Application of Machine Learning in Turbulent Combustion for Aviation Gas Turbine Combustor Design

2021 ◽  
Author(s):  
Vishwas Verma ◽  
Kiran Manoharan ◽  
Jaydeep Basani
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Flow Field ◽  
Gas Turbine ◽  
Computational Cost ◽  
Data Driven ◽  
Navier Stokes ◽  
Computational Time ◽  
Reacting Flow ◽  
Data Driven Approach

Abstract Numerical simulation of gas turbine combustors requires resolving a broad spectrum of length and time scales for accurate flow field and emission predictions. Reynold’s Averaged Navier Stokes (RANS) approach can generate solutions in few hours; however, it fails to produce accurate predictions for turbulent reacting flow field seen in general combustors. On the other hand, the Large Eddy Simulation (LES) approach can overcome this challenge, but it requires orders of magnitude higher computational cost. This limits designers to use the LES approach in combustor development cycles and prohibits them from using the same in numerical optimization. The current work tries to build an alternate approach using a data-driven method to generate fast and consistent results. In this work, deep learning (DL) dense neural network framework is used to improve the RANS solution accuracy using LES data as truth data. A supervised regression learning multilayer perceptron (MLP) neural network engine is developed. The machine learning (ML) engine developed in the present study can compute data with LES accuracy in 95% lesser computational time than performing LES simulations. The output of the ML engine shows good agreement with the trend of LES, which is entirely different from RANS, and to a reasonable extent, captures magnitudes of actual flow variables. However, it is recommended that the ML engine be trained using broad design space and physical laws along with a purely data-driven approach for better generalization.

scholarly journals Contributions towards Data driven Deep Learning methods to predict Steady State Fluid Flow in mechanical Heart Valves

2021 ◽  
Vol 7 (2) ◽  
pp. 625-628
Author(s):  
Jan Oldenburg ◽  
Julian Renkewitz ◽  
Michael Stiehm ◽  
Klaus-Peter Schmitz
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Steady State ◽  
Flow Field ◽  
Heart Valves ◽  
Data Driven ◽  
High Shear ◽  
Blood Flows ◽  
The Neural Network ◽  
Data Driven Approach

Abstract It is commonly accepted that hemodynamic situation is related with cardiovascular diseases as well as clinical post-procedural outcome. In particular, aortic valve stenosis and insufficiency are associated with high shear flow and increased pressure loss. Furthermore, regurgitation, high shear stress and regions of stagnant blood flow are presumed to have an impact on clinical result. Therefore, flow field assessment to characterize the hemodynamic situation is necessary for device evaluation and further design optimization. In-vitro as well as in-silico fluid mechanics methods can be used to investigate the flow through prostheses. In-silico solutions are based on mathematical equitation’s which need to be solved numerically (Computational Fluid Dynamics - CFD). Fundamentally, the flow is physically described by Navier-Stokes. CFD often requires high computational cost resulting in long computation time. Techniques based on deep-learning are under research to overcome this problem. In this study, we applied a deep-learning strategy to estimate fluid flows during peak systolic steady-state blood flows through mechanical aortic valves with varying opening angles in randomly generated aortic root geometries. We used a data driven approach by running 3,500 two dimensional simulations (CFD). The simulation data serves as training data in a supervised deep learning framework based on convolutional neural networks analogous to the U-net architecture. We were able to successfully train the neural network using the supervised data driven approach. The results showing that it is feasible to use a neural network to estimate physiological flow fields in the vicinity of prosthetic heart valves (Validation error below 0.06), by only giving geometry data (Image) into the Network. The neural network generates flow field prediction in real time, which is more than 2500 times faster compared to CFD simulation. Accordingly, there is tremendous potential in the use of AIbased approaches predicting blood flows through heart valves on the basis of geometry data, especially in applications where fast fluid mechanic predictions are desired.

scholarly journals Artificial Neural Network and Machine Learning Based Methods for Population Estimation of Rohingya Refugees: Comparing Data-Driven and Satellite Image-Driven Approaches

2019 ◽  
Vol 06 (04) ◽  
pp. 439-455 ◽  
Author(s):  
Nahian Ahmed ◽  
Nazmul Alam Diptu ◽  
M. Sakil Khan Shadhin ◽  
M. Abrar Fahim Jaki ◽  
M. Ferdous Hasan ◽  
...  
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Artificial Neural Network ◽  
Satellite Image ◽  
Population Estimation ◽  
Data Driven ◽  
Gradient Boosting ◽  
Ann Model ◽  
Data Driven Approach ◽  
Artificial Neural

Manual field-based population census data collection method is slow and expensive, especially for refugee management situations where more frequent censuses are necessary. This study aims to explore the approaches of population estimation of Rohingya migrants using remote sensing and machine learning. Two different approaches of population estimation viz., (i) data-driven approach and (ii) satellite image-driven approach have been explored. A total of 11 machine learning models including Artificial Neural Network (ANN) are applied for both approaches. It is found that, in situations where the surface population distribution is unknown, a smaller satellite image grid cell length is required. For data-driven approach, ANN model is placed fourth, Linear Regression model performed the worst and Gradient Boosting model performed the best. For satellite image-driven approach, ANN model performed the best while Ada Boost model has the worst performance. Gradient Boosting model can be considered as a suitable model to be applied for both the approaches.

Exploring Elastoplastic Constitutive Law of Microstructured Materials Through Artificial Neural Network—A Mechanistic-Based Data-Driven Approach

2020 ◽  
Vol 87 (9) ◽  
Author(s):  
Hang Yang ◽  
Hai Qiu ◽  
Qian Xiang ◽  
Shan Tang ◽  
Xu Guo
Keyword(s):  
Neural Network ◽  
Artificial Intelligence ◽  
Machine Learning ◽  
Complex Loading ◽  
Plasticity Theory ◽  
Constitutive Law ◽  
Data Driven ◽  
Reverse Loading ◽  
Data Driven Approach ◽  
Elastoplastic Materials

Abstract In this paper, a data-driven approach for constructing elastoplastic constitutive law of microstructured materials is proposed by combining the insights from plasticity theory and the tools of artificial intelligence (i.e., constructing yielding function through ANN) to reduce the required amount of data for machine learning. Illustrative examples show that the constitutive laws constructed by the present approach can be used to solve the boundary value problems (BVPs) involving elastoplastic materials with microstructures under complex loading paths (e.g., cyclic/reverse loading) effectively. The limitation of the proposed approach is also discussed.

scholarly journals Machine-learning computation of distance modulus for local galaxies

2020 ◽  
Vol 635 ◽  
pp. A124
Author(s):  
A. A. Elyiv ◽  
O. V. Melnyk ◽  
I. B. Vavilova ◽  
D. V. Dobrycheva ◽  
V. E. Karachentseva
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Angular Size ◽  
Data Driven ◽  
Gradient Boosting ◽  
Distance Modulus ◽  
Explanatory Variables ◽  
Photometric Redshift ◽  
The Neural Network ◽  
Data Driven Approach

Context. Quickly growing computing facilities and an increasing number of extragalactic observations encourage the application of data-driven approaches to uncover hidden relations from astronomical data. In this work we raise the problem of distance reconstruction for a large number of galaxies from available extensive observations. Aims. We propose a new data-driven approach for computing distance moduli for local galaxies based on the machine-learning regression as an alternative to physically oriented methods. We use key observable parameters for a large number of galaxies as input explanatory variables for training: magnitudes in U, B, I, and K bands, corresponding colour indices, surface brightness, angular size, radial velocity, and coordinates. Methods. We performed detailed tests of the five machine-learning regression techniques for inference of m−M: linear, polynomial, k-nearest neighbours, gradient boosting, and artificial neural network regression. As a test set we selected 91 760 galaxies at z <  0.2 from the NASA/IPAC extragalactic database with distance moduli measured by different independent redshift methods. Results. We find that the most effective and precise is the neural network regression model with two hidden layers. The obtained root–mean–square error of 0.35 mag, which corresponds to a relative error of 16%, does not depend on the distance to galaxy and is comparable with methods based on the Tully–Fisher and Fundamental Plane relations. The proposed model shows a 0.44 mag (20%) error in the case of spectroscopic redshift absence and is complementary to existing photometric redshift methodologies. Our approach has great potential for obtaining distance moduli for around 250 000 galaxies at z <  0.2 for which the above-mentioned parameters are already observed.

A comparative study of data-driven and physics-based gas turbine fault recognition approaches

2021 ◽  
pp. 1748006X2198964
Author(s):  
Juan Luis Pérez-Ruiz ◽  
Igor Loboda ◽  
Iván González-Castillo ◽  
Víctor Manuel Pineda-Molina ◽  
Karen Anaid Rendón-Cortés ◽  
...  
Keyword(s):  
Anomaly Detection ◽  
Gas Turbine ◽  
Recognition Accuracy ◽  
Data Driven ◽  
Fault Identification ◽  
Fault Recognition ◽  
Data Driven Approach ◽  
Diagnostic Approaches ◽  
Artificial Neural Network Ann ◽  
Two Stages

The present paper compares the fault recognition capabilities of two gas turbine diagnostic approaches: data-driven and physics-based (a.k.a. gas path analysis, GPA). The comparison takes into consideration two differences between the approaches, the type of diagnostic space and diagnostic decision rule. To that end, two stages are proposed. In the first one, a data-driven approach with an artificial neural network (ANN) that recognizes faults in the space of measurement deviations is compared with a hybrid GPA approach that employs the same type of ANN to recognize faults in the space of estimated fault parameter. Different case studies for both anomaly detection and fault identification are proposed to evaluate the diagnostic spaces. They are formed by varying the classification, type of diagnostic analysis, and deviation noise scheme. In the second stage, the original GPA is reconstructed replacing the ANN with a tolerance-based rule to make diagnostic decisions. Here, two aspects are under analysis: the comparison of GPA classification rules and whole approaches. The results reveal that for simple classifications both spaces are equally accurate for anomaly detection and fault identification. However, for complex scenarios, the data-driven approach provides on average slightly better results for fault identification. The use of a hybrid GPA with ANN for a full classification instead of an original GPA with tolerance-based rule causes an increase of 12.49% in recognition accuracy for fault identification and up to 54.39% for anomaly detection. As for the whole approach comparison, the application of a data-driven approach instead of the original GPA can lead to an improvement of 12.14% and 53.26% in recognition accuracy for fault identification and anomaly detection, respectively.

Combustion Tuning for a Gas Turbine Power Plant Using Data-Driven and Machine Learning Approach

2021 ◽  
Vol 143 (3) ◽  
Author(s):  
Suhui Li ◽  
Huaxin Zhu ◽  
Min Zhu ◽  
Gang Zhao ◽  
Xiaofeng Wei
Keyword(s):  
Machine Learning ◽  
Power Plant ◽  
Gas Turbine ◽  
Operating Conditions ◽  
Data Driven ◽  
Ann Model ◽  
Promising Alternative ◽  
Combustion Performance ◽  
Wide Range ◽  
Gas Turbine Power Plant

Abstract Conventional physics-based or experimental-based approaches for gas turbine combustion tuning are time consuming and cost intensive. Recent advances in data analytics provide an alternative method. In this paper, we present a cross-disciplinary study on the combustion tuning of an F-class gas turbine that combines machine learning with physics understanding. An artificial-neural-network-based (ANN) model is developed to predict the combustion performance (outputs), including NOx emissions, combustion dynamics, combustor vibrational acceleration, and turbine exhaust temperature. The inputs of the ANN model are identified by analyzing the key operating variables that impact the combustion performance, such as the pilot and the premixed fuel flow, and the inlet guide vane angle. The ANN model is trained by field data from an F-class gas turbine power plant. The trained model is able to describe the combustion performance at an acceptable accuracy in a wide range of operating conditions. In combination with the genetic algorithm, the model is applied to optimize the combustion performance of the gas turbine. Results demonstrate that the data-driven method offers a promising alternative for combustion tuning at a low cost and fast turn-around.

scholarly journals Data-driven predictions of a multiscale Lorenz 96 chaotic system using machine-learning methods: reservoir computing, artificial neural network, and long short-term memory network

2020 ◽  
Vol 27 (3) ◽  
pp. 373-389 ◽  
Author(s):  
Ashesh Chattopadhyay ◽  
Pedram Hassanzadeh ◽  
Devika Subramanian
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Short Term Memory ◽  
Data Driven ◽  
Reservoir Computing ◽  
Short Term ◽  
Term Memory ◽  
Machine Learning Methods ◽  
Long Short Term Memory ◽  
Lorenz 96

Abstract. In this paper, the performance of three machine-learning methods for predicting short-term evolution and for reproducing the long-term statistics of a multiscale spatiotemporal Lorenz 96 system is examined. The methods are an echo state network (ESN, which is a type of reservoir computing; hereafter RC–ESN), a deep feed-forward artificial neural network (ANN), and a recurrent neural network (RNN) with long short-term memory (LSTM; hereafter RNN–LSTM). This Lorenz 96 system has three tiers of nonlinearly interacting variables representing slow/large-scale (X), intermediate (Y), and fast/small-scale (Z) processes. For training or testing, only X is available; Y and Z are never known or used. We show that RC–ESN substantially outperforms ANN and RNN–LSTM for short-term predictions, e.g., accurately forecasting the chaotic trajectories for hundreds of numerical solver's time steps equivalent to several Lyapunov timescales. The RNN–LSTM outperforms ANN, and both methods show some prediction skills too. Furthermore, even after losing the trajectory, data predicted by RC–ESN and RNN–LSTM have probability density functions (pdf's) that closely match the true pdf – even at the tails. The pdf of the data predicted using ANN, however, deviates from the true pdf. Implications, caveats, and applications to data-driven and data-assisted surrogate modeling of complex nonlinear dynamical systems, such as weather and climate, are discussed.

A note on no-free-lunch theorem

2021 ◽  
Author(s):  
Lidong Wu
Keyword(s):  
Machine Learning ◽  
Theoretical Result ◽  
Data Driven ◽  
Np Hard ◽  
Free Lunch ◽  
No Free Lunch Theorem ◽  
Data Driven Approach ◽  
Hard Problems ◽  
No Free Lunch ◽  
Np Hard Problems

The No-Free-Lunch theorem is an interesting and important theoretical result in machine learning. Based on philosophy of No-Free-Lunch theorem, we discuss extensively on the limitation of a data-driven approach in solving NP-hard problems.

Data-driven seismic response prediction of structural components

2021 ◽  
pp. 875529302110533
Author(s):  
Huan Luo ◽  
Stephanie German Paal
Keyword(s):  
Experimental Data ◽  
Seismic Response ◽  
Computational Cost ◽  
Data Driven ◽  
Computational Time ◽  
Lateral Stiffness ◽  
Hysteretic Model ◽  
Structural Components ◽  
Rc Columns ◽  
Detailed Model

Lateral stiffness of structural components, such as reinforced concrete (RC) columns, plays an important role in resisting the lateral earthquake loads. The lateral stiffness relates the lateral force to the lateral deformation, having a critical effect on the accuracy of the lateral seismic response predictions. The classical methods (e.g. fiber beam–column model) to estimate the lateral stiffness require calculations from section, element, and structural levels, which is time-consuming. Moreover, the shear deformation and bond-slip effect may also need to be included to more accurately calculate the lateral stiffness, which further increases the modeling difficulties and the computational cost. To reduce the computational time and enhance the accuracy of the predictions, this article proposes a novel data-driven method to predict the laterally seismic response based on the estimated lateral stiffness. The proposed method integrates the machine learning (ML) approach with the hysteretic model, where ML is used to compute the parameters that govern the nonlinear properties of the lateral response of target structural components directly from a training set composed of experimental data (i.e. data-driven procedure) and the hysteretic model is used to directly output the lateral stiffness based on the computed parameters and then to perform the seismic analysis. We apply the proposed method to predict the lateral seismic response of various types of RC columns subjected to cyclic loading and ground motions. We present the detailed model formulation for the application, including the developments of a modified hysteretic model, a hybrid optimization algorithm, and two data-driven seismic response solvers. The results predicted by the proposed method are compared with those obtained by classical methods with the experimental data serving as the ground truth, showing that the proposed method significantly outperforms the classical methods in both generalized prediction capabilities and computational efficiency.

Sign in / Sign up

Export Citation Format

Share Document