MRI observables via univariate and bivariate component functions, approximated through generalized polynomial chaos expansions. The numerical results demonstrate the efficiency of the proposed technique, requiring significantly fewer deterministic simulations compared with traditional MC methods and other surrogate model-assisted MC techniques utilizing machine learning algorithms, all while maintaining high accuracy in SAR statistics. Specifically, the proposed framework constructs surrogate models of a local SAR with an average relative error of 0.28% using 289 simulations, outperforming the machine learning-based surrogate modeling techniques considered in this study. Furthermore, the SAR statistics obtained by the proposed framework reveal fluctuations of up to 30% in SAR values within specific head regions. These findings highlight the critical importance of considering dielectric property uncertainties to ensure MRI safety, particularly in 7 T MRI systems.

RESUMEN

The application of optimal control theory in practical engineering is often limited by the modeling cost and complexity of the mathematical model of the controlled plant, and various constraints. To bridge the gap between the theory and practice, this paper proposes a model-free direct method based on the sequential sampling and updating of surrogate model, and extends the ability of direct method to solve model-free optimal control problems with general constraints. The algorithm selects sample points from the current actual trajectory data to update the surrogate model of controlled plant, and solve the optimal control problem of the constantly refined surrogate model until the result converges. The presented initial and subsequent sampling strategies eliminate the dependence on the model. Furthermore, the new stopping criteria ensure the overlap of final actual and planned trajectories. The several examples illustrate that the presented algorithm can obtain constrained solutions with greater accuracy and require fewer sample data.

RESUMEN

We propose a genetic algorithm for optimizing oil skimmer assignments, introducing a tailored repair operation for constrained assignments. Methods essentially involve simulation-based evaluation to ensure adherence to South Korea's regulations. Results show that the optimized assignments, compared to current ones, reduced work time on average and led to a significant reduction in total skimmer capacity. Additionally, we present a deep neural network-based surrogate model, greatly enhancing efficiency compared to simulation-based optimization. Addressing inefficiencies in mobilizing locations that store oil skimmers, further optimization aimed to minimize mobilized locations and was validated through scenario-based simulations resembling actual situations. Based on major oil spills in South Korea, this strategy significantly reduced work time and required locations. These findings demonstrate the effectiveness of the proposed genetic algorithm and mobilized location minimization strategy in enhancing oil spill response operations.

RESUMEN

An intelligent optimization technology was proposed to mitigate prevalent multi-defects, particularly failure, wrinkling, and springback in sheet metal forming. This method combined deep neural networks (DNNs), genetic algorithms (GAs), and Monte Carlo simulation (MCS), collectively as DNN-GA-MCS. Our primary aim was to determine intricate process parameters while elucidating the intricate relationship between processing methodologies and material properties. To achieve this goal, variable blank holder force (VBHF) trajectories were implemented into five sub-stroke steps, facilitating adjustments to the blank holder force via numerical simulations with an oil pan model. The Forming Limit Diagram (FLD) predicted by machine learning algorithms based on the Generalized Incremental Stress State Dependent Damage (GISSMO) model provided a robust framework for evaluating sheet failure dynamics during the stamping process. Numerical results confirmed significant improvements in formed quality: compared with the average value of training sets, the improvements of 18.89%, 13.59%, and 14.26% are achieved in failure, wrinkling, and springback; in the purposed two-segmented mode VBHF case application, the average value of three defects is improved by 12.62%, and the total summation of VBHF is reduced by 14.07%. Statistical methodologies grounded in material flow analysis were applied, accompanied by the proposal of distinctive optimization strategies for the die structure aimed at enhancing material flow efficiency. In conclusion, our advanced methodology exhibits considerable potential to improve sheet metal forming processes, highlighting its significant effect on defect reduction.

RESUMEN

The advent of digital twins facilitates the generation of high-fidelity replicas of actual systems or assets, thereby enhancing the design's performance and feasibility. When developing digital twins, precise measurement data is essential to ensure alignment between the actual and digital models. However, inherent uncertainties in sensors and models lead to disparities between observed and predicted (simulated) behaviors. To mitigate these uncertainties, this study originally proposes a multi-objective optimization strategy utilizing a Gaussian process regression surrogate model, which integrates various uncertain parameters, such as load angle, bucket cylinder stroke, arm cylinder stroke, and boom cylinder stroke. This optimization employs a genetic algorithm to indicate the Pareto frontiers regarding the pressure exerted on the boom, arm, and bucket cylinders. Subsequently, TOPSIS is applied to ascertain the optimal candidate among the identified Pareto optima. The findings reveal a substantial congruence between the experimental and numerical outcomes of the devised virtual model, in conjunction with the TOPSIS-derived optimal parameter configuration.

RESUMEN

Crash scenario-based testing is crucial for assessing autonomous driving safety. However, existing studies on scenario generation tend to prioritize concrete scenarios for direct testing, neglecting the construction of fundamentally functional scenarios with a broader range. Police-reported historical crash data is a valuable supplement, yet detecting all potential crash scenarios is laborious. In order to address this issue, this study proposes an adaptive search sampling framework based on deep generative model and surrogate model (SM) to extract master scenario samples from police-reported historical crash data. The framework starts with selecting representative samples from the full crash dataset as initial master scenario samples using various sampling techniques. Evaluation indexes are then constructed, and derived scenario samples are synthesized using the deep generative model. To enhance efficiency, an SM is established to replace the generative model's training and data generation process. Based on the SM, an adaptive search sampling method is developed, which iteratively adjusts the sampling strategy using the Similarity Score to achieve comprehensive sampling. Experimental results demonstrate the notable advantage of the adaptive search sampling method over other sampling methods. Furthermore, statistical analysis and visualization assessments confirm the effectiveness and accuracy of the proposed method.

Asunto(s)

RESUMEN

The clinical performance of biodegradable polymer stents implanted in blood vessels is affected by uneven degradation. Stress distribution plays an important role in polymer degradation, and local stress concentration leads to the premature fracture of stents. Numerical simulations combined with in vitro experimental validation can accurately describe the degradation process and perform structural optimization. Compared with traditional design techniques, optimization based on surrogate models is more scientifically effective. Three stent structures were designed and optimized, with the effective working time during degradation as the optimization goal. The finite element method was employed to simulate the degradation process of the stent. Surrogate models were employed to establish the functional relationship between the design parameters and the degradation performance. The proposed function models accurately predicted the degradation performance of various stents. The optimized stent structures demonstrated improved degradation performance, with the kriging model showing a better optimization effect. This study provided a novel approach for optimizing the structural design of biodegradable polymer stents to enhance degradation performance.

RESUMEN

This study aims to enhance the degradation uniformity of PLGA sinus stents to minimize fracture risk caused by stress corrosion. Symmetric stent structures were introduced and compared to sinusoidal structure in terms of stress and degradation uniformity during implantation and degradation processes. Three surrogate models were employed to optimize the honeycomb-like structure. Results showed honeycomb-like structures exhibited the superior stress distribution and highest degradation uniformity. The kriging model achieved the smallest error and degradation uniformity of 83.24%. In conclusion, enhancing the symmetry of stent structures improves degradation uniformity, and the kriging model has potential for the optimization of stent structures.

RESUMEN

The human tongue has highly variable morphology. Its role in regulating respiratory flows and deposition of inhaled aerosols remains unclear. The objective of this study was to quantify the uncertainty of nanoparticle deposition from the variability in tongue shapes and positions and to rank the importance of these morphological factors. Oropharyngeal models with different tongue postures were reconstructed by modifying an existent anatomically accurate upper airway geometry. An LRN k-ω model was applied to solve the multiregime flows, and the Lagrangian tracking approach with near-wall treatment was used to simulate the behavior and fate of inhaled aerosols. Once the database of deposition rates was completed, a surrogate model was trained using Gaussian process regression with polynomial kernels and was validated by comparing its predictions to new CFD simulations. Input sensitivity analysis and output updateability quantification were then performed using the surrogate model. Results show that particle size is the most significant parameter in determining nanoparticle deposition in the upper airway. Among the morphological factors, the shape variations in the central tongue had a higher impact on the total deposition than those in the back tongue and glottal aperture. When considering subregional deposition, mixed sensitivity levels were observed among morphological factors, with the back tongue being the major factor for throat deposition and the central tongue for oral deposition. Interaction effects between flow rate and morphological factors were much higher than the effects from individual parameters and were most significant in the throat (pharyngolaryngeal region). Given input normal variances, the nanoparticle deposition exhibits logarithmical normal distributions, with much lower uncertainty in 100-nm than 2-nm aerosols.

RESUMEN

Individuals typically spend most of their lives indoors, predominantly in spaces like offices and residences. Consequently, prolonged indoor exposure underscores the critical significance of maintaining optimal indoor air quality (IAQ) to safeguard one's health. The primary impediment to attaining efficient regulation of Indoor Air Quality (IAQ) is the challenge of monitoring the IAQ parameters, particularly within the immediate vicinity of an individual's breathing space. Current heating, ventilation, and air conditioning systems lack the ability to rapidly predict and optimize the quality of indoor air. The objective of this study is to acquire the distribution features of indoor pollutants and precise indoor environment data in order to efficiently forecast and enhance the IAQ. To achieve this objective, a proposed surrogate model was developed using computational fluid dynamics (CFD). Notably, the Kriging surrogate model can rapidly predict IAQ while using a limited number of CFD runs. CFD are widely used as numerical simulation methods to obtain the accurate information. Surrogate models can rapidly forecast indoor environmental conditions using CFD simulation data simultaneously. Optimization algorithms can efficiently achieve desirable indoor ambient conditions, offering highly effective and intelligent control techniques for indoor atmospheric ventilation.

Asunto(s)

RESUMEN

Inverse mechanical parameter identification enables the characterization of ultrasoft materials, for which it is difficult to achieve homogeneous deformation states. However, this usually involves high computational costs that are mainly determined by the complexity of the forward model. While simulation methods like finite element models can capture nearly arbitrary geometries and implement involved constitutive equations, they are also computationally expensive. Machine learning models, such as neural networks, can help mitigate this problem when they are used as surrogate models replacing the complex high fidelity models. Thereby, they serve as a reduced order model after an initial training phase, where they learn the relation of in- and outputs of the high fidelity model. The generation of the required training data is computationally expensive due to the necessary simulation runs. Here, active learning techniques enable the selection of the "most rewarding" training points in terms of estimated gained accuracy for the trained model. In this work, we present a recurrent neural network that can well approximate the output of a viscoelastic finite element simulation while significantly speeding up the evaluation times. Additionally, we use Monte-Carlo dropout based active learning to identify highly informative training data. Finally, we showcase the potential of the developed pipeline by identifying viscoelastic material parameters for human brain tissue.

RESUMEN

Precision medicine is a promising framework for generating evidence to improve health and health care. Yet, a gap persists between the ever-growing number of statistical precision medicine strategies for evidence generation and implementation in real-world clinical settings, and the strategies for closing this gap will likely be context-dependent. In this paper, we consider the specific context of partial compliance to wound management among patients with peripheral artery disease. Using a Gaussian process surrogate for the value function, we show the feasibility of using Bayesian optimization to learn optimal individualized treatment rules. Further, we expand beyond the common precision medicine task of learning an optimal individualized treatment rule to the characterization of classes of individualized treatment rules and show how those findings can be translated into clinical contexts.

Asunto(s)