Evgeny Burnaev | Skolkovo Institute of Science and Technology (original) (raw)

Papers by Evgeny Burnaev

ArXiv, 2021

YUEFAN SHEN, The State Key Lab of CAD&CG, Zhejiang University HONGBO FU, The School of Creative M... more YUEFAN SHEN, The State Key Lab of CAD&CG, Zhejiang University HONGBO FU, The School of Creative Media, City University of Hong Kong ZHONGSHUO DU and XIANG CHEN, The State Key Lab of CAD&CG, Zhejiang University EVGENY BURNAEV, Skolkovo Institute of Science and Technology DENIS ZORIN, New York University KUN ZHOU, The State Key Lab of CAD&CG, Zhejiang University YOUYI ZHENG∗, The State Key Lab of CAD&CG, Zhejiang University

The choice of shape parameterisation enormously impact on the design space and optimal solution i... more The choice of shape parameterisation enormously impact on the design space and optimal solution in the aerodynamic optimisation. Three parameterisation methods, PARSEC, the Class/Shape Function Transformation (CST) and MACROS Dimension Reduction (DR), which is a novel parameterisation method, are studied in this paper. Comparison studies of these methods are performed in terms of accuracy of inverse fitting and effect on constructing design space. The results show that MACROS DR has excellent capability of dimension reduction and significantly high accuracy of inverse fitting. The CST and PARSEC methods can provide higher flexibility than MACROS DR comparing their design space.

In Japanese. About Construction of Low Dimensional Structures from Object Surface Description usi... more In Japanese. About Construction of Low Dimensional Structures from Object Surface Description using Feature Extraction Techniques, i.e. so-called Effective Dimension Reduction technique to parameterise geometry of an airfoil in order to better predict its aerodynamic characteristics and perform surrogate based optimization.

Surrogate modeling is widely used in many engineering problems. Data sets often have Car-tesian p... more Surrogate modeling is widely used in many engineering problems. Data sets often have Car-tesian product structure (for instance factorial design of experiments with missing points). In such case the size of the data set can be very large. Therefore, one of the most popular algorithms for approxi-mation–Gaussian Process regression–can be hardly applied due to its computational complexity. In this paper a computationally efficient approach for constructing Gaussian Process regression in case of data sets with Cartesian product structure is presented. Efficiency is achieved by using a special structure of the data set and operations with tensors. Proposed algorithm has low computational as well as memory complexity compared to existing algorithms. In this work we also introduce a regular-ization procedure allowing to take into account anisotropy of the data set and avoid degeneracy of regression model.

В настоящее время одним из основных средств уменьшения сроков проектирова- ния и снижения затрат ... more В настоящее время одним из основных средств уменьшения сроков проектирова- ния и снижения затрат на разработку различных образцов как гражданской, так и военной техни- ки является использование систем автоматизиро- ванного проектирования: трехмерного проекти- рования (CAD), моделирования и инженерного анализа (CAE), управления данными об изде- лии (PDM) и др. Как известно, CAD-системы дают возможность создать 3D-образ объекта, но в большинстве случаев не позволяют определить совокупности физических свойств проектируе- мого изделия. Эта задача решается с помощью CAE-систем, которые по имеющейся модели из- делия позволяют рассчитать его технические и эксплуатационные характеристики. Однако большинство CAE-систем не приближает облик изделия и его характеристики к оговоренному в техническом задании – конструктор совместно с инженером-расчетчиком на основе анализа ре- зультатов, полученных в CAE-системе, должны самостоятельно определить новый облик изде-
лия.
Таким образом, для эффективного решения
задач по созданию новых образцов техники не- обходимо автоматизировать сам процесс поиска оптимального облика и внутренних свойств изде- лия. Для этого нужно связать CAD- и CAE-сис- темы, создав единую среду, а также применить формализованные методики научного поиска, используя методы оптимизации и анализа дан- ных.
Программный комплекс pSeven позволяет решить эти задачи. В частности, с его помощью можно интегрировать различные CAПР и исполь- зовать их совместно с современным математичес- ким аппаратом – алгоритмами анализа данных и оптимизации, автоматизируя тем самым про- цессы инженерного анализа и оптимизации пара- метров изделия. Лежащая в основе pSeven алго- ритмическая библиотека MACROS включает в себя развитый инструментарий для проведения оптимизации и анализа данных:
• передовые математические методы интеллекту- ального анализа данных, снижения размерности, анализа чувствительности;
• современные высокоэффективные алгоритмы оптимизации, которые позволяют решать слож- нейшие оптимизационные задачи за минимальное время и число итераций;
• инструменты для построения метамоделей по данным как натурных, так и вычислительных экспериментов, а также для анализа построенных моделей. Использование программного комплекса pSeven влечет за собой улучшение технических характеристик проектируемых изделий и умень- шение числа дорогостоящих натурных и ресурсо- емких вычислительных экспериментов.
Пользовательский интерфейс пакета pSeven представлен как графической средой (pSeven GUI), так и командной оболочкой (pSeven Shell). Оба ин- терфейса полностью эквивалентны функционально и дают пользователю доступ ко всем возможностям комплекса. В дальнейшем речь будет идти только о графической среде pSeven GUI.

One of the activities of the EUROCOPTER flight test department is to measure loads for different ... more One of the activities of the EUROCOPTER flight test department is to measure loads for different components (through load gauges) and flight configurations. In this paper, a flight configuration includes both the manoeuvre and general information about the helicopter itself (helicopter weights, longitudinal/lateral centre of gravity locations, altitudes) and the air characteristics in which the helicopter is flying (Outside Air Temperature). A flight configuration is described by the Flight Configuration Parameters (FCP). Information, obtained from this activity, is stored in a specific database. In the scope of CHAMALO (acronym of Calculation of Helicopter Approximated MAcros LOads) project, EUROCOPTER is interested in building surrogate models from the existing load database to estimate missing loads from FCP. Both the objective and challenge is to automatically build accurate and robust surrogate models from the load database, which explain the relations between the input FCP and the output static and dynamic loads. MACROS tool (developed by DATADVANCE) is used for surrogate modelling. This tool includes a wide range of well-known techniques (e.g., Splines, Linear Regression, Gaussian Process Regression), original techniques (e.g., HDA-High Dimensional Approximation), an automatic selection of the appropriate approximation type based on built-in decision tree and data properties, and a flexible support for accelerated training, smoothing, handling multiple output components, etc. Once the best surrogate models for each load gauge (66 in total) and manoeuvre family (32 in total) were constructed, CHAMALO software was developed by DATADVANCE for automatic prediction of helicopter static and dynamic loads as a function of FCP. It is concluded that EUROCOPTER considers this approach to be very promising. In fact, about 50% of missing loads, which need to be estimated, may be calculated by CHAMALO with a sufficient accuracy, drastically reducing the time and manpower needed for such analysis. Further studies are planned to increase this percentage.

In this paper we examine problem of predictive maintenance in complex technical systems. We propo... more In this paper we examine problem of predictive maintenance in complex technical systems. We propose two approaches for anticipation of rare events (typically faults): 1) degradation detection and trending, and 2) failure discrimination based on classification techniques. The applicability of the approaches is illustrated on the real-world test cases from aircraft operations based on the data granted by AIRBUS.

In order to construct a nonlinear regression model we have to accurately (in some sense) initiali... more In order to construct a nonlinear regression model we have to accurately (in some sense) initialize parameters of the model. In this work we performed comparison of several widely used methods and several newly developed approached for initialization of parameters of a regression model, represented as a decom􏱵 position in a linear dictionary of some parametric functions (sigmoids). We proposed a general deterministic approach for initialization, providing repeatability of results, reduction of a learning time and in some cases increase of a regression model accuracy; we developed two new algorithms (based on a piecewise􏱵linear approximation and based on local properties of approximable dependency) in the framework of the proposed approach; we developed randomized initialization algorithm (spherical initialization) for effective approxi􏱵 mation of high􏱵dimensional dependencies; we improved the classical initialization method SCAWI (by locat􏱵 ing centers of sigmoids in sample points), providing a regression model accuracy improvement on specific classes of dependencies (smooth functions and discontinuous functions with a number of local peculiarities in an input domain) when using RProp algorithm for learning; we performed comparison of classical and newly proposed initialization methods and highlighted the most efficient ones.

When aircraft composite structures (fuselage, wing panels, etc.) are designed, stability analyses... more When aircraft composite structures (fuselage, wing panels, etc.) are designed, stability analyses is performed for different geometries, loadings, composites. For laminated composites the feasibility rules and the discreteness of orientations lead to tremendous amount of combinations that should be analysed in order to prevent instability and fulfil strength criteria. Population-based optimization algorithms is often used for automation of this design process. Unlike gradient-based methods, such algorithms require many function evaluations making design process intractable due to significant computational cost of stability analysis. Computational burden can be decreased by using surrogate models. Since many stability criteria are defined as minimum of several modes, then the response function has discontinuities. Therefore surrogate models should adapt to such peculiarities in order to accurately approximate stability criteria depending on composite material properties and loadings. Two original methods to tackle with this behaviour are presented and extensively compared on the problem of Airbus in-house stability tools approximation. These methods use the idea of ensembles of approximators. Both methods provide significantly higher accuracy than conventional methods and one of the methods turned out to be significantly more accurate than other. Therefore, proposed methods can be used for construction of accurate surrogate models to replace the computationally expensive stability analyses. 1 Introduction Structural optimization for composite thin-walled structures often exhibits large computational times due to repetitive stability analysis. For composite structure made of thin plates or thin shallow shells (such as aircraft fuselage), buckling computation is of primary importance since it is one of the critical sizing constraints when minimizing the weight. This computational burden becomes even critical when we address laminated composites and when the stacking sequence is optimized [1]. The prescribed orientations and the feasibility rules for stacking sequences make this kind of problems NP-complete. To solve such problems, one often goes to population-based heuristics

We consider the problem of optimal estimation of the value of a vector parameter θ = (θ 0 ,. .. ,... more We consider the problem of optimal estimation of the value of a vector parameter θ = (θ 0 ,. .. , θn) of a drift term in a fractional Brownian motion represented by a finite sum n i=0 θ i ϕ i (t) over known functions ϕ i (t), i = 0,. .. , n. For the value of the parameter θ, we obtain a maximum likelihood estimate as well as Bayesian estimates for normal and uniform prior distrubitions.

Sobol' indices are a common metric of dependency in sensitivity analysis. It is used as a measure... more Sobol' indices are a common metric of dependency in sensitivity analysis. It is used as a measure of confidence of input variables influence on the output of the analyzed mathematical model. We consider a problem of selection of experimental design points for Sobol' indices estimation. Based on the concept of D-optimality, we propose a method for constructing an adaptive design of experiments, effective for the calculation of Sobol' indices from Polynomial Chaos Expansions. We provide a set of applications that demonstrate the efficiency of the proposed approach.

Problem of aircraft structural components (wing, fuselage, tail) optimization is considered. Solu... more Problem of aircraft structural components (wing, fuselage, tail) optimization is considered. Solution of this problem is very computationally intensive, since it requires at each iteration a two-level process: first from previous iteration, an update step at full component level must be performed in order to take into account internal loads and their sensitivities in the whole structure involved by changes in local geometry. Second, numerous local analyzes are run on isolated elements (for example, super stiffeners) of structural components in order to calculate mechanical strength criteria and their sensitivities, depending on current internal loads. An optimization step is then performed from combined global-local sensitivities. This bi-level global-local optimization process is then repeated until convergence of load distribution in the whole structure. Numerous calculations of mechanical strength criteria are necessary for local analyzes, resulting in great increase of the time between two iterations. In this work an effective method for speeding up the optimization process was proposed. The method uses surrogate models of optimization constraints (mechanical strength criteria) and provides a reduction of structure optimization computational time from several days to a few hours.

This work concerns a construction of surrogate models for a specific aerodynamic data base. This ... more This work concerns a construction of surrogate models for a specific aerodynamic data base. This data base is generally available from wind tunnel testing or from CFD aerodynamic simulations and contains aerodynamic coefficients for different flight conditions and configurations (such as Mach number, angle-of-attack, vehicle configuration angle) encountered over different space vehicles mission. The main peculiarity of aerodynamic data base is a specific design of experiment which is a union of grids of low and high fidelity data with considerably different sizes. Universal algorithms can't approximate accurately such significantly non-uniform data. In this work a fast and accurate algorithm was developed which takes into account different fidelity of the data and special design of experiments.

We consider the regression problem, i.e. prediction of a real valued function. A Gaussian process... more We consider the regression problem, i.e. prediction of a real valued function. A Gaussian process prior is imposed on the function, and is combined with the training data to obtain predictions for new points. We introduce a Bayesian regularization on parameters of a covariance function of the process, which increases quality of approximation and robustness of the estimation. Also an approach to modeling nonstationary cova􏱵 riance function of a Gaussian process on basis of linear expansion in parametric functional dictionary is pro􏱵 posed. Introducing such a covariance function allows to model functions, which have non􏱵homogeneous behaviour. Combining above features with careful optimization of covariance function parameters results in unified approach, which can be easily implemented and applied. The resulting algorithm is an out of the box solution to regression problems, with no need to tune parameters manually. The effectiveness of the method is demonstrated on various datasets.

The paper is concerned with the sequential online anomaly detection problem for a dynamical syste... more The paper is concerned with the sequential online anomaly detection problem for a dynamical system driven by a quasiperiodic stochastic process. We propose a multicomponent time series model and an effective online decomposition algorithm to approximate the components of the model. Assuming the stationarity of the obtained components , we approach the anomaly detection problem on a per-component basis and propose two online anomaly detection schemes corresponding to two real-world scenarios. Experimental results for decomposition and detection algorithms for synthesized and real-world datasets are provided to demonstrate the efficiency of our anomaly detection framework.

Предоставлены широкие возможности программного комплекса PSE/MACROS автоматиза- ции инженерных ра... more Предоставлены широкие возможности программного комплекса PSE/MACROS автоматиза- ции инженерных расчетов, содержащего современные высокоэффективные алгоритмы опти- мизации, средства интеграции с различными CAD/CAE системами, разнообразные средства многодисциплинарного анализа и визуализации данных. Представлена совокупность применяе- мых в конкретном проекте инструментов в удобном виде во встроенной графической среде. Показано, что при решении сложнейших задач внедрение комплекса существенно сокращает сроки проектирования, что позволяет улучшить технические и эксплуатационные характери- стики проектируемых изделий.

Recurrent Neural Networks (RNNs) are extensively used for time-series modeling and prediction. We... more Recurrent Neural Networks (RNNs) are extensively used for time-series modeling and prediction. We propose an approach for automatic construction of a binary classifier based on Long Short-Term Memory RNNs (LSTM-RNNs) for detection of a vehicle passage through a checkpoint. As an input to the classifier we use multidimensional signals of various sensors that are installed on the checkpoint. Obtained results demonstrate that the previous approach to handcrafting a classifier, consisting of a set of deterministic rules, can be successfully replaced by an automatic RNN training on an appropriately labelled data.

Anomalies (unusual patterns) in time-series data give essential, and often actionable information... more Anomalies (unusual patterns) in time-series data give essential, and often actionable information in critical situations. Examples can be found in such fields as healthcare, intrusion detection, finance, security and flight safety. In this paper we propose new conformalized density-and distance-based anomaly detection algorithms for a one-dimensional time-series data. The algorithms use a combination of a feature extraction method, an approach to assess a score whether a new observation differs significantly from a previously observed data, and a probabilistic interpretation of this score based on the conformal paradigm.

A number of important applied problems in engineering , finance and medicine can be formulated as... more A number of important applied problems in engineering , finance and medicine can be formulated as a problem of anomaly detection. A classical approach to the problem is to describe a normal state using a one-class support vector machine. Then to detect anomalies we quantify a distance from a new observation to the constructed description of the normal class. In this paper we present a new approach to the one-class classification. We formulate a new problem statement and a corresponding algorithm that allow taking into account a privileged information during the training phase. We evaluate performance of the proposed approach using a synthetic dataset, as well as the publicly available Microsoft Malware Classification Challenge dataset.

General predictive models do not provide a measure of confidence in predictions without Bayesian ... more General predictive models do not provide a measure of confidence in predictions without Bayesian assumptions. A way to circumvent potential restrictions is to use conformal methods for constructing non-parametric confidence regions, that offer guarantees regarding validity. In this paper we provide a detailed description of a computationally efficient conformal procedure for Kernel Ridge Regression (KRR), and conduct a comparative numerical study to see how well conformal regions perform against the Bayesian confidence sets. The results suggest that conformalized KRR can yield predictive confidence regions with specified coverage rate, which is essential in constructing anomaly detection systems based on predictive models.

ArXiv, 2021

In this teaching notes (in Russian) we consider continuous and discrete wavelet transformation an... more In this teaching notes (in Russian) we consider continuous and discrete wavelet transformation and its application to signal analysis.