Uncertainty Quantification Techniques Research Articles

Sustainable Design of Pavements: Predicting Pavement Service Life

Pavement service life is an important factor that affects both the whole life cost and carbon footprint of a pavement. The service life of a pavement is affected by several different parameters which can be broadly classified into climate conditions, binder and mixture properties, pavement design, workmanship, and maintenance strategies. The current practice for determining service life of pavements involves the use of pavement design tools, which are used while constructing a new pavement or performing a reconstruction/resurfacing or pavement maintenance. In addition, field measurements using ground penetration radar, falling weight deflectometer, traffic speed deflectometer, and other techniques are also used to assess the condition of an existing pavement. The information from these measurements is then combined with pavement design software to predict potential pavement service life. The accuracy of the predicted pavement service life is affected by the associated uncertainties in the parameters that affect pavement life. The following paper presents various approaches that could be potentially used to determine the associated uncertainties in the estimation of pavement service life. The various uncertainty quantification techniques have been applied to a specific design, and the outcomes are discussed in this paper. The Monte Carlo simulation method, a system-level uncertainty quantification technique, can estimate a probabilistic pavement service life. The other uncertainty quantification schemes are software specific and provide probabilistic life factors by assumed statistical distributions. Hence, the Monte Carlo simulation technique could be one potential method that can be used for estimating a generalized pavement service utilizing predictions from various design software.

Uncertainty quantification in multi-class image classification using chest X-ray images of COVID-19 and pneumonia

This paper investigates uncertainty quantification (UQ) techniques in multi-class classification of chest X-ray images (COVID-19, Pneumonia, and Normal). We evaluate Bayesian Neural Networks (BNN) and the Deep Neural Network with UQ (DNN with UQ) techniques, including Monte Carlo dropout, Ensemble Bayesian Neural Network (EBNN), Ensemble Monte Carlo (EMC) dropout, across different evaluation metrics. Our analysis reveals that DNN with UQ, especially EBNN and EMC dropout, consistently outperform BNNs. For example, in Class 0 vs. All, EBNN achieved a UAcc of 92.6%, UAUC-ROC of 95.0%, and a Brier Score of 0.157, significantly surpassing BNN's performance. Similarly, EMC Dropout excelled in Class 1 vs. All with a UAcc of 83.5%, UAUC-ROC of 95.8%, and a Brier Score of 0.165. These advanced models demonstrated higher accuracy, better discriaminative capability, and more accurate probabilistic predictions. Our findings highlight the efficacy of DNN with UQ in enhancing model reliability and interpretability, making them highly suitable for critical healthcare applications like chest X-ray imageQ6 classification.

Surrogate based design space exploration and exploitation for an efficient airfoil optimization under uncertainties using transition models

The pursuit of sustainable, zero-emission air travel is heavily dependent on the creation of energy-efficient aircraft. Key strategies for achieving this sustainability in aviation include reducing fuel consumption through low-drag designs harnessing laminar flow. However, designing aircraft with laminar flow characteristics is complex due to their sensitivity to environmental and operational factors. This study tackles the challenge of developing energy-efficient aircraft by using computational fluid dynamics models and sophisticated optimization techniques that account for uncertainty. Our approach demonstrates the effectiveness of surrogate-based optimization and uncertainty quantification in optimizing airfoil drag for a natural laminar airfoil (NLF) design. We use surrogate models, trained with data from detailed airfoil simulations, which include a boundary layer code coupled with a linear stability method and a newly developed transition transport model. Transition location predicted using transition models facilitate an accurate drag prediction used in the optimization process. The accuracy of these surrogate models is enhanced through active sampling strategies. Our robust optimization method considers uncertainties in environmental and operational conditions, offering a deeper insight into their effects on crucial design parameters. Unlike traditional deterministic aerodynamic design optimization, our findings highlight the efficacy and precision of uncertainty-based optimization in achieving robust NLF airfoil designs over large (exploration mode) and small (exploitation mode) design spaces. Investigating design space parameterization based on the size of design variables reveals significant differences in optimal airfoil configurations. The optimized designs we propose favor delayed transition, in contrast to deterministic designs which often result in significant loss of laminarity when facing uncertainties. This study represents a significant advancement in aerospace engineering, providing a practical and effective methodology for creating energy-efficient airfoil designs. The application of these advanced optimization and uncertainty quantification techniques shows great potential for the wider field of aerospace engineering, paving the way for more resilient and robust aircraft designs.

Optimization and data mining for shock-induced mixing enhancement inside scramjet using stochastic deep-learning flowfield prediction

One of the significant challenges in supersonic combustion ramjet engines lies in effective mixing of the fuel, primarily due to the high momentum of supersonic inflow at the combustor. Among the various fluid phenomena associated with fuel mixing, it is well recognized that the interaction between the fuel-injection jet plume and oblique shock waves can significantly enhance mixing efficiency. However, the most suitable interaction for optimal mixing enhancement remains yet to be clarified. The present study conducts model-based optimization and data mining for shock-induced mixing enhancement of angled-slot injection in a two-dimensional scramjet combustor. Stochastic deep-learning flowfield prediction has been utilized to enable fast and reliable evaluations of a substantial number of designs. Prediction errors can be estimated without requiring correct data owing to uncertainty quantification techniques. Data mining and sensitivity analysis, coupled with flowfield prediction, have revealed the optimal shock interaction with the fuel jet plume characterized by a pronounced downstream recirculation region. The mechanism that drives mixing enhancement through this recirculation region has been discussed based on the results of optimization and sensitivity analysis. This study has yielded valuable insights for the future design of scramjet injectors. Furthermore, the effectiveness of the model-based design and analysis has been demonstrated through the present study, showcasing its potential for guiding future developments in scramjet technology.

Application of uncertainty quantification techniques in the framework of process safety studies: Advanced dispersion simulations

AbstractIn the framework of process safety studies, consequence assessment of accidental scenarios is a crucial step affecting the eventual risk profile associated with the facilities under analysis. Conventional models used for consequence assessment are based on integral models, and may not be adequate to cope with the dynamic evolution of accidental scenarios and their three‐dimensional features. On the other hand, consequence assessment models based on computational fluid dynamics (CFD) approaches are promising to cope with complex scenarios and environments, but setting the simulation introduces relevant uncertainties associated with both the input data, assumptions, and with the modelling of physical effects involved. In the present study, uncertainty quantification (UQ) techniques are applied to support advanced safety studies based on CFD simulations of hazardous gas dispersion. Firstly, the accidental scenarios are characterized by defining release scenarios and conditions and quantifying source terms using integral models. At the same time, input meteorological data are gathered. This enables the development of high‐fidelity CFD simulations of gas dispersion based on different input sets and eventually the implementation of UQ techniques. The generalized polynomial chaos (gPC) expansion is employed to obtain hazardous gas concentration based on the variation of wind direction and speed. The present method is applied for the analysis of a real plant featuring a complex layout. The results show the advantages of the present approach by quantifying the influence of meteorological conditions and providing indications for supporting the development of protection systems and emergency measures.

Measuring Residual Stresses with Crack Compliance Methods: An Ill-Posed Inverse Problem with a Closed-Form Kernel

By means of relaxation methods, residual stresses can be obtained by introducing a progressive cut or a hole in a specimen and by measuring and elaborating the strains or displacements that are consequently produced. If the cut can be considered a controlled crack-like defect, by leveraging Bueckner’s superposition principle, the relaxed strains can be modeled through a weighted integral of the residual stress relieved by the cut. To evaluate residual stresses, an integral equation must be solved. From a practical point of view, the solution is usually based on a discretization technique that transforms the integral equation into a linear system of algebraic equations, whose solutions can be easily obtained, at least from a computational point of view. However, the linear system is often significantly ill-conditioned. In this paper, it is shown that its ill-conditioning is actually a consequence of a much deeper property of the underlying integral equation, which is reflected also in the discretized setting. In fact, the original problem is ill-posed. The ill-posedness is anything but a mathematical sophistry; indeed, it profoundly affects the properties of the discretized system too. In particular, it induces the so-called bias–variance tradeoff, a property that affects many experimental procedures, in which the analyst is forced to introduce some bias in order to obtain a solution that is not overwhelmed by measurement noise. In turn, unless it is backed up by sound and reasonable physical assumptions on some properties of the solution, the introduced bias is potentially infinite and impairs every uncertainty quantification technique. To support these topics, an illustrative numerical example using the crack compliance (also known as slitting) method is presented. The availability of the Linear Elastic Fracture Mechanics Weight Function for the problem allows for a completely analytical formulation of the original integral equation by which bias due to the numerical approximation of the physical model is prevented.

Unsupervised stochastic learning and reduced order modeling for global sensitivity analysis in cardiac electrophysiology models

Background and Objective:Numerical simulations in electrocardiology are often affected by various uncertainties inherited from the lack of precise knowledge regarding input values including those related to the cardiac cell model, domain geometry, and boundary or initial conditions used in the mathematical modeling. Conventional techniques for uncertainty quantification in modeling electrical activities of the heart encounter significant challenges, primarily due to the high computational costs associated with fine temporal and spatial scales. Additionally, the need for numerous model evaluations to quantify ubiquitous uncertainties increases the computational challenges even further. Methods:In the present study, we propose a non-intrusive surrogate model to perform uncertainty quantification and global sensitivity analysis in cardiac electrophysiology models. The proposed method combines an unsupervised machine learning technique with the polynomial chaos expansion to reconstruct a surrogate model for the propagation and quantification of uncertainties in the electrical activity of the heart. The proposed methodology not only accurately quantifies uncertainties at a very low computational cost but more importantly, it captures the targeted quantity of interest as either the whole spatial field or the whole temporal period. In order to perform sensitivity analysis, aggregated Sobol indices are estimated directly from the spectral mode of the polynomial chaos expansion. Results:We conduct Uncertainty Quantification (UQ) and global Sensitivity Analysis (SA) considering both spatial and temporal variations, rather than limiting the analysis to specific Quantities of Interest (QoIs). To assess the comprehensive performance of our methodology in simulating cardiac electrical activity, we utilize the monodomain model. Additionally, sensitivity analysis is performed on the parameters of the Mitchell-Schaeffer cell model. Conclusions:Unlike conventional techniques for uncertainty quantification in modeling electrical activities, the proposed methodology performs at a low computational cost the sensitivity analysis on the cardiac electrical activity parameters. The results are fully reproducible and easily accessible, while the proposed reduced-order model represents a significant contribution to enhancing global sensitivity analysis in cardiac electrophysiology.

Machine learning techniques for predictive modelling and uncertainty quantification of the mechanical properties of woven carbon fibre composites

Carbon fibre composites owe their exceptional mechanical properties to the characteristics of their constituents; fibres and resin. However, uncertainty in these individual constituent properties can introduce significant variations in the macroscale mechanical performance of composites. These variations are amplified in the presence of various plies in the composite and pose a challenge to quantify during manufacturing stages. In this study, uncertain constituent material properties are propagated to the macroscale composite through multiscale homogenisation to compute the varied resultant thin-shell stiffness matrices. An in-house-built automated simulation environment has been used to create a physics-conforming trainable database for model training, testing, validation and uncertainty quantification using Gaussian Process Regression (GPR), Artificial Neural Networks (ANNs), and Multiple Linear Regression (MLR). Through rigorous comparative analyses, the superior performance of GPR is demonstrated when compared to alternative techniques with R² ≥ 0.99 and NRMSE < 10⁻⁷. Further, the findings approve the validity of GPR for extrapolatory predictions. For instance, a randomly sampled [-5, 5] % variation of constituent material properties leads to extreme uncertainties in macroscale ± 0.067 %, for axial stiffness with excellent GPR predictive accuracy of R² = 0.99. This slightly increases to ± 0.14 % for a variation of [-10, 10] %. Moreover, the extreme uncertainties indicate that the direct axial and bending stiffnesses are more sensitive to variations in constituent properties than others. Additional investigations presented in the paper produce predictive models and uncertainty bounds for all stiffness entries that are of high importance for macroscale property control during manufacturing.

Nuclear safety Enhanced: A Deep dive into current and future RAVEN applications

As the horizon of nuclear energy expands with the advent of small modular reactors, Generation IV reactors, and fusion reactors, there is a growing perspective that the licensing process could benefit from a more comprehensive approach. Moving beyond traditional deterministic and probabilistic risk assessment analyses might pave the way for a novel safety analysis paradigm propelled by the increasing computational power at our disposal.This paper explores different methodologies that can improve the outcomes of nuclear safety analysis. These range from uncertainty quantification techniques, aimed at enhancing the precision of safety margins, to deploying dynamic event trees by driving system code simulations, capturing the potential evolutions of severe accidents. These methodologies offer a better understanding of the management and consequences of nuclear accident scenarios, significantly improving the accuracy and efficiency of safety predictions compared to traditional methods.Specific case studies illustrate the practical application of these advanced techniques, demonstrating substantial improvements in predicting and managing the dynamics of severe accidents. These findings underscore the effectiveness of these methodologies in enhancing risk assessment capabilities and informing decision-making processes for nuclear safety management. The paper also emphasizes the importance of adaptability and continuous evolution, a call for action to address emerging nuclear safety concerns and highlights the utility of advanced tools like RAVEN.

Autobalanced Multitask Node Embedding Framework for Intelligent Education.

Recently, online education has become popular. Many e-learning platforms have been launched with various intelligent services aimed at improving the learning efficiency and effectiveness of learners. Graphs are used to describe the pairwise relations between entities, and the node embedding technique is the foundation of many intelligent services, which have received increasing attention from researchers. However, the graph in the intelligent education scenario has three noteworthy properties, namely, heterogeneity, evolution, and lopsidedness, which makes it challenging to implement ecumenical node embedding methods on it. In this article, an autobalanced multitask node embedding model is proposed, named MNE, and applied to the interaction graph, settling a few actual tasks in intelligent education. More specifically, MNE builds two purpose-built self-supervised node embedding learning tasks for heterogeneous evolutive graphs. Edge-specific reconstruction tasks are built according to the semantic information and properties of the heterogeneous edges, and an evolutive weight regression task is designed, aiding the model to perceive the evolution of learners' implicit cognitive states. Then, both aleatoric and epistemic uncertainty quantification techniques are introduced, achieving both task- and node-level weight estimation and instructing subtask autobalancing. Experimental results on real-world datasets indicate that the proposed model outperforms the state-of-the-art graph embedding methods on two assessment tasks and demonstrates the validity of the proposed multitask framework and subtask balancing mechanism. Our implementations are available at https://github.com/ccnu-mathits/MNE4HEN.

Reliability-Based Topology Optimization with a Proportional Topology for Reliability

This research proposes an efficient technique for reliability-based topology optimization (RBTO), which deals with uncertainty and employs proportional topology optimization (PTO) to achieve the optimal reliability structure. The recent technique, called proportional topology optimization for reliability (PTOr), uses Latin hypercube sampling (LHS) for uncertainty quantification. The difficulty of the double-loop nested problem in uncertainty quantification (UQ) with LHS can be alleviated by the power of PTO, enabling RBTO to be performed easily. The rigorous advantage of PTOr is its ability to accomplish topology optimization (TO) without gradient information, making it faster than TO with evolutionary algorithms. Particularly, for reliability-based topology design, evolutionary techniques often fail to achieve satisfactory results compared to gradient-based techniques. Unlike recent PTOr advancement, which enhances the RBTO performance, this achievement was previously unattainable. Test problems, including an aircraft pylon, reveal its performances. Furthermore, the proposed efficient framework facilitates easy integration with other uncertainty quantification techniques, increasing its performance in uncertainty quantification. Lastly, this research provides computer programs for the newcomer studying cutting-edge knowledge in engineering design, including UQ, TO, and RBTO, in a simple manner.

Empirical validation of Conformal Prediction for trustworthy skin lesions classification

Background and objective:Uncertainty quantification is a pivotal field that contributes to realizing reliable and robust systems. It becomes instrumental in fortifying safe decisions by providing complementary information, particularly within high-risk applications. existing studies have explored various methods that often operate under specific assumptions or necessitate substantial modifications to the network architecture to effectively account for uncertainties. The objective of this paper is to study Conformal Prediction, an emerging distribution-free uncertainty quantification technique, and provide a comprehensive understanding of the advantages and limitations inherent in various methods within the medical imaging field. Methods:In this study, we developed Conformal Prediction, Monte Carlo Dropout, and Evidential Deep Learning approaches to assess uncertainty quantification in deep neural networks. The effectiveness of these methods is evaluated using three public medical imaging datasets focused on detecting pigmented skin lesions and blood cell types. Results:The experimental results demonstrate a significant enhancement in uncertainty quantification with the utilization of the Conformal Prediction method, surpassing the performance of the other two methods. Furthermore, the results present insights into the effectiveness of each uncertainty method in handling Out-of-Distribution samples from domain-shifted datasets. Our code is available at: github.com/jfayyad/ConformalDx. Conclusions:Our conclusion highlights a robust and consistent performance of conformal prediction across diverse testing conditions. This positions it as the preferred choice for decision-making in safety-critical applications.

An Optimized Uncertainty-Aware Training Framework for Neural Networks.

Uncertainty quantification (UQ) for predictions generated by neural networks (NNs) is of vital importance in safety-critical applications. An ideal model is supposed to generate low uncertainty for correct predictions and high uncertainty for incorrect predictions. The main focus of state-of-the-art training algorithms is to optimize the NN parameters to improve the accuracy-related metrics. Training based on uncertainty metrics has been fully ignored or overlooked in the literature. This article introduces a novel uncertainty-aware training algorithm for classification tasks. A novel predictive uncertainty estimate-based objective function is defined and optimized using the stochastic gradient descent method. This new multiobjective loss function covers both accuracy and uncertainty accuracy (UA) simultaneously during training. The performance of the proposed training framework is compared from different aspects with other UQ techniques for different benchmarks. The obtained results demonstrate the effectiveness of the proposed framework for developing the NN models capable of generating reliable uncertainty estimates.

Deep learning‐based extraction of surface wave dispersion curves from seismic shot gathers

AbstractMulti‐channel analysis of surface waves is a seismic method employed to obtain useful information about shear‐wave velocities in the near surface. A fundamental step in this methodology is the extraction of dispersion curves from dispersion spectra, with the latter usually obtained by applying specific processing algorithms onto the recorded shot gathers. Although the extraction process can be automated to some extent, it usually requires extensive quality control, which can be arduous for large datasets. We present a novel approach that leverages deep learning to identify a direct mapping between seismic shot gathers and their associated dispersion curves (both fundamental and first higher order modes), therefore by‐passing the need to compute dispersion spectra. Given a site of interest, a set of 1D compressional and shear velocities and density models are created using prior knowledge of the local geology; pairs of seismic shot gathers and Rayleigh‐wave phase dispersion curves are then numerically modelled and used to train a simplified residual network. The proposed approach is shown to achieve high‐quality predictions of dispersion curves on a synthetic test dataset and is, ultimately, successfully deployed on a field dataset. Various uncertainty quantification and convolutional neural network visualization techniques are also presented to assess the quality of the inference process and better understand the underlying learning process of the network. The predicted dispersion curves are inverted for both the synthetic and field data; in the latter case, the resulting shear‐wave velocity model is plausible and consistent with prior geological knowledge of the area. Finally, a comparison between the manually picked fundamental modes with the predictions from our model allows for a benchmark of the performance of the proposed workflow.

Experimental Covariance Determination for Critical Integral Experiments

Integral benchmarks for criticality safety and nuclear data validation require expensive uncertainty quantification studies. In general, uncertainty quantification techniques ignore correlations between experiments and shared components. Experiments, such as the Thermal/Epithermal eXperiments (TEX) campaigns, consist of many shared components, such as the Jemima highly enriched uranium (HEU) fuel plates, which create a strong correlation in their uncertainties. While these correlations are known to exist, they are often not estimated because of the complexity of such calculations. This paper describes an intuitive method of determining the covariance for each of the experimental components, providing a correlation for each family of components across the multiple cases examined within a benchmark. A proof-of-principle study using the TEX-HEU experimental campaign was performed and verified that the covariance and correlation matrices can be calculated with information commonly found in the International Criticality Safety Benchmark Evaluation Project benchmarks. This study showed that the introduction of model and experimental covariances reduces the χ2 per degree of freedom from 2.203 to 1.179, indicating that the omission causes overly pessimistic bias quantifications. This technique can be seamlessly integrated to current benchmark evaluations as well as reevaluations of legacy benchmarks.

A unifying perspective on non-stationary kernels for deeper Gaussian processes

The Gaussian process (GP) is a popular statistical technique for stochastic function approximation and uncertainty quantification from data. GPs have been adopted into the realm of machine learning (ML) in the last two decades because of their superior prediction abilities, especially in data-sparse scenarios, and their inherent ability to provide robust uncertainty estimates. Even so, their performance highly depends on intricate customizations of the core methodology, which often leads to dissatisfaction among practitioners when standard setups and off-the-shelf software tools are being deployed. Arguably, the most important building block of a GP is the kernel function, which assumes the role of a covariance operator. Stationary kernels of the Matérn class are used in the vast majority of applied studies; poor prediction performance and unrealistic uncertainty quantification are often the consequences. Non-stationary kernels show improved performance but are rarely used due to their more complicated functional form and the associated effort and expertise needed to define and tune them optimally. In this perspective, we want to help ML practitioners make sense of some of the most common forms of non-stationarity for Gaussian processes. We show a variety of kernels in action using representative datasets, carefully study their properties, and compare their performances. Based on our findings, we propose a new kernel that combines some of the identified advantages of existing kernels.

Constraining the carbon cycle in JULES-ES-1.0

Abstract. Land surface models are an important tool in the study of climate change and its impacts, but their use can be hampered by uncertainties in input parameter settings and by errors in the models. We apply uncertainty quantification (UQ) techniques to constrain the input parameter space and corresponding historical simulations of JULES-ES-1.0 (Joint UK Land Environment Simulator Earth System), the land surface component of the UK Earth System Model, UKESM1.0. We use an ensemble of historical simulations of the land surface model to rule out ensemble members and corresponding input parameter settings that do not match modern observations of the land surface and carbon cycle. As JULES-ES-1.0 is computationally expensive, we use a cheap statistical proxy termed an emulator, trained on the ensemble of model runs, to rule out parts of the parameter space where the simulator has not yet been run. We use history matching, an iterated approach to constraining JULES-ES-1.0, running an initial ensemble and training the emulator, before choosing a second wave of ensemble members consistent with historical land surface observations. We successfully rule out 88 % of the initial input parameter space as being statistically inconsistent with observed land surface behaviour. The result is a set of historical simulations and a constrained input space that are statistically consistent with observations. Furthermore, we use sensitivity analysis to identify the most (and least) important input parameters for controlling the global output of JULES-ES-1.0 and provide information on how parameters might be varied to improve the performance of the model and eliminate model biases.

Uncertainty Quantification of Soil Organic Carbon Estimation from Remote Sensing Data with Conformal Prediction

Soil organic carbon (SOC) contents and stocks provide valuable insights into soil health, nutrient cycling, greenhouse gas emissions, and overall ecosystem productivity. Given this, remote sensing data coupled with advanced machine learning (ML) techniques have eased SOC level estimation while revealing its patterns across different ecosystems. However, despite these advances, the intricacies of training reliable and yet certain SOC models for specific end-users remain a great challenge. To address this, we need robust SOC uncertainty quantification techniques. Here, we introduce a methodology that leverages conformal prediction to address the uncertainty in estimating SOC contents while using remote sensing data. Conformal prediction generates statistically reliable uncertainty intervals for predictions made by ML models. Our analysis, performed on the LUCAS dataset in Europe and incorporating a suite of relevant environmental covariates, underscores the efficacy of integrating conformal prediction with another ML model, specifically random forest. In addition, we conducted a comparative assessment of our results against prevalent uncertainty quantification methods for SOC prediction, employing different evaluation metrics to assess both model uncertainty and accuracy. Our methodology showcases the utility of the generated prediction sets as informative indicators of uncertainty. These sets accurately identify samples that pose prediction challenges, providing valuable insights for end-users seeking reliable predictions in the complexities of SOC estimation.

A quantitative comparison study for structural flexibility identification using Accelerometric and computer vision-based vibration data

For civil engineering structures, the dynamics-based structural health monitoring is commonly based on inverse analysis of structural acceleration response data to identify the structural parameters. As the measurement of structural displacement response becomes more feasible and cost-effective thanks to advances in computer vision techniques, a comparison of the displacement and acceleration response data for structural identification has become of considerable interest among researchers. This paper presents a comprehensive analysis and comparison of structural flexibility identification based on displacement versus acceleration response data through uncertainty quantification technique. Firstly, theoretical derivation reveals that the energy spectral density of acceleration response is equal to the fourth power of the frequency of the energy spectral density of displacement response, which affects the distribution of measurement noise across different frequency bands. Secondly, the uncertainty quantification method for modal flexibility from subspace-based system identification is utilized to compare the derivation process of covariance estimation on modal flexibility. Finally, numerical analysis and experimental example are carried out to compare the structural flexibility identification results when using the displacement versus acceleration. It shows that more precise identification results of lower-order modal parameters are obtained when using displacement, which further improves the precision of modal flexibility identification. The displacement is more suitable for identifying the modal flexibility when the measurement noises are at the same level.

Axial Compressor Map Generation Leveraging Autonomous Self-Training Artificial Intelligence. Phase 2

Abstract This paper discusses a study performed by SoftInWay as part of a Phase II SBIR project funded by NASA. In contrast with the Phase I project (Burlaka and Moroz, 2023, “Axial Compressor Map Generation Leveraging Autonomous Self-Training Artificial Intelligence,” ASME J. Eng. Gas Turbines Power, 145(1), p. 011001) where three discrete compressors were considered, the Phase II study was focused on addressing the problem of axial compressor long development time and cost with the use of AI models capable of predicting the geometry and performance of various multistage axial compressors with multiple variable vanes. The applicability of the AI models to various compressors enables the opportunity to avoid iterations between engine cycle analysis and compressor design. In this paper, automated compressor design and performance generation workflows are described. The approach for autonomous selection of the architectures and hyperparameters of Machine Learning (ML) models is explained. The uncertainty quantification techniques are considered. The developed ML-powered methods for compressor geometry prediction are discussed. The ML models' accuracy values and representations of typical geometry and performance predictions are given. The utilization of the ML models in engine cycle analysis is discussed.