Residual Modes Research Articles

MSRA-G: Combination of multi-scale residual attention network and generative adversarial networks for hyperspectral image classification

Deep learning-based technology has been introduced to increase the classification accuracy of hyperspectral imagery (HSI). Nevertheless, it is still a challenging issue to derive a satisfying classification accuracy from limited training samples. A novel method (MSRA-G) that combines multi-scale residual attention (MSRA) with Generative Adversarial Networks (GANs) was proposed. In view of the low classification accuracy with limited training samples, the is first used to generate more separable synthetic samples. A network is then proposed to extract multi-scale context information for improving HSI classification. The proposed method constructs two multi-scale feature extraction modules to identify high-level spatial–spectral features based on the 3D–2D hybrid network. In addition, the residual connection mode and the attention mechanism are combined to establish the channel and spatial residual attention modules. Different weights are assigned to different features in the channel dimension and spatial dimension, and the features are selectively learned. Furthermore, to verify the performance of MSRA-G, experiments were carried out on three publicly available HSI datasets of Indian Pines, University of Pavia and Salinas Valley. The experimental results show that our proposed MSRA-G is superior to several popular classification models. It can still achieve satisfactory classification accuracies, even in the case of insufficient training samples.

A non-intrusive dynamic sensitivity-based substructure model updating method for nonlinear systems

In nonlinear finite element model updating (FEMU), one of the main challenges is how to improve the updating efficiency for large-scale structures. The computational time of updating depends on the degree-of-freedom of the finite element model, and the optimization algorithm that the user chooses. In this work, a novel dynamic response sensitivity-based nonlinear FEMU method using the substructures is presented to reduce the time consuming of computation and updating. Considering the residual structural modes, the enhanced Hurty/Craig-Bampton (EHCB) substructure method is developed to establish a more precise initial numerical model, which also has high computational efficiency. A non-intrusive algorithm based on the real and imaginary perturbation of updating parameters is implemented to compute the nonlinear substructural response sensitivities, and further reduce the iteration loops in the nonlinear FEMU process. Simulation examples with smooth and non-smooth localized nonlinearities are conducted to verify the accuracy, robustness, and efficiency of the proposed method. Results indicated that the proposed method can effectively and efficiently update the nonlinear finite element model with high accuracy.

Residual dynamic mode decomposition: robust and verified Koopmanism

Dynamic mode decomposition (DMD) describes complex dynamic processes through a hierarchy of simpler coherent features. DMD is regularly used to understand the fundamental characteristics of turbulence and is closely related to Koopman operators. However, verifying the decomposition, equivalently the computed spectral features of Koopman operators, remains a significant challenge due to the infinite-dimensional nature of Koopman operators. Challenges include spurious (unphysical) modes and dealing with continuous spectra, which both occur regularly in turbulent flows. Residual dynamic mode decomposition (ResDMD), introduced by Colbrook & Townsend (Rigorous data-driven computation of spectral properties of Koopman operators for dynamical systems. 2021. arXiv:2111.14889), overcomes such challenges through the data-driven computation of residuals associated with the full infinite-dimensional Koopman operator. ResDMD computes spectra and pseudospectra of general Koopman operators with error control and computes smoothed approximations of spectral measures (including continuous spectra) with explicit high-order convergence theorems. ResDMD thus provides robust and verified Koopmanism. We implement ResDMD and demonstrate its application in various fluid dynamic situations at varying Reynolds numbers from both numerical and experimental data. Examples include vortex shedding behind a cylinder, hot-wire data acquired in a turbulent boundary layer, particle image velocimetry data focusing on a wall-jet flow and laser-induced plasma acoustic pressure signals. We present some advantages of ResDMD: the ability to resolve nonlinear and transient modes verifiably; the verification of learnt dictionaries; the verification of Koopman mode decompositions; and spectral calculations with reduced broadening effects. We also discuss how a new ordering of modes via residuals enables greater accuracy than the traditional modulus ordering (e.g. when forecasting) with a smaller dictionary. This result paves the way for more significant dynamic compression of large datasets without sacrificing accuracy.

A new residual distribution hydrodynamics solver for astrophysical simulations

ABSTRACT Many astrophysical systems can only be accurately modelled when the behaviour of their baryonic gas components is well understood. The residual distribution (RD) family of partial differential equation (PDE) solvers produce approximate solutions to the corresponding fluid equations. We present a new implementation of the RD method. The solver efficiently calculates the evolution of the fluid, with up to second order accuracy in both time and space, across an unstructured triangulation, in both 2D and 3D. We implement a novel variable time stepping routine, which applies a drifting mechanism to greatly improve the computational efficiency of the method. We conduct extensive testing of the new implementation, demonstrating its innate ability to resolve complex fluid structures, even at very low resolution. We can resolve complex structures with as few as 3–5 resolution elements, demonstrated by Kelvin–Helmholtz and Sedov blast tests. We also note that we find cold cloud destruction time scales consistent with those predicted by a typical PPE solver, albeit the exact evolution shows small differences. The code includes three residual calculation modes, the LDA, N, and blended schemes, tailored for scenarios from smooth flows (LDA), to extreme shocks (N), and both (blended). We compare our RD solver results to state-of-the-art solvers used in other astrophysical codes, demonstrating the competitiveness of the new approach, particularly at low resolution. This is of particular interest in large scale astrophysical simulations, where important structures, such as star forming gas clouds, are often resolved by small numbers of fluid elements.

Narrow Linewidth Half-Open-Cavity Random Laser Assisted by a Three-Grating Ring Resonator for Strain Detection.

A stabilized narrow-linewidth random fiber laser for strain detection, based on a three-grating ring (TGR) resonator and half-open-cavity structure, is proposed and investigated experimentally. The half-open-cavity structure proved to provide double optical gain of erbium-doped fiber, which was beneficial to increase the photon lifetime as well as further narrow the linewidth. Meanwhile, the stability and frequency noise of narrow lasing output was improved by suppressing the competition-induced undesired residual random lasing modes with the TGR resonator. The TGR resonator is composed of a double-cavity fiber Bragg grating Fabry–Perot (FBG-FP) interferometer, a section of single-mode fiber, and a circulator. The specially designed double-cavity FBG-FP interferometer embedded in the TGR resonator acted as the strain-sensing element and improved the resolution of the dynamic strain. A stable ultra-narrow linewidth of about 205 Hz was obtained. The frequency noise was reduced to about 2 Hz/√Hz. A high dynamic strain measuring resolution of 35 femto-strain (fε)/√Hz was achieved.

Fingerprint and Universal Markovian Closure of Structured Bosonic Environments.

We exploit the properties of chain mapping transformations of bosonic environments to identify a finite collection of modes able to capture the characteristic features, or fingerprint, of the environment. Moreover we show that the countable infinity of residual bath modes can be replaced by a universal Markovian closure, namely, a small collection of damped modes undergoing a Lindblad-type dynamics whose parametrization is independent of the spectral density under consideration. We show that the Markovian closure provides a quadratic speedup with respect to standard chain mapping techniques and makes the memory requirement independent of the simulation time, while preserving all the information on the fingerprint modes. We illustrate the application of the Markovian closure to the computation of linear spectra but also to nonlinear spectral response, a relevant experimentally accessible many body coherence witness for which efficient numerically exact calculations in realistic environments are currently lacking.

Vibration-modelling-based fault feature analysis for incipient damage identification of sun gear

Different types of incipient damage to gears need to be distinguished for the reliable and safe operation of mechanical equipment. However, the identification of various types of incipient gear damage is still a challenge for the fault diagnosis of planetary gearboxes. To identify the various types of incipient damage in a sun gear, fault feature analysis is conducted for various damage conditions in a planetary gearbox. First, the internal excitation aroused by gear damage is computed and the effects of the typical damage modes of a sun gear on the internal excitation are demonstrated. Then, the internal excitation of the damaged sun gear is applied into the dynamic model of the planetary gear train to simulate distributed damage or local damage. After that, with the vibration responses obtained from the dynamic model, a vibration signal model is proposed to simulate the vibration signals of the planetary gearbox. Furthermore, the vibration signals are analysed over the time domain and frequency domain. The fault features of the planetary gearbox are characterized by using residual signals and variational mode decomposition signals, which are useful for identifying the various incipient damage types of a sun gear. Finally, the simulated results are validated by experimental tests.

Low Frequency-Noise Ring Random Fiber Laser With a Dual-Cavity FBG Fabry-Perot Filter

A low frequency-noise ring random fiber laser (RFL) based on random Bragg grating array (RBGA) feedback is proposed. A dual-cavity fiber Bragg grating Fabry-Perot (FBG-FP) filter, which matches the wavelength of RBGA, is specially designed. Due to the narrower bandwidth, flatter response and steeper rolloff, the dual-cavity FBG-FP filter can significantly suppress the competition-induced frequency jitter and undesired residual random lasing modes, and contribute to establish stable lasing and further reduce frequency noise of RFL. The frequency noise of the proposed RFL is ∼3 Hz/√Hz above 1 kHz, and the frequency jitter is ∼3.11 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−11</sup> over 10 s. The 3-dB linewidth is ∼490 Hz with a ∼40 dB of optical signal-to-noise ratio.

Optimization of the location of piezoelectric actuator and sensor in active vibration control using Multi-Verse Optimizer algorithm

In the process of using piezoelectric actuators and sensors to actively control the vibration of flexible structures, the controllability and spillover effect of the system are often the focus of attention, and the location of piezoelectric actuators and sensors is one of the key factors affecting controllability and spillover effect. In this paper, the residual modes are introduced into the location optimization criteria of actuators and sensors, so that the energy can drive as much as possible the eigenmodes that need to be controlled, and avoid driving the residual eigenmodes. At the same time, a modified Multi-Verse Optimizer (MVO) algorithm is used to optimize the location of piezoelectric actuators and sensors. The important results obtained in this paper are compared with the results obtained by the finite element simulation software to verify the accuracy of the relevant derivation results, and the optimization simulation of the position of piezoelectric actuator and the simulation of vibration control are carried out. The simulation results showed that, the MVO algorithm is very suitable for solving the optimization problems that these criteria are non-convex and difficult to derive, and the resulting optimized layout can effectively suppress vibration.

Compact and broadband silicon TE-pass polarizer based on tapered directional coupler.

We demonstrate a novel TE-pass polarizer, to the best of our knowledge, on a silicon-on-insulator (SOI) platform. The device's working principle is based on the phase-matched coupling of the unwanted TM0 mode in an input waveguide to the TM1 mode in a tapered directional coupler (DC), which is then guided through a low-loss bend (180-degree) and scattered in a terminator section with low back reflections. However, the input TE0 mode is routed through the tapered section uncoupled with negligible loss. An S-bend is added before the output for filtering any residual TM0 mode present in the input waveguide. Tapering the DC helps maintain phase matching for broadband operation and increases the tolerance toward fabrication errors. The measurement shows low insertion loss (IL < 0.44 dB), high extinction ratio (ER > 15 dB), and wide bandwidth (BW = 80 nm). The overall device length is only 13 µm. A high performing TE-pass polarizer (IL < 0.89, ER > 30, and BW = 100 nm) is also demonstrated by cascading two proposed polarizers.

A refined Bloch mode synthesis framework for fast and accurate analysis of electroelastic metamaterials with piezoelectric resonant shunt damping

Bloch mode synthesis (BMS) is a model order reduction method that was originally formulated to accelerate band-structure calculations of periodic elastic materials. In this paper, the BMS framework is expanded to offer a platform suitable for analysis of electroelastic metamaterials with piezoelectric resonant shunt damping. First, a transformation of coordinates formed by a set of dominant fixed-interface modes and a set of static-constraint modes is used in combination with Bloch periodicity boundary conditions to derive the classical BMS method in the context of electroelastic metamaterials. Then, a set of residual modes that is simply truncated in the classical BMS, is augmented into the reduction process, leading to an enhanced BMS method, denoted RMH-BMS.Currently available model order reduction techniques for electroelastic metamaterials adopt a reduction basis obtained from one of the limiting eigenvalue problems associated with short-circuit (SC) and open-circuit (OC) boundary conditions. In either case, the electrical circuit parameters are ignored during modal reduction. In this paper, it is shown that such approximation leads to inaccurate band-structure predictions for electroelastic metamaterials equipped with periodic array of resonant shunt circuits. To this end, two versions of the proposed BMS methods are evaluated and compared. In the first version, the reduced order model is constructed following the widely used OC basis approach. The second version, on the other hand, exploits a coupled basis approach, which requires knowledge of the electrical circuit parameters prior to the interior modal reduction step. A comprehensive error analysis is conducted to assess the computational performances of the BMS methods, and the results revealed that whether the classical BMS or the RMH-BMS is employed to construct the reduced order model, the coupled basis approach produces significantly more accurate results compared to the OC basis approach. The results presented in this work expose the limitations of implementing the OC basis approach in wave propagation analysis of electroelastic metamaterials with piezoelectric resonant shunt damping and shed light on the associated consequences in predicting the bandgap properties of such metamaterials.

A Hybrid Data-Driven Approach for Multistep Ahead Prediction of State of Health and Remaining Useful Life of Lithium-Ion Batteries.

In this paper, a novel multistep ahead predictor based upon a fusion of kernel recursive least square (KRLS) and Gaussian process regression (GPR) is proposed for the accurate prediction of the state of health (SoH) and remaining useful life (RUL) of lithium-ion batteries. The empirical mode decomposition is utilized to divide the battery capacity into local regeneration (intrinsic mode functions) and global degradation (residual). The KRLS and GPR submodels are employed to track the residual and intrinsic mode functions. For RUL, the KRLS predicted residual signal is utilized. The online available experimental battery aging data are used for the evaluation of the proposed model. The comparison analysis with other methodologies (i.e., GPR, KRLS, empirical mode decomposition with GPR, and empirical mode decomposition with KRLS) reveals the distinctiveness and superiority of the proposed approach. For 1-step ahead prediction, the proposed method tracks the trajectory with the root mean square error (RMSE) of 0.2299, and the increase of only 0.2243 RMSE is noted for 30-step ahead prediction. The RUL prediction using residual signal shows an increase of 3 to 5% in accuracy. This proposed methodology is a prospective approach for an efficient battery health prognostic.

Mode selection and amplification from optical frequency combs by optical injection locking: a numerical study of noise.

Noise, except for residual side modes, in mode selection and amplification from optical frequency combs by optical injection locking has not been studied in detail, to our knowledge. We report a numerical study of noise behaviors, including both residual side modes and the noise between them. It reveals that comb laser injection can introduce excessive noise, compared to optical injection with a single mode master laser through the interference between residual side modes and resonances within Arnol'd tongues created by periodic forcing in the optical phase, which can be a severe problem for the case of sub-GHz comb mode spacing. The dependences of residual side mode rejection and phase noise variance on comb mode spacing, seeding power, and detuning are discussed to help in low-noise system design and operation.

A 12.1 TOPS/W Quantized Network Acceleration Processor With Effective-Weight-Based Convolution and Error-Compensation-Based Prediction

In this article, a quantized network acceleration processor (QNAP) is proposed to efficiently accelerate CNN processing by eliminating most unessential operations based on algorithm-hardware co-optimizations. First, an effective-weight-based convolution (EWC) is proposed to distinguish a group of effective weights (EWs) to replace the other unique weights. Therefore, the input activations corresponding to the same EW can be accumulated first and then multiplied by the EW to reduce amounts of multiplication operations, which is efficiently supported by the dedicated process elements in QNAP. The experimental results show that energy efficiency is improved by 1.59x-3.20x compared with different UCNN implementations. Second, an error-compensation-based prediction (ECP) method adopts trained compensated values to replace partly unimportant partial sums to further reduce potentially redundant addition operations caused by the ReLU function. Compared with SnaPEA and Pred on AlexNet, 1.23x and 1.75x higher energy efficiencies (TOPS/W) are achieved by ECP, respectively, with marginal accuracy loss. Third, the residual pipeline mode is proposed to efficiently implement residual blocks with a 1.5x lower memory footprint, a 1.18x lower power consumption, and a 13.15% higher hardware utilization on average than existing works. Finally, the QNAP processor is fabricated in the TSMC 28-nm CMOS process with a core area of 1.9 mm². Benchmarked with AlexNet, VGGNet, GoogLeNet, and ResNet on ImageNet at 470 MHz and 0.9 V, the processor achieves 117.4 frames per second with 131.6-mW power consumption on average, which outperforms the state-of-the-art processors by 1.77x-24.20x in energy efficiency.

Empirical Mode Decomposition and Fourier analysis of Caspian Sea level's time series

The present study described Caspian Sea's water level changes by using the recorded data in three ports namely Anzali, Nowshahr, and Amirabad. The data are the mean values of daily records within a three-year period. Primary determinations revealed that the water level drops to the lowest amount in the winter but then starts to increase towards the spring. Afterwards, the water level peaks in the summer and again it decreases in the autumn. It was also detected that water level decreased 22 cm annually in the given period. Fourier Transform and Empirical Mode Decomposition (EMD) were adopted for the decomposition of time series. Fourier analysis indicated that the dominant sinusoidal component possessed a period and amplitude of 294 days and 0.148 m, respectively. EMD yielded six Intrinsic Mode Functions (IMFs) and one residual mode. The average period of each IMF was calculated through zero-down crossing method. Afterwards, the spectrum of each IMF and their summit characteristics were extracted. A maximum value of 0.127 m was observed at the frequency of 0.0034 cycle/day in the IMF6 spectrum, which perfectly corresponds to the dominant sinusoidal component in time series.

A free‐interface component mode synthesis of moving contact multibodies and its application to gear contacts

AbstractMoving contact components are common in mechanical systems, and they exhibit a large rigid motion with linear‐elastic deflection under contact boundary conditions. The finite element method can capture local deformation and high gradient stresses, but finely meshed contact interfaces lead to numerous degrees of freedom. Model order reduction techniques for multibodies are effective, but moving contact interfaces also introduce multiple interface degrees of freedom. In the present method, the interface degrees of freedom are retained to ensure the accuracy of the local response induced by contacts, but the computational scale is not further enlarged. The motion of moving contact components is described by the floating frame of reference formulation. A free‐interface component mode synthesis method is proposed for model order reduction. Kept free‐interface normal modes represent global flexibility and residual modes accurately compensate for the local quasi‐static responses of omitted modes induced by contacts. The contribution of residual modes is represented by interface load coordinates, which can be directly evaluated by contact forces. The contacts are calculated with the present contact algorithm. The accuracy and efficiency of the proposed method are fully demonstrated through correlation with the finite element method in gear meshing simulations.

Ultra-broadband LP11 mode converter with high purity based on long-period fiber grating and an integrated Y-junction.

We report an ultra-broadband LP11 mode converter with high purity based on integrated two shunt-wound long-period fiber gratings (LPFGs) and an adiabatic Y-junction, together with a high-order-mode bandpass filter. Two shunt-wound LPFGs are inscribed by CO2 laser in a two-mode fiber to achieve a 10 dB bandwidth of 50 nm and 51 nm at resonance wavelengths of 1530 nm and 1570 nm, respectively. Meanwhile, the Y-junction fabricated by lithography can be operated over S + C+L band to combine the converted LP11 mode. The presented ultra-broadband mode converter is able to achieve a mode conversion efficiency of 95%, together with a wavelength-dependent loss of less than 3 dB over the S + C+L band. This device has low modal crosstalk of 17 dB between the LP01 and LP11 modes, because most of the residual LP01 mode is further filtered by a high-order-mode bandpass filter at the output port of the Y-junction. The insertion loss of mode converter is estimated to be lower than 2.7 dB, due to the use of low loss polymer material during the fabrication. The proposed ultra-broadband LP11 mode converter with high purity is promising for the application of ultra-broadband mode-division-multiplexing transmission systems.

Fire-exposed stones in constructions: Residual strength, performance loss and damage mode shift due to mineralogical transformation and micro-cracking

The heritage restoration and the design of modern stone masonry structures bear in mind the material response to fire, the residual strength and post-fire resilience. In this attempt, the decomposition of fire-exposed natural and engineered stones and their mineralogical transformations are reported. The fire-induced petrographic change is further correlated to rise in shrinkage and porosity, and loss of mechanical properties, such as strength, toughness and stiffness. Stones of different nature (limestone, sandstone, quartzite, polymer/quartz-based engineered stone) are selected to trigger various thermal degradation mechanisms. Also, the stones were sorted considering their sustainability profile since all stones, except the artificial ones, were collected from local quarries and linked to the regional heritage and architecture. An integrated inspection methodology is designed that implements non-destructive monitoring on mechanical tests. A novel proof-loading concept is introduced at which the residual performance is assessed exclusively based on ultrasound pulse velocity (UPV) and early acoustic emission (AE) data. The damage mechanisms that dominate fracture and the shift of failure from strong brittle to weak pseudo-ductile with an early transition from crack initiation to propagation are determined based on AE features analysis.

Cavity-induced exciton localization and polariton blockade in two-dimensional semiconductors coupled to an electromagnetic resonator

Recent experiments have demonstrated strong light-matter coupling between electromagnetic nanoresonators and pristine sheets of two-dimensional semiconductors, and it has been speculated whether these systems can enter the quantum regime operating at the few-polariton level. To address this question, we present a first-principles microscopic quantum theory for the interaction between excitons in an infinite sheet of two-dimensional material and a localised electromagnetic resonator. We find that the light-matter interaction breaks the symmetry of the otherwise translation-invariant system and thereby effectively generates a localised exciton mode, which is coupled to an environment of residual exciton modes. This dissipative coupling increases with tighter lateral confinement, and our analysis reveals this to be a potential challenge in realising nonlinear exciton-exciton interaction. Nonetheless, we predict that polariton blockade due to nonlinear exciton-exciton interactions is well within reach for nanoresonators coupled to transition-metal dichalcogenides, provided that the lateral resonator mode confinement can be sufficiently small that the nonlinearity overcomes the polariton dephasing caused by phonon interactions.

Combining Residual and LSTM Recurrent Networks for Transportation Mode Detection Using Multimodal Sensors Integrated in Smartphones

In recent years, with the rapid development of public transportation, the ways people travel has become more diversified and complicated. Transportation mode detection, as a significant branch of human activity recognition (HAR), is of great importance in analyzing human travel patterns, traffic prediction and planning. Though many works have been devoted to transportation mode detection, there remains challenge for accurate and robust transportation pattern identification. In this paper, we propose a residual and LSTM recurrent networks-based transportation mode detection algorithm using multiple light-weight sensors integrated in commodity smartphones. Feature representation learning is adopted separately on multiple preprocessed sensor data using deep residual and LSTM network, which can enhance the identification accuracy and support one or more sensors. Residual units are introduced to accelerate the learning speed and enhance the accuracy of transportation mode detection. Furthermore, we also leverage the attention model to learn the significance of different features and different timesteps to enhance the recognition accuracy. Extensive experimental results on three datasets indicate that using our proposed model can achieve the best recognition accuracy for eight transportation modes including being stationary, walking, running, cycling, taking a car, taking a bus, taking a subway and taking a train, which outperforms other benchmark algorithms.