Compressible Model Research Articles

Mechanical properties and environmental implications of excess-sulfate cement concrete with phosphogypsum-based cold-bonded fine aggregates

Recycling waste phosphogypsum (PG) into phosphogypsum-based cold-bonded aggregates (PCBAs) for eco-concrete production has received extensive attention due to its sustainability for construction industry development. Currently, PCBAs mainly replace the coarse aggregates in concrete while the fine aggregates remain natural sand (NS). This study substitutes NS with phosphogypsum-based cold-bonded fine aggregates (PCBFAs) for the preparation of excess-sulfate cement concrete (ESCC), and the compressive strength, tensile splitting strength, flexural strength, and stress-strain curve were studied by considering replacement rations (0%, 25%, 50%, 75%, 100%). The results indicate that when the replacement ratio is 100%, the compressive strength, splitting tensile strength, and flexural strength of ESCC were reduced by 17.2%, 26.9%, and 29.2% respectively. The addition of PCBFAs will make the ascending segment of the stress-strain curve flatter and make the descending segment steeper. Based on the tested results, the prediction formulas of the mechanical properties and the uniaxial compressive stress-strain model of the concrete under different replacement ratios are proposed. Microstructure analysis indicates that the PCBFAs gradually integrate with the cement matrix, reducing the defects in the interface transition zone. Furthermore, ESCC effectively immobilizes impurities contained in PG. The research findings demonstrate the feasibility of using PCBFAs as a substitute for NS.

PSD estimation and modal parameter identification for random vibrations with blade tip timing measurement

Abstract Blade tip timing (BTT) is a vibration measurement technique for blade health monitoring. Most of the existing BTT analysis methods are suitable for deterministic vibration signals but are ineffective for random vibration signals that often occur in practice. Statistical analysis of BTT data is significant for random vibration analysis and improving blade monitoring efficiency. This study proposes a compressive model for power spectral density (PSD) estimation and modal parameter identification. The efficiencies of three compressive sensing algorithms, including the least absolute shrinkage and selection operator (Lasso), nonnegative least squares (NLS), and nonnegative orthogonal matching pursuit, are compared. The effects of the duration of the signal and the frequency resolution on the quality of the estimated PSD and the identified parameters are discussed. According to the analysis, to obtain accurate damping ratios, it is recommended that the duration of the signal be greater than 3000 revolutions. A Q criterion based on the half-power bandwidth is proposed to determine the set of frequency resolutions. Numerical and field tests were conducted to verify the proposed method. The results indicate that the NLS algorithm is recommended to use. The root-mean-square errors of the identified natural frequencies and damping ratios obtained by the proposed method were 0.065 Hz and 0.023%, respectively. The proposed method was verified at different rotational speeds in a field test, demonstrating the capability of the method over a wide rotational speed range and providing more opportunities to detect blade damage.

A multiaxial multiscale compressive thermo-mechanical damage model for Fiber Reinforced Cementitious Composites (FRCC)

In the current study, a multiaxial multiscale compressive stress-strain model for Fiber Reinforced Cementitious Composites (FRCC) was innovatively developed. The proposed model was on the basis of the theory of thermodynamics, multiscale theory and internal variable theory. The free energy function and dissipation function were established to characterize the elastic and plastic deformation of FRCC, respectively. The multiscale behaviors of FRCC in meso-scale and macro-scale was taken into account. In meso-scale, the energy dissipation functions of fiber in tension and bond-slip deformation in the interface transition zone (ITZ) were established. The resistance to cracking of cementitious matrix through the fiber bond stress were taken into account. In macro-scale, the thermo-mechanical coupling contributions of plastic damage, thermal damage, yield surface and hardening evolution were creatively considered in the potential function within thermodynamics framework. The influence of types and volume fractions of fibers on the reinforcement mechanism was discussed. When the temperature is higher than the melting point of fiber, the fiber influence coefficients was employed. The sensitivity of yield surface was discussed. The numerical predictions of this developed model were carried out for validation. The nonlinear stress-strain responses of concrete were accurately captured at temperatures ranges from 20 °C to 800 °C compared with the experimental data, which indicates the accuracy and applicability of the proposed model.

Numerical investigation on Taylor bubble mass transfer in microchannel based on CO2 gas with the consideration of gas compressibility

This paper focused on the mass transfer of Taylor bubbles composed of pure carbon dioxide in organic solvent in microchannels. Combined with the Henry’s law, the carbon dioxide concentration at the two-phase interface was dynamically updated by using a compressible fluid model and considering the variation of gas density in the bubble. The critical bubble shape factor could help to judge the dominant mechanism of bubble deformation. The entire mass transfer process was divided into three stages: the rapid dissolution stage, the two-end dissolution stage, and the diffusion-like dissolution stage. The gas–liquid mass transfer capacity and influencing factors of each stage were analyzed, and the dominant mass transfer mechanism of each stage was summarized. It was found that the non-uniform degree of gas density in the bubble was related to the bubble Reynolds number and bubble volume by analyzing the distribution variation of solute density.

A Pruning and Distillation Based Compression Method for Sonar Image Detection Models

Accurate underwater target detection is crucial for the operation of autonomous underwater vehicles (AUVs), enhancing their environmental awareness and target search and rescue capabilities. Current deep learning-based detection models are typically large, requiring substantial storage and computational resources. However, the limited space on AUVs poses significant challenges for deploying these models on the embedded processors. Therefore, research on model compression is of great practical importance, aiming to reduce model parameters and computational load without significantly sacrificing accuracy. To address the challenge of deploying large detection models, this paper introduces an automated pruning method based on dependency graphs and successfully implements efficient pruning on the YOLOv7 model. To mitigate the accuracy degradation caused by extensive pruning, we design a hybrid distillation method that combines output-based and feature-based distillation techniques, thereby improving the detection accuracy of the pruned model. Finally, we deploy the compressed model on an embedded processor within an AUV to evaluate its performance. Multiple experiments confirm the effectiveness of our proposed method in practical applications.

Mechanical properties, micro-mechanisms, and constitutive models of seawater sea-sand engineered cementitious composites

Three mixing proportions of engineered cementitious composites (ECCs) were prepared using seawater, sea-sand, and PVA fibers, corresponding to three strength grades of C30, C40, and C50, respectively. The mechanical properties of these three kinds of seawater sea-sand engineered cementitious composites (SS-ECCs), including compressive strength, tensile strength, flexural strength, elastic modulus, ultimate elongation, uniaxial compressive and tensile stress-strain relationship, and so on, were analyzed by compressive, tensile, and flexural tests. The tensile and flexural strength of SS-ECCs are influenced by their ductility. The SS-ECC with C40 strength grade exhibited the best ductility, with an average ultimate tensile strain reaching 3.42 %, and possessed the highest tensile and flexural strength. The C50 specimen had the worst ductility, but its tensile strength and flexural strength came in second place. As the strength grade increased, the elastic modulus gradually increased, while the Poisson's ratio decreased. Uniaxial compressive and tensile constitutive models for SS-ECCs were established. The uniaxial compressive constitutive model was divided into a nonlinear ascending stage and a bilinear descending stage. XRD and TGA results revealed that the improvement in the strength of SS-ECCs was mainly attributed to the increased content of Ca(OH)2 in the reaction products. Additionally, the gypsum content in the raw materials decreased with the increasing strength grade, resulting in a reduction in the generation of ettringite (AFt).

Hyperforin improves matrix stiffness induced nucleus pulposus inflammatory degeneration by activating mitochondrial fission

ObjectiveThe continuously increasing extracellular matrix stiffness during intervertebral disc degeneration promotes disease progression. In an attempt to obtain novel treatment methods, this study aims to investigate the changes in nucleus pulposus cells under the stimulation of a stiff microenvironment. DesignRNA sequencing and metabolomics experiments were combined to evaluate the primary nucleus pulposus and screen key targets under mechanical biological stimulation. Additionally, small molecules work in vitro were used to confirm the target regulatory effect and investigate the mechanism. In vivo, treatment effects were validated using a rat caudal vertebrae compression model. ResultsOur research results revealed that by activating TRPC6, hyperforin, a herbaceous extract can rescue the inflammatory phenotype caused by the stiff microenvironment, hence reducing intervertebral disc degeneration (IDD). Mechanically, it activates mitochondrial fission to inhibit PFKFB3. ConclusionIn summary, this study reveals the important bridging role of TRPC6 between mechanical stiffness, metabolism, and inflammation in the context of nucleus pulposus degeneration. TRPC6 activation with hyperforin may become a promising treatment for IDD.

Research on the residual compressive properties of concrete with preload after sustained high temperatures

The containment vessel of nuclear power plants, concrete structures near the furnaces and roller lines, and chimneys, are exposed to high temperatures (120 °C to 350 °C) for long term. Research on the compressive properties of concrete after short-term high temperatures (usually within 3 h) has been widely conducted, while that of concrete after sustained high temperatures is scarce. In this research, the effects of temperature (120 °C to 350 °C), heating duration (3 h to 32 h) and initial stress level (0.2 to 0.6, ratio of initial applied stress to compressive strength at room temperature) on the failure mode, compressive strength, peak strain and stress-strain relationship of concrete after sustained high temperatures were studied. The results show that with increasing temperature and heating duration, concrete compressive strength gradually decreases, while concrete peak strain gradually increases. At heating duration of 24 h, the properties of concrete have basically stabilised. Preloading helps the concrete to resist the adverse effects of high temperatures by inhibiting the dehydration of cement matrix and the cracking of interfacial transition zone. For concrete with preload, transient thermal strain and short-term creep at high temperatures could not be recovered after cooling, which causes a significant increase in the peak strain of concrete. A compressive stress-strain model for concrete with preload after sustained high temperatures was developed, which consists of pre-creep ascending segment, creep segment, post-creep ascending segment and descending segment. Comparison of the prediction results with the test results proves that the model has high accuracy.

Experimental and numerical investigation of the fracture behavior in granite specimens with two co-planar side flaws under biaxial loading

It is well known that the structural plane plays a decisive role in the mechanical properties of rock mass. To better understand the cracking mechanism and stress characteristics of co-planar intermittent double-flawed granite specimens with various dip angles (Abbreviated as flawed specimens), two co-planar flaws were prefabricated on both sides of the granite, and biaxial compression tests were conducted. The failure characteristics and local microstrain were recorded by the camera and strain acquisition instrument. The evolution law of mechanical parameters of flawed specimens was analyzed. Simultaneously, a two-dimensional particle flow code (PFC2D) was used to establish a biaxial compression model of flawed specimens. The crack initiation and propagation of the specimens were explained by the crack hot spot, the evolution law of crack, and the evolution law of the number of microcracks. It is found that (a) as flaw dip angle (α) increases, peak strength and elastic modulus initially decrease and then increase, with reductions ranging from 17.6% to 46.1% and 15.6% to 48.2%, respectively. (b) Increasing lateral stress (σ2) triggers a shift in failure mode from a tensile-shear mixed failure to a shear failure. The cohesion (c) and friction angle (φ) of flawed specimens were notably smaller compared to those of intact specimens, with reductions correlating with the α. (c) Flawed specimens exhibit pronounced contact force concentration and stress concentration at the flaw tip. The analysis of the microstrain confirms stress concentration at the flaw tip, leading to failure. However, various σ2 induces the differences in the magnitude and extent of tip stress.

Learned Video Compression with Adaptive Temporal Prior and Decoded Motion-aided Quality Enhancement

Learned video compression has drawn great attention and shown promising compression performance recently. In this article, we focus on the two components in the learned video compression framework, the conditional entropy model and quality enhancement module, to improve compression performance. Specifically, we propose an adaptive spatial-temporal entropy model for image, motion, and residual compression, which introduces a temporal prior to reduce temporal redundancy of latents and an additional modulated mask to evaluate the similarity and perform refinement. In addition, a quality enhancement module is proposed for predicted frame and reconstructed frame to improve frame quality and reduce the bitrate cost of residual coding. The module reuses decoded optical flow as a motion prior and utilizes deformable convolution to mine high-quality information from the reference frame in a bit-free manner. The two proposed coding tools are integrated into a pixel-domain residual coding–based compression framework to evaluate their effectiveness. Experimental results demonstrate that our framework achieves competitive compression performance in the low-delay scenario compared with recent learning-based methods and traditional H.265/HEVC in terms of Peak Signal-to-Noise Ratio (PSNR) and Multi-Scale Structural Similarity Index (MS-SSIM). The code is available at OpenLVC.

Аналитическая модель малых колебаний сжимаемой магмы с реологией Максвелла в питающей системе вулкана. Часть 2. Осцилляции вертикальной скорости

The analytical solution for vertical magma movements in a volcanic conduit within the occurrence of low-frequency volcanic seismic events is presented. Magma is described by Maxwell's compressible body model. When the density of the magmatic melt is disturbed, for example, when dense magma enters from deep layers or the melt degasses at a certain depth, density oscillations may occur in the channel as a reaction to this event. For the magma conduit of the simplest cylindrical shape, the magma density and two components of the velocity of movement are subject to oscillations. In this case, the vertical component of the velocity experiences forced oscillations, both under the influence of density oscillations and under the influence of the initiating disturbance. All these oscillations are harmonic damped oscillations, the damping coefficient of which is determined by the relaxation time of the magmatic melt, and the natural frequency depends on the physical characteristics of the magmatic melt and the geometric dimensions of the conduit. Melt density oscillations lead to periodic variations in the lithostatic pressure drop, which in turn causes vertical movements of the melt, the most amplitude along the axis of the magma conduit. The model is used to describe crater surface displacements observed on the surface of the Santiaguito volcano crater.

Real-time modeling of the riveting process forces for aircraft panel structures

The riveting of large panels in aircraft manufacturing requires an automatic drilling and riveting machine capable of precise and stable operation. However, the significant nonlinearity and time-varying nature of riveting process forces severely impact the control capacity and operational accuracy of these machines. So, the objective is to provide technical assurance for the precise operation of the machine and ensure high-quality riveting of the panels. Therefore, a real-time modeling method for the automatic riveting process forces on aircraft panels is proposed, which focuses on the rivet die's downward compression displacement as the core variable and accounts for the non-uniform deformation of rivets. This method assumes a constant volume for the rivet rod. It combines power-law hardening and Coulomb's friction theories to establish three sub-models for different riveting stages: the elastic stage, the plastic deformation stage, and the driven head formation stage. Further, considering the continuity of rivet's deformation, rivet rod's cross-section stress, and the driven head dimensions changes with the driven head's downward compression displacement, a novel real-time continuous dynamic load model for the full compression riveting process of a single-rivet is developed. The results indicate that the model is effective, with a maximum deviation of 7.50 % between experimental and predicted values, demonstrating its accuracy and reliability. In conclusion, this research provides a real-time, continuous description of the riveting process force, enabling automatic machines to promptly detect defects and deformations in the riveting.

Nonequilibrium–diffusion limit of the compressible Euler radiation model in ℝ3

This article studies the nonequilibrium–diffusion limit of the compressible Euler model arising in radiation hydrodynamics in with the general initial data. Combining the moment method with Hilbert expansion, we show that the radiative intensity can be approximated by the sum of interior solution and initial layer. We also show that the solution satisfied by the density, temperature, and velocity can be approximated by the interior solutions. Our results can be considered as an extension from in arXiv.2312.15208 by Ju, Li, and Zhang to . In contrast to arXiv.2312.15208, we get the exact convergence rates by studying the error system derived from the primitive system, the zeroth‐order to the second‐order about the radiative intensity, and the zeroth‐order about the hydrodynamics.

Understanding cavitation bubble collapse and rebound near a solid wall

The collapse of cavitation bubbles is known for its intense and aggressive behavior, producing concentrated bursts of energy that can cause erosion, noise, and damage to metallic surfaces and mechanical equipment. However, this concentrated energy also offers significant potential for various applications in biomedical science, industrial engineering, and related technologies. This study presents a comprehensive numerical investigation of the collapse and rebound of cavitation bubbles near a wall using a conservative, compressible multiphase flow model. Laser-generated cavitation bubbles growing, collapsing, and rebounding in distilled water near a wall at three standoff distances were experimentally studied to validate the numerical method. The results show good agreement between the bubble shapes and radii from the numerical simulations and experiments. Additionally, velocity of microjets and thermodynamics of the bubble collapse were validated against the reference data in the literature. Subsequently, the pressure and temperature distributions and jet shapes were examined in detail, exploring the impact of standoff distances ranging from 0.2 to 1.8. A quantitative analysis of the pressure and temperature produced on the wall, as well as the jet velocity and bubble movement during the collapse, was systematically conducted. Jet velocities observed ranged from 30 m/s to 130 m/s, and numerical simulations were compared with experimental data from the literature, showing good agreement. Maximum temperature and pressure induced at the wall center can reach up to 616.2 K and 139.5 MPa, respectively. The detailed analyses in this study provide significant insights into the behavior of cavitation bubbles and the effect of standoff distance on their collapse and rebound near a wall.

Compressive stress-strain model for concrete-filled filament-wound FRP tubes with arbitrary fiber orientation: An analytical approach

An analytical model is proposed to predict the non-linear hoop stress- strain response of multi-laminated fiber reinforced polymer (FRP) tube when subjected to internal pressure. This model incorporates the arbitrary stacking of fiber direction and thickness of individual lamina in FRP tube to effectively calculate confinement pressure exerted by FRP tube on solid concrete columns under uniaxial compression. Next, the model is augmented with Jiang and Teng’s analysis-oriented model to predict the axial behavior of concrete filled FRP tube solid columns with fibers oriented in any arbitrary direction. The predictions by the proposed model are then compared with numerous experimental axial stress-strain/load-strain response of FRP-confined concrete columns collected from the literature. It is found that the present model can closely predicts with the experimental outcomes, outperforming several existing analytical models.

Joint parameter estimation algorithm for polarization-sensitive channel compression arrays based on atomic norm minimization

To reduce the complexity of direction-of-arrival (DOA) algorithms in polarization-sensitive arrays and maintain high estimation accuracy while decreasing runtime, this paper proposes a channel compression model based on orthogonal dipole arrays. In addition, a DOA estimation algorithm using atomic norm minimization (ANM) is introduced for this model. This algorithm performs eigenvalue decomposition on the covariance matrix of the data received from the polarization-sensitive channel compressed array. A new observation vector is constructed for the ANM problem under the channel compression model using the obtained eigenvalues and eigenvectors. Subsequently, a Toeplitz matrix is constructed from the optimal solution of the semi-positive definite programming (SDP) problem. The Vandermonde decomposition of the Toeplitz matrix provides estimates of the signal DOA parameters. By combining the vectorization result of the covariance matrix and the least-squares calculation, the polarization auxiliary angle and polarization phase angle information of the incident source are also obtained. Simulation experiments compare the root-mean-squared error (RMSE) of each algorithm at different angular intervals and signal-to-noise ratios (SNR) to confirm the validity of estimating DOA and polarization parameters.

Discrete element modeling and experimental study of biomechanical properties of cotton stalks in machine-harvested film-stalk mixtures

To address the current problems of low accuracy and poor reliability of the discrete element model of cotton stalks, as well as the difficulty of guiding the design and optimization of the equipment through simulations, the discrete element modeling and physical-mechanical tests of cotton stalks in machine harvested film-stalk mixtures are carried out. The peak tensile force Fjmax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm j}^{\\max }$$\\end{document}, the peak pressure Fymax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm y}^{\\max }$$\\end{document}, the peak bending force Fwmax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm w}^{\\max }$$\\end{document}, the peak shear force Fjmax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm j}^{\\max }$$\\end{document}, and the force-displacement (F–x) curves of cotton stalks are obtained from the physical tests. The discrete element model of double-layer cotton stalks based on the flat-joint model is established with the PFC3D\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$^{\\rm 3D}$$\\end{document} software. The Fymax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm y}^{\\max }$$\\end{document} is taken as the response value, and the microscopic parameters of the cotton stalk model are used as the test factors, then the Plackett–Burman test, the steepest climb test, and the Box–Behnken test are sequentially designed using Design-Expert software. The second-order regression model describing the relationship between the Fymax\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$F_{\\rm y}^{\\max }$$\\end{document} and the microscopic parameters is established. The optimal parameter combinations of the microscopic parameters are obtained, and then they are utilized to construct the compression, bending, and shear models of cotton stalks and to carry out the validation tests. The results confirm that the established discrete element model could accurately characterize the biomechanical properties of cotton stalks and that the parameter calibration method is reasonable, which could provide a reference for the discrete element modeling of cotton stalks and other stalks, and also offer a theoretical basis for the research of the crushing and separation mechanism of the film-stalk mixtures and the development of the equipment.

Information Density Enhancement Using Lossy Compression in DNA Data Storage.

This study develops two deoxyribonucleic acid (DNA)lossy compression models, Models A and B, to encode grayscale images into DNA sequences, enhance information density, and enable high-fidelity image recovery. These models, distinguished by their handling of pixel domains and interpolation methods, offer a novel approach to data storage for DNA. Model A processes pixels in overlapped domains using linear interpolation (LI), whereas Model B uses non-overlapped domains with nearest-neighbor interpolation (NNI). Through a comparative analysis with Joint Photographic Experts Group (JPEG)compression, the DNA lossy compression models demonstrate competitive advantages in terms of information density and image quality restoration. The application of these models to the Modified National Institute of Standards and Technology (MNIST) dataset reveals their efficiency and the recognizability of decompressed images, which is validated by convolutional neural network (CNN) performance. In particular, Model B2, a version of Model B, emerges as an effective method for balancing high information density (surpassing over 20 times the typical densities of two bits per nucleotide) with reasonably good image quality. These findings highlight the potential of DNA-based data storage systems for high-density and efficient compression, indicating a promising future for biological data storage solutions.

Concrete Compressive Strength Prediction by Ensemble Machine Learning Approach

The prediction of concrete compressive strength is a crucial aspect of ensuring the structural integrity and durability of construction projects. In recent years, machine learning approaches have improved upon the limitations of empirical formulas and laboratory testing methods for predicting concrete compressive strength. This study utilizes ensemble machine-learning techniques, such as Bagging, XGBoost, and Stacking models, to enhance the accuracy of concrete compressive strength prediction models. A five-fold cross-validation technique was applied to mitigate the problems of underfitting or overfitting in the regression model. Furthermore, various statistical indices were employed to compare the forecasting performance of these ensemble techniques. The prediction performance of this research revealed that XGBoost achieved the highest R-squared value of 93%, followed by Stacking and Bagging regression models at 92%. Consequently, this research underscores the potential of ensemble techniques as valuable tools in the domain of civil engineering, paving the way for more reliable and efficient construction practices.

High Performance CNN Accelerators based on Hardware and Algorithm Co-Optimization

The efficacy of convolutional neural networks (CNNs) has led to their extensive application in picture recognition and categorization. Nevertheless, embedded systems face challenges when it comes to storing the massive volumes of weight data required by CNNs in their on-chip memory. Pruning a convolutional neural network (CNN) model can reduce its size with no impact on accuracy; however, when used with a parallel architecture, the resulting model will be slower to execute. A CNN compression method that is hardware-centric is introduced in this paper. A model of a deep neural network (DNN) will include "pruning layers (P-layers)" and "no-pruning layers (NP-layers)" attached to it. For effective and parallel processing, an NP-layer employs a regular distribution for its weights. Pruning causes a P-layer to have an irregular form, yet the high compression ratio it yields makes the shape undesirable. Using uniform and gradual quantization methods, we can achieve a processing efficiency/compression ratio trade-off with a small loss of precision. Integrating multiple parallel finite impulse response (FIR) filters into an NP-layer distributed convolutional architecture, we further improve the regular model. The P-layer irregular sparse model is suggested to use an ADF-based processing element. The suggested compression strategy and hardware architecture can be utilised by a hardware/algorithm co-optimization (HACO) method to run an NP→P hybrid compressed CNN model on FPGAs. The VGG-16 network achieves a compression ratio of 27.5× with a top-5 accuracy loss of 0.44% by using a hardware accelerator on a single FPGA chip without off-chip memory. A result of 83.0 FPS, or 1.8 times quicker, was achieved when the compressed VGG-16 model was implemented on a Xilinx VCU118 evaluation board for image applications. Index Terms - CNN, Accelerators, Co-Optimization.