Manifold Model Research Articles

Implementation of a high-accuracy manifold element modelling scheme for dynamic fracture analysis under thermal-mechanical shock

In this study, a high-accuracy numerical manifold modelling scheme is proposed for simulating dynamic thermoelastic fracture problems of two-dimensional stationary cracks in linear elastic solids. The advantages of the numerical manifold method in discretizing the physical model with multiple discontinuities, and those of the precise time step integration method in time integral are integrated together in the proposed numerical scheme. The implementation details of the proposed modelling scheme are well presented. Dynamic stress intensity factor considering the inertial effect is computed. Through three dynamic fracture examples with pure thermal shock, heat flux shock and mechanical shock, respectively, the validity of the present scheme is first verified. Then the proposed scheme is applied for dynamic fracture analysis under mixed thermal-mechanical shock, in which different loading forms, such as step loading, linear loading and harmonic loading, are considered. The results show that the present modelling scheme is extremely stable and convergence. The advantages of the present scheme on accuracy, efficiency, time-step-independence over commonly used Newmark method are also fully demonstrated.

Volumetric adaptive slicing of manifold mesh for rapid prototyping based on relative volume error

PurposeCompared with cusp height and area deviation ratio, volume error (VE) caused by the layer height could represent the stair-case effect more comprehensively. The proposed relative volume error (RVE)-based adaptive slicing method takes VE rather than cusp height as slicing criteria, which can improve part surface quality for functionalized additive manufacturing.Design/methodology/approachThis paper proposes a volumetric adaptive slicing method of manifold mesh for rapid prototyping based on RVE. The pre-height sequences of manifold mesh are first preset to reduce the SE by dividing the whole layer sequence into several parts. A breadth-first search-based algorithm has been developed to generate a solid voxelization to get VE. A new parameter RVE is proposed to evaluate the VE caused by the sequence of the layer positions. The RVE slicing is conducted by iteratively adjusting the layer height sequences under different constraint conditions.FindingsThree manifold models are used to verify the proposed method. Compared with uniform slicing with 0.2 mm layer height, cusp height-based method and area deviation-based method, the standard deviations of RVE of all three models are improved under the proposed method. The surface roughness measured by the confocal laser scanning microscope proves that the proposed RVE method can greatly improve part surface quality by minimizing RVE.Originality/valueThis paper proposes an RVE-based method to balance the surface quality and print time. RVE could be calculated by voxelized parts with required accuracy at a very fast speed by parallel.

CO Emission Modeling in a Heavy Duty Annular Combustor Operating With Natural Gas

Abstract This paper summarizes the development of a large-eddy simulation (LES)-based approach for the prediction of CO emission in an industrial gas turbine combustor. Since the operating point of the modern combustors is really close to the extinction limit, the availability of a tool able to detect the onset of high-CO production can be useful for the proper definition of the combustion chamber air split or to introduce design improvements for the premixer itself. The accurate prediction of CO cannot rely on the flamelet assumption, representing the fundament of the modern combustion models. Consequently, in this work, the extended turbulent flame speed closure (ETFSC) of the standard flamelet generated manifold (FGM) model is employed to consider the effect of the heat loss and the strain rate on the flame brush. Moreover, a customized CO-Damköhler number is introduced to decouple the in-flame CO production region from the postflame contribution where the oxidation takes place. A fully premixed burner working at representative values of pressure and flame temperature of an annular combustor is selected for the validation phase of the process. The comparison against the experimental data shows that the process is not only able to capture the trend but also to predict CO in a quantitative manner. In particular, the interaction between the flame and the air fluxes at some critical sections of the combustor, leading the CO emission from the equilibrium value to the super-equilibrium, has been correctly reproduced.

Flamelet Versus Detailed Chemistry Large Eddy Simulation for a Liquid-Fueled Gas Turbine Combustor: A Comparison of Accuracy and Computational Cost

Abstract A Honeywell liquid-fueled gas turbine test combustor at idle conditions is numerically investigated in simcenterstar-ccm+ version 2020.3. This work presents large eddy simulation (LES) results using both the flamelet generated manifold (FGM) and detailed chemistry combustion models. Both take advantage of a hybrid chemical (HyChem) mechanism which has previously demonstrated very good accuracy for real fuels such as Jet-A with only 47 species. The objective of this work is to investigate the ability of FGM and detailed chemistry modeling to capture pollutant formation in an aero-engine combustor. Comparisons for NOx, CO, unburned hydrocarbons (UHC), and soot are made, along with the radial temperature profile. To fully capture potential emissions, a soot moment model and Zeldovich NOx model are employed along with radiation. A comparison of results with and without chemistry acceleration techniques for detailed chemistry is included. Then, computational costs are assessed by comparing the performance and scalability of the simulations with each of the combustion models. It is found that the detailed chemistry case with clustering can reproduce nearly identical results to detailed chemistry without any acceleration if CO is added as a clustering variable. With the Lagrangian model settings chosen for this study, the detailed chemistry results compared more favorably with the experimental data than FGM; however, there is uncertainty in the secondary breakup parameters. Sensitivity of the results to a key parameter in the spray breakup model is provided for both FGM and complex chemistry (CC). By varying this breakup rate, the FGM case can predict CO, NOx, and UHC equally well. The smoke number, however, is predicted most accurately by CC. The cost for running detailed chemistry with clustering is found to be about four times that of FGM for this combustor and chemical mechanism.

The manifold embedded selective pseudo-labeling algorithm and transfer learning of small sample dataset

Special scene classification and identification tasks are not easily fulfilled to obtain samples, which results in a shortage of samples. The focus of current researches lies in how to use source domain data (or auxiliary domain data) to build domain adaption transfer learning models and to improve the classification accuracy and performance of small sample machine learning in these special and difficult scenes. In this paper, a model of deep convolution and Grassmann manifold embedded selective pseudo-labeling algorithm (DC-GMESPL) is proposed to enable transfer learning classifications among multiple small sample datasets. Firstly, DC-GMESPL algorithm uses satellite remote sensing image sample data as the source domain to extract the smoke features simultaneously from both the source domain and the target domain based on the Resnet50 deep transfer network. This is done for such special scene of the target domain as the lack of local sample data for forest fire smoke video images. Secondly, DC-GMESPL algorithm makes the source domain feature distribution aligned with the target domain feature distribution. The distance between the source domain and the target domain feature distribution is minimized by removing the correlation between the source domain features and re-correlation with the target domain. And then the target domain data is pseudo-labeled by selective pseudo-labeling algorithm in Grassmann manifold space. Finally, a trainable model is constructed to complete the transfer classification between small sample datasets. The model of this paper is evaluated by transfer learning between satellite remote sensing image and video image datasets. Experiments show that DC-GMESPL transfer accuracy is higher than DC-CMEDA, Easy TL, CMMS and SPL respectively. Compared with our former DC-CMEDA, the transfer accuracy of our new DC-GMESPL algorithm has been further improved. The transfer accuracy of DC-GMESPL from satellite remote sensing image to video image has been improved by 0.50%, the transfer accuracy from video image to satellite remote sensing image has been improved by 8.50% and then, the performance has been greatly improved.

Influences of the dead end on the flow characteristics at the exhaust manifold of a marine diesel engine

A 3D Computational Fluid Dynamics (CFD) model of an exhaust manifold with and without a dead end has been developed to investigate the impacts of its geometry on the flow structure and the pressure distribution within the manifold. The model differs from previously studied models principally for its ability to approach the realistic operating principle of an engine as the modelled exhaust valves of the investigated engine open and close according to the firing order. The experimental results of an exhaust manifold without a dead end has been used to validate the CFD model through the pressure distribution and the flow structure. The outcomes demonstrated that the developed CFD model concurred well with the experimental data. The effects of the dead end on the exhaust manifold were then investigated using the validated CFD model. The study has revealed that the addition of a dead end (i) provides a smoother pressure distribution inside the manifold and increase in the efficiency of the turbocharger and (ii) decreases the pressure inside of the interconnection pipes of cylinders while the exhaust gas discharges. Moreover, the results disclose a smoother discharge of exhaust gases leading to a more effective sweeping of the exhaust gas thorough the cylinder without causing any exhaust backpressure. Furthermore, the dead end reduces the turbulence kinetic energy at the blind end of the exhaust manifold resulting in a decrease of pressure loss within. The abovementioned findings regarding to the flow structure and the pressure distribution within the exhaust manifold improve the efficiency of the engine.

The effects of swirl number on the combustion characteristics in an internally-staged combustor

The three-dimensional turbulent swirling flame in an internally-staged combustor is numerically investigated. Four cases over a range of swirl intensities are explored by the Flamelet Generated Manifold model in this paper. Special attention is paid to analyzing the variation of the flow field, temperature, major species concentrations and emissions. The results clearly show the effects of swirl number on the size of the center recirculation zone, fuel distribution and combustion characteristics. When the third premixed stage swirl number increases from 0.6 to 1.2, the axial length of the center recirculation zone decreases by 3.7%, while the radial length increases by 6.9%. The characteristics of the flow field play an important role in the spatial distribution of the fuel, which further affects the temperature distribution in the combustor. The backflow effect is enhanced, resulting in a greater concentration of fuel at the outlet of the swirler. After the maximum temperature is reached at the exit position of the pilot stage, the temperature decreases compared to the peak temperature downstream as the proportion of premixed combustion mode increases. This creates a high concentration region of OH at the outlet of the pilot stage injector and the heat release region is squeezed upstream. At the same time, the volume of the high-temperature region downstream of the pilot stage is reduced. In addition, as the swirl number of the third premixed stage increases from 0.6 to 1.2, the emissions of NO and CO decrease by 28.7% and 75%, respectively.

Large eddy simulation of premixed turbulent combustion using a non-adiabatic, strain-sensitive flamelet approach

In this study, a novel strained, non-adiabatic flamelet generated manifold (FGM) model is developed based on a counterflow premixed flame. The different levels of strain and heat loss are introduced into the flamelets by varying the stagnation point strain rate and burnt temperature of the counterflow configuration. The strain rate response of these flamelets is found to exhibit different extinction patterns for different ranges of burnt temperature. A priori studies are performed to compare the phase-space flame structures of the strained and unstrained flamelets. It is found that the strained flamelet degenerates towards the unstrained flamelet with burnt temperature decreasing. A bifurcation burnt temperature is further defined that separates the two types of flamelet quenching that are driven by strain and heat loss. An appropriate set of tabulation parameters is chosen that preserves the uniqueness of mapping between the manifold and the populated flamelets. The external boundaries of the tabulations are defined, where the proposed model degenerates into the conventional unstrained/non-adiabatic FGM manifold. Internal tabulation boundaries are defined to distinguish different combustion modes among stable combustion, strain-induced quenching, and heat-loss-induced quenching. The model is validated using large eddy simulation (LES) of a confined turbulent premixed jet flame. The proposed FGM model vastly increases accuracy of predictions when both strain and heat loss effects are included. Without accounting for these effects, there are discrepancies in the prediction of flame height and temperature as compared with experimental data. An analysis of the dominant physical process (strain, heat loss) that contributes to the evolution of the turbulent flame is performed.

Experimental and numerical study on the characteristics of a penetrating exhaust manifold for low-speed two-stroke marine diesel engine

Penetrating exhaust manifold systems are commonly used in low-speed two-stroke marine diesel engines owing to their potential for energy conservation. However, few studies in the literature have reported on this type of manifold system, and there are few guidelines for optimizing the configuration or predicting the performance via a mean-line model. This study investigates the loss characteristics and flow mechanism of the penetrating exhaust manifold of a low-speed two-stroke marine diesel engine using experimental and numerical methods. The loss is characterized by the entropy production, which indicates a decrease in the energy grade. The results show that the minimum loss can be achieved when the penetration depth and diffusion angle of the manifold are approximately 0.5 and 6°, respectively. Moreover, the loss coefficient is maximally reduced by 14.0 % with penetrating exhaust manifold. The flow field analysis shows that the trend of the flow loss is ‘decrease-increase’ owing to the interaction among three factors: the mixing loss due to the shearing of the injection flow with the main flow, the momentum loss due to the impingement on the bottom of the main pipe, and the separation loss in the branch pipe owing to the flow divergence. This investigation can provide guidance for the configuration optimization and loss modeling of the exhaust manifold for low-speed two-stroke marine engines.

CONCEPT FOR A PERSONA DRIVEN RECOMMENDATION TOOL FOR PROCESS MODELLING APPROACHES

AbstractIn order to ensure successful product development processes, manifold modelling approaches have been developed, which cover a wide range of aspects such as responsibilities, duration of activities and dependencies. Still, an industry standard does not exist. Users of process modelling approaches are driven by different targets depending on the respective role. Currently, practitioners need to evaluate strengths and weaknesses of each approach by themselves and find little guidance for the selection.As a consequence, users might select unsuitable approaches and do not get the expected result. Thus, the intended applications of the model such as analyses or an optimization of the process are hampered. This could heavily affect companies´ success by product or project failures.The paper shows the concept of a recommendation tool that enables a suitable and effective selection of process modelling approaches. Key element is the description of relevant use cases and personas that represent the various needs of both different company types as well as different roles within such as process modellers and users. By identifying the most relevant case, each practitioner will be successfully guided to the most suitable modelling approach.

Emission prediction and analysis on CH4/NH3/air swirl flames with LES-FGM method

Ammonia, made up of 17.8% hydrogen, is one of carbon free fuels which can be used in large-scale power plants. However, ammonia combustion faces some challenges due to the high NOx emission. This study investigates the emission characteristics of CH4/NH3/air co-firing premixed swirling flames with NH3 mole fraction up to 60% in a model combustor. Large eddy simulation (LES) coupled with flamelet generated manifold (FGM) model which considers the transport equations of slow chemical emission component was performed, with experimental validation in both flame structure and global emission. The simulation shows an overall good agreement with experimental results. It is found that the NO emission is improved with the NO transport equation calculated during the simulation. The results imply that the nitrogenous components, residence time and temperature are important factors that influence NO concentration profile. The residence time of inner recirculation indicates the NO level for CH4/NH3/air flames, and local high NO concentration is mainly caused by local long residence time. The OH and NO are correlative in premixed CH4/NH3/air flames because these components are both strongly related to the temperature. N2O has a correlation with NH and HNO components. The radical components such as HNO, O, OH etc. also influence the emission formation process. When the NH3 ratio is closed to 40%, the concentration of H, OH, O2 and HNO are all adequate to produce maximum NO.

Manifold Modeling in Machine Learning

Manifold Modeling in Machine Learning

A new beginning of intelligence research. Designing the playground

Here we present several points for designing a probable playground concerning a new beginning of intelligence research within the XXI Century: the nature, definition, and measurement of the construct of interest, its development across the lifespan, its enhancement by varied means, and its place within the already identified human psychological traits. Predictions can go wrong when those who make them 1) assume that trends will be linear, 2) use script-writing assuming that they know what the responses to any trend will be, and 3) conflate primary facts with their interpretation. With these pitfalls in mind, we predict: 1) a proliferation of alternate models of the positive manifold; 2) The derailment of the field in the next decade or two with a new trendy research angle; 3) The gradual abandonment of classic IQ tests for intelligence research in favor of alternative measurements. We see a bright future for intelligence research, but dark spots cannot be discarded.

Cartesian Mesh Generation with Local Refinement for Immersed Boundary Approaches

In this paper, an efficient and robust Cartesian Mesh Generation with Local Refinement for an Immersed Boundary Approach is proposed, whose key feature is the capability of high Reynolds number simulations by the use of wall function models, bypassing the need for accurate boundary layer discretization. Starting from the discrete manifold model of the object to be analyzed, the proposed model generates Cartesian adaptive grids for a CFD simulation, with minimal user interactions; the most innovative aspect of this approach is that the automatic generation is based on the segmentation of the surfaces enveloping the object to be analyzed. The aim of this paper is to show that this automatic workflow is robust and enables to get quantitative results on geometrically complex configurations such as marine vehicles. To this purpose, the proposed methodology has been applied to the simulation of the flow past a BB2 submarine, discretized by non-uniform grid density. The obtained results are comparable with those obtained by classical body-fitted approaches but with a significant reduction of the time required for the mesh generation.

Driver distraction analysis using face pose cues

Vehicle driver distraction is one of the major reasons for road accidents. Involvement with a co-passenger, use of in-vehicle devices or phone leads to a situation where the driver head pose varies and the eye is off the road. A low cost early warning system should reduce the distracted driving instances, thus making our roads safer. Face pose information forms an important cue to determine driver distraction. The main objective of this work is to analyse the distractions of the driver based on his/her face pose cues. A straight pose or slight variation would indicate a non-distracted driver, while a large pose variation from the center would indicate a high probability for a distracted driver. Face pose database of vehicle drivers is developed and is bench-marked. A clustered two layer approach on Gabor features is proposed. A five layer convolutional network with three fully connected layers is also used to bench-mark the data. The proposed clustered two-layer approach with Gabor features and SVM classifier provides better results in driver distraction analysis when compared to the deep learning approach and other manifold approaches. The improved accuracy could be attributed to the improved modeling of manifold in our approach, better class discrimination of the Gabor features together with better classification provided by the SVM classifier.

On Recovery Guarantees for One-Bit Compressed Sensing on Manifolds

This paper studies the problem of recovering a signal from one-bit compressed sensing measurements under a manifold model; that is, assuming that the signal lies on or near a manifold of low intrinsic dimension. We provide a convex recovery method based on the Geometric Multi-Resolution Analysis and prove recovery guarantees with a near-optimal scaling in the intrinsic manifold dimension. Our method is the first tractable algorithm with such guarantees for this setting. The results are complemented by numerical experiments confirming the validity of our approach.

DEEP GENERATIVE STORM MODEL FOR DYNAMIC IMAGING.

We introduce a novel generative smoothness regularization on manifolds (SToRM) model for the recovery of dynamic image data from highly undersampled measurements. The proposed generative framework represents the image time series as a smooth non-linear function of low-dimensional latent vectors that capture the cardiac and respiratory phases. The non-linear function is represented using a deep convolutional neural network (CNN). Unlike the popular CNN approaches that require extensive fully-sampled training data that is not available in this setting, the parameters of the CNN generator as well as the latent vectors are jointly estimated from the undersampled measurements using stochastic gradient descent. We penalize the norm of the gradient of the generator to encourage the learning of a smooth surface/manifold, while temporal gradients of the latent vectors are penalized to encourage the time series to be smooth. The main benefits of the proposed scheme are (a) the quite significant reduction in memory demand compared to the analysis based SToRM model, and (b) the spatial regularization brought in by the CNN model. We also introduce efficient progressive approaches to minimize the computational complexity of the algorithm.

Dynamic Imaging Using a Deep Generative SToRM (Gen-SToRM) Model.

We introduce a generative smoothness regularization on manifolds (SToRM) model for the recovery of dynamic image data from highly undersampled measurements. The model assumes that the images in the dataset are non-linear mappings of low-dimensional latent vectors. We use the deep convolutional neural network (CNN) to represent the non-linear transformation. The parameters of the generator as well as the low-dimensional latent vectors are jointly estimated only from the undersampled measurements. This approach is different from traditional CNN approaches that require extensive fully sampled training data. We penalize the norm of the gradients of the non-linear mapping to constrain the manifold to be smooth, while temporal gradients of the latent vectors are penalized to obtain a smoothly varying time-series. The proposed scheme brings in the spatial regularization provided by the convolutional network. The main benefit of the proposed scheme is the improvement in image quality and the orders-of-magnitude reduction in memory demand compared to traditional manifold models. To minimize the computational complexity of the algorithm, we introduce an efficient progressive training-in-time approach and an approximate cost function. These approaches speed up the image reconstructions and offers better reconstruction performance.

Noncircularity-based generalized shift invariance for estimation of angular parameters of incoherently distributed sources

In this paper, a reduced-rank angular parameters estimation algorithm is proposed for incoherently distributed (ID) noncircular sources based on a uniform linear array (ULA), which addresses the problems of central direction-of-arrival (DOA) estimation and angular spread estimation. Firstly, the noncircularity property of the signals is utilized to establish an extended generalized array manifold (GAM) model based on the first-order Taylor series approximation. Then, the central DOAs of source signals are obtained based on the generalized shift invariance property of the array manifold and the reduced-rank principle. Next, the angular spreads are estimated from the central moments of the angular distribution. Compared with the existing algorithm without exploiting the noncircularity information, the proposed one can achieve a higher accuracy and handle more sources. In addition, it can deal with a general scenario where different sources have different angular distribution shapes. Furthermore, the approximate stochastic Cramer-Rao bound (CRB) of the concerned problem is derived. Simulation results are provided to demonstrate the performance of the proposed algorithm.

Depth Map Reconstruction and Enhancement With Local and Patch Manifold Regularized Deep Depth Priors

The depth map captured by depth sensors (e.g., the time of flight (ToF) and Kinect) is often prone to low resolution, degradation, noise, and poor quality. This paper proposes a novel model for the robust depth estimation of RGB-D images through local and nonlocal manifold regularizations. The first stage called deep depth prior manifold (DDPM), is inspired partly by the deep depth prior (DDP) model, that is a deep convolutional neural network (CNN) integrated with a local manifold regularization term. The local neighboring relationships between depth pixels and color images are employed to promote smoothing in the results. The Laplacian Eigenmap technique used for local manifold modeling produces over-smooth depth map. To improve the quality of the reconstructed image, a nonlocal manifold modeling stage was suggested, where the similarity between the depth and the corresponding color image is determined by characterizing their matching aspects. These objectives are aggregated within an optimization problem. Moreover, to extract edges better considering visual nonlocal characteristics, the structured low-rank Hankel approximation was adopted to better eliminate depth degradations, and to extract highly promoted edges and sharp points. Three types of the degradations were handling in this work, containing undersampling, ToF-like, and Kinect-like degradations. Experimental results indicate that the proposed method outperformed the state-of-the-art restoration techniques on standard benchmark images, in terms of well-known criteria like PSNR.