A Novel Quantization and Model Compression Approach for Hardware Accelerators in Edge Computing

Massive computational complexity and memory requirement of artificial intelligence models impede their deploy-ability on edge computing devices of the Internet of Things(IoT).While Power-of-Two(PoT)quantization is pro-posed to improve the efficiency for edge inference of Deep Neural Networks(DNNs),existing PoT schemes require a huge amount of bit-wise manipulation and have large memory overhead,and their efficiency is bounded by the bottleneck of computation latency and memory footprint.To tackle this challenge,we present an efficient inference approach on the basis of PoT quantization and model compression.An integer-only scalar PoT quantization(IOS-PoT)is designed jointly with a distribution loss regularizer,wherein the regularizer minimizes quantization errors and training disturbances.Additionally,two-stage model compression is developed to effectively reduce memory requirement,and alleviate bandwidth usage in communications of networked heterogenous learning systems.The product look-up table(P-LUT)inference scheme is leveraged to replace bit-shifting with only indexing and addition operations for achieving low-latency computation and implementing efficient edge accelerators.Finally,comprehensive experiments on Residual Networks(ResNets)and efficient architectures with Canadian Institute for Advanced Research(CIFAR),ImageNet,and Real-world Affective Faces Database(RAF-DB)datasets,indicate that our approach achieves 2×∼10×improvement in the reduction of both weight size and computation cost in comparison to state-of-the-art methods.A P-LUT accelerator prototype is implemented on the Xilinx KV260 Field Programmable Gate Array(FPGA)platform for accelerating convolution operations,with performance results showing that P-LUT reduces memory footprint by 1.45×,achieves more than 3×power efficiency and 2×resource efficiency,compared to the conventional bit-shifting scheme.

Customized Optimization for Vehicle Acoustic Statistical Energy Analysis

Statistical Energy Analysis(SEA) is one of the conventional tools for predicting vehicle high-frequency acoustic responses.This study proposes a new method that can provide customized optimization solutions to meet NVH targets based on the specific needs of different project teams during the initial project stages.This approach innovatively integrates dynamic optimization,Radial Basis Function(RBF),and Fuzzy Design Variables Genetic Algorithm(FDVGA) into the optimization process of Statistical Energy Analysis(SEA),and also takes vehicle sheet metal into account in the optimization of sound packages.In the implementation process,a correlation model is established through Python scripts to link material density with acoustic parameters,weight,and cost.By combining Optimus and VaOne software,an optimization design workflow is constructed and the optimization design process is successfully executed.Under various constraints related to acoustic performance,weight and cost,a globally optimal design is achieved.This technology has been effectively applied in the field of Battery Electric Vehicle(BEV).

Corrosion behavior of the friction-stir-welded joints of 2A14-T6 aluminum alloy

The corrosion behavior of friction-stir-welded 2A14-T6 aluminum alloy was investigated by immersion testing in immersion exfoliation corrosion（EXCO） solution. Electrochemical measurements（open circuit potential, potentiodynamic polarization curves, and electrochemical impedance spectroscopy）, scanning electron microscopy, and energy dispersive spectroscopy were employed for analyzing the corrosion mechanism. The results show that, compared to the base material, the corrosion resistance of the friction-stir welds is greatly improved, and the weld nugget has the highest corrosion resistance. The pitting susceptibility originates from the edge of Al-Cu-Fe-Mn-Si phase particles as the cathode compared to the matrix due to their high self-corrosion potential. No corrosion activity is observed around the θ phase（Al2Cu） after 2 h of immersion in EXCO solution.

Multi-objective comprehensive optimization of fuel consumption and emission for hybrid electric vehicles

Aiming to reduce fuel consumption and emissions of a dual-clutch hybrid electric vehicle during cold start, multiobjective optimization for fuel consumption and HC/CO emission from a TWC(three-way catalytic converter) outlet is presented in this paper. DP(dynamic programming) considering dual-state variables is proposed based on the Bellman optimality principle. Both the battery SOC(state of charge) and the temperature of TWC monolith are considered in the algorithm simultaneously. In this way the global optimal control strategy and the Pareto optimal solution of multi-objective function are derived. Simulation results show that the proposed method is able to promote the TWC light-off significantly by decreasing the engine's load and improving exhaust temperature from the outlet of the engine, in comparison with original DP considering the single battery SOC. Compared to the results achieved by rule-based control strategy, fuel economy and emission of TWC outlet for cold start are optimized comprehensively. Each indicator of Pareto solution set shows the significant improvement.

Study on Control of Pneumatic Selecting and Shifting Actuators Using Rapid Control Prototyping

A gearbox in-the-loop control platform using dSPACE real-time system is designed for the study on the control technology of pneumatic selecting and shifting actuators based on rapid control prototyping.The operational principle of such actuators was analyzed using dSPACE hardware and software,resulting in a better knowledge of the logical relationship among solenoid valves,gear positions of cylinders and system input/output.Based on these,a control model was developed under the Matlab/Simulink environment and rapidly improved to meet requirements through experiments.Relevant tests have shown that analysis efficiency on selecting and shifting actuators could be raised and development of control strategy facilitated.

Multi-Layer Reconstruction Errors Autoencoding and Density Estimate for Network Anomaly Detection

Anomaly detection is an important method for intrusion detection.In recent years,unsupervised methods have been widely researched because they do not require labeling.For example,a nonlinear autoencoder can use reconstruction errors to attain the discrimination threshold.This method is not effective when the model complexity is high or the data contains noise.The method for detecting the density of compressed features in a hidden layer can be used to reduce the influence of noise on the selection of the threshold because the density of abnormal data in hidden layers is smaller than normal data.However,compressed features may lose some of the high-dimensional distribution information of the original data.In this paper,we present an efficient anomaly detection framework for unsupervised anomaly detection,which includes network data capturing,processing,feature extraction,and anomaly detection.We employ a deep autoencoder to obtain compressed features and multi-layer reconstruction errors,and feeds them the same to the Gaussian mixture model to estimate the density.The proposed approach is trained and tested on multiple current intrusion detection datasets and real network scenes,and performance indicators,namely accuracy,recall,and F1-score,are better than other autoencoder models.

Wear and Corrosion Properties of Mo Surface-modified Layer in TiNi Alloy Prepared by Plasma Surface Alloying

In order to improve the wear resistance and restrain nickel release of TiNi alloys,the Mo modified layers on TiNi substrates were obtained using the double glow plasma surface alloying technique.Scanning electron microscopy（SEM）,glow discharge optical emission spectroscopy（GDOES） and X-ray diffraction（XRD） were employed to investigate the morphology,composition and structure.Microhardness test and scratch test were performed to analyze the microhardness and coating/substrate adhesion.Tribological and electrochemical behaviors of the Mo modified layers on TiNi were tested by the reciprocating wear instrument and electrochemical measurement system.The Ni concentrations in Hanks’ solution where surface electrochemical tests took place were measured by mass spectrometry.The surface-modified layer contained a Mo deposition layer and a Mo diffusion layer.The X-ray diffraction analysis revealed that the modified layers were composed of Mo,MoTi,Mo Ni,and Ti2Ni.The microhardnesses of the Mo modified layers treated at 900 ℃ and 950 ℃ were 832.8 HV and 762.4 HV,respectively,which was about 3 times the microhardness of the TiNi substrate.Scratch tests indicated that the modified layers possessed good adhesion with the substrate.Compared with as-received TiNi alloy,the modified alloys exhibited significant improvement of wear resistance against Si3N4 with low normal loads during the sliding tests.Mass spectrometry displayed that the Mo alloy layers had successfully inhibited the Ni release into the body.

Development and application of magnesium alloy parts for automotive OEMs:A review

China is currently vigorously implementing the“energy conservation and emission reduction”and“dual carbon”strategies.As the most resource-advantaged light metal material in China,Magnesium(Mg)alloy is progressively expanding its application in automobile,rail transportation,aerospace,medical,and electronic products.Chongqing University,Shanghai Jiaotong University,and Australian National University have conducted extensive research on the preparation,properties,and processes of Mg alloys.In the past 20 years,the proportion of Mg alloy in the automotive industry has gradually expanded,whereas currently the design and development of Mg alloy parts for automobiles has rarely been reported.Thus,the application models and typical parts cases of Mg alloy are summarized mainly from the four systems of the whole vehicle(body system,chassis system,powertrain system,interior,and exterior system).Subsequently,two actual original equipment manufacturers(OEM)cases are used to introduce the development logic of reliable die-cast Mg alloy,including forward design,formability analysis,process design analysis,structural redesign,manufacturing,and testing,aiming to share the methods,processes,and focus of attention of automotive OEMs for developing Mg alloy parts to enhance the confidence and motivation of applying Mg alloy in automotive field.Eventually,the multiple challenges faced by Mg alloy materials are sorted out and how to face these challenges are discussed.National policies and regulations,environmental protection and energy saving,and consumer demand will continue to promote the application of Mg.

Estimation of Vehicle Speed Based on Wheel Speeds from ASR System in Four-Wheel Drive Vehicles

Three major methods currently in the use of determining vehicle speed based on wheel speeds, the minimum wheel speed, minimum wheel speed corrected by slope method and the Kalman filter method, are analyzed, with merits and defects of each approach stated. Through simulations, the Kalman filter method based on minimum wheel speed shows improved accuracy, in addition to better adaptivity to vehicle reference speed. It also can be used to acceleration ship regulation （ASR） in part-time four-wheel drive vehicles.

Effect of Header Configuration on Refrigerant Distribution in Parallel Flow Microchannel Evaporators

The refrigerant flow distribution in the parallel flow microchannel evaporators is experimentally investigated to study the effect of header configuration.Six different configurations are tested in the same evaporator by installing insertion device and partition plate in the header to ensure the consistency of the other structure parameters.The results show that the uniformity of refrigerant flow distribution and the heat transfer rate are greatly improved by reducing the sectional area of header.The heat transfer rate can increase by 67.93%by reducing the sectional area of both inlet and outlet headers.The uniformity of refrigerant flow distribution and the heat transfer rate become worse after installing the partition plate in the insertion devices and changing the inner structure of the header further.

Prediction and Analysis of Vehicle Interior Road Noise Based on Mechanism and Data Series Modeling

Currently,the inexorable trend toward the electrification of automobiles has heightened the prominence of road noise within overall vehicle noise.Consequently,an in-depth investigation into automobile road noise holds substantial practical importance.Previous research endeavors have predominantly centered on the formulation of mechanism models and data-driven models.While mechanism models offer robust controllability,their application encounters challenges in intricate analyses of vehicle body acoustic-vibration coupling,and the effective utilization of accumulated data remains elusive.In contrast,data-driven models exhibit efficient modeling capabilities and can assimilate conceptual vehicle knowledge,but they impose stringent requirements on both data quality and quantity.In response to these considerations,this paper introduces an innovative approach for predicting vehicle road noise by integrating mechanism-driven and data-driven methodologies.Specifically,a series model is devised,amalgamating mechanism analysis with data-driven techniques to predict vehicle interior noise.The simulation results from dynamic models serve as inputs to the data-driven model,ultimately generating outputs through the utilization of the Long Short-Term Memory with Autoencoder(AE-LSTM)architecture.The study subsequently undertakes a comparative analysis between different dynamic models and data-driven models,thereby validating the efficacy of the proposed series vehicle road noise prediction model.This series model,encapsulating the rigid-flexible coupling dynamic model and AE-LSTM series model,not only demonstrates heightened computational efficiency but also attains superior prediction accuracy.

Sound Transmission Loss of Helmholtz Resonators with Elastic Bottom Plate

Helmholtz resonators are widely used to control low frequency noise propagating in pipes.In this paper,the elastic bottom plate of Helmholtz resonator is simplified as a single degree of freedom(SDOF)vibration system with acoustic excitation,and a one-dimensional lumped-parameter analytical model was developed to accurately characterize the structure-acoustic coupling and sound transmission loss(STL)of a Helmholtz resonator with an elastic bottom plate.The effect of dynamical parameters of elastic bottom plate on STL is analyzed by utilizing the model.A design criterion to circumvent the effect of wall elasticity of Helmholtz resonators is proposed,i.e.,the structural natural frequency of the wall should be greater than three times the resonant frequency of the resonator to avoid the adverse effects of wall elasticity.This study can provide guidance for the rapid and effective design of Helmholtz resonators.

Laminar Flame Instability of n-Hexane,n-Octane,and n-Decane in Spherical Expanding Flames

This paper focuses on the laminar flame instability of three high molecular weight n-alkanes,namely n-hexane,n-octane,and n-decane.The experiment was carried out in a constant volume combustion bomb to get the flame images.The critical radius under different conditions was extracted using the image processing program.Combined with the existing critical Peclet number theory,the dominant factors of flame instability under current conditions for three n-alkanes can be figured out.Moreover,the average cell size(equivalent cell radius,R_(cell))was extracted to provide quantitative analysis of the flame cellular structure,based on the method developed in this work.The theoretical R_(cell)were also calculated and compared with the experimental results to validate the proposed method.

Towards Human‑Vehicle Interaction: Driving Risk Analysis Under Different Driver Vigilance States and Driving Risk Detection Method

The driver's behavior plays a crucial role in transportation safety.It is widely acknowledged that driver vigilance is a major contributor to traffic accidents.However,the quantitative impact of driver vigilance on driving risk has yet to be fully explored.This study aims to investigate the relationship between driver vigilance and driving risk,using data recorded from 28 drivers who maintain a speed of 80 km/h on a monotonous highway for 2 hours.The k-means and linear fitting methods are used to analyze the driving risk distribution under different driver vigilance states.Additionally,this study proposes a research framework for analyzing driving risk and develops three classification models(KNN,SVM,and DNN)to recognize the driving risk status.The results show that the frequency of low-risk incidents is negatively correlated with the driver's vigilance level,whereas the frequency of moderate-risk and high-risk incidents is positively correlated with the driver's vigilance level.The DNN model performs the best,achieving an accuracy of 0.972,recall of 0.972,precision of 0.973,and f1-score of 0.972,compared to KNN and SVM.This research could serve as a valuable reference for the design of warning systems and intelligent vehicles.