Based on the original welding test, the numerical simulation of stud impact welding with arbitrary Lagrangian-Eulerian (ALE) method was carried out. The numerical simulation results were highly similar to the actual welding test results, so the simulation results could truly reflect the actual welding test. Combined with the simulation datas, the distribution of shear stress, compressive stress and corresponding strain at different positions of the sample section were analyzed, which proves that the stress wave was an important mechanism for the formation of wavy interface.

Based on the high-frequency micro-vibration platform, the mechanical vibration-assisted cold metal transfer welding (CMT) test of 6082 aluminum alloy was carried out to explore the influence of high-frequency micro-vibration CMT process on the microstructure and properties of joints. The results show that after applying vibration, the weld reinforcement increased slightly, the weld width decreased, and the penetration depth increased. When the vibration frequency was 1 119 Hz, the weld reinforcement and penetration depth increased by 4.5% and 23% respectively. The vibration narrowed the width of the heat-affected zone, and the grain size near the fusion line had reduced from the original 20 µm to 15 µm. After vibration was applied, the bubble escape speed was accelerated, the growth trend of columnar crystals was suppressed, the crystal grains were refined, the hardness of the joint softening zone was improved, and the hardness distribution of the welded joint was more uniform. When the vibration frequency was 1 119 Hz, the Vickers hardness of the weld softening zone reached the highest value of 77. Vibration can effectively reduce the residual stress of the welded joint and the heat-affected zone. When the vibration frequency was 1119 Hz, the average value of the maximum residual stress of the welded joint and the heat-affected zone had reduced by 9.6% and 6.3% respectively, and the mechanical properties of the welded joint were improved.

Finite element simulation method was used to simulate the electric field of 400 kA aluminum electrolytic cell under the traditional cathode steel bar structure and the special-shaped cathode steel bar structure, which mainly included the cathode voltage drop and the level current of aluminum liquid. An improved structure of cathode steel bar was proposed by analyzing the abnormal structure of the cathode steel bar. The cathode steel bar has the characteristics of bulge and extension. The result shows that the improved cathode steel bar can further reduce the voltage and level current of aluminum electrolytic cell, thus improving the current efficiency of electrolytic cell, reducing the energy consumption of a single cell by about 33.6 kW•h, and improving the stability of electrolytic aluminum.

The stability problem was investigated for a class of time-delayed neural networks driven by G−Brownian motion (G−DNNs). The random noise does not always obeys a normal distribution, in order to better describe the actual situation, G−Brownian motion was used to describe the noise, and the influence of discrete-time observations of G-Brownian motion noise on the stability of time-delayed neural networks was analyzed. For exponentially stable time-delayed neural networks, the random noise driven by G-Brownian motion was introduced, the upper bound of random noise intensity was given by using G−random analysis theory, Gronwall inequality, and Borel-Cantelli lemma, and when the noise intensity was less than the upper bound, the stability speed of stochastic delayed recurrent neural networks was faster than that of the original. The stability of stochastic delayed recurrent neural networks was further analyzed in the case of discrete-time noise. By using the G−Itô formula, scaling technique, and some basic inequalities, the upper bound was obtained for the discrete-time step size of the noise and the stability of the stochastic delayed recurrent neural networks was further accelerated. The validity of theoretical results was verified by an example.

In order to improve the measurement accuracy of electronic sphygmogram, a measure scheme using self-adaptive segmental recognition of Korotkoff sound was proposed based on the principle of Korotkoff sound method. Firstly, the dual ADC was used to collect sound and pressure signals, and the static pressure equation was fitted in real time. Then, the abrupt point in the pressure signal was identified and the signal of Korotkoff sound was segmtioned. Then, the threshold value is calculated adaptively based on the normal distribution 3σ principle, and the feature points of the middle segment Korotkoff sound were extracted. Finally, heart rate was used as the search step to identify systolic and diastolic blood pressure points. The experimental results show that the standard deviation of systolic blood pressure based on this method is 2.18 mmHg (1 mmHg=133.322 Pa), the standard deviation of systolic blood pressure is 2.74 mmHg, the standard deviation of diastolic blood pressure is −0.93 mmHg, and the standard deviation of diastolic blood pressure is 2.24 mmHg, which prove that the scheme has higher accuracy and stability, and provide a direction for the optimization and improvement of electronic measurement of Korotkoff sound.

In order to improve the stability and robustness of self-driving vehicle, a trajectory tracking control method was proposed based on the combination of self-adaptive proportional integral derivative (PID) of radial basis function neural network (RBFNN−PID) and model predictive control (MPC). A simulation model of intelligent vehicle longitudinal speed control and lateral control was established based on self-adaptive RBFNN−PID algorithm, MPC algorithm and vehicle dynamics model. Based on that, the lateral MPC and LQR−PID control algorithm were used as benchmarks to demonstrate the superiority of the presented control method in trajectory tracking. The simulation results show that the presented control method has higher accuracy than the comparing group. Finally, the hardware-in-the-loop verification of the proposed control method is carried out. The results show that the proposed trajectory tracking control algorithm is effective and advanced in trajectory tracking accuracy and stability.

As a component of traffic system, the level intersection is the node most likely to cause traffic congestion and traffic accidents. Based on the causes of congestion and the current traffic organization survey of the level intersection of Fushou Road and Xinxing Road in Liaoyuan City, were the road canalization and the optimization of signal timing phase for the intersection of Fushou Road and Xinxing Road were conducted. Indicators such as capacity, saturation, service level, queue length and vehicle delay were used for evaluation and analysis, and VISSIM simulation was used to verify the scheme. The results show that the intersection capacity of the optimized scheme was improved by 43.78%. The saturation level decreased by 45.04%. The level of service was improved to level 2. Vehicle delays decreased by 33.01%. The maximum queue length decreased by 33.63%. The traffic condition at the intersection of Fushou Road and Xinxing Road in Liaoyuan City was improved, which can demonstrate the effectiveness of the scheme.

Aiming at the problems of slow speed, low accuracy and poor robustness in recognition, detection and segmentation of stacked automobile parts, a fast detection and instance segmentation method based on improved Mask R−CNN algorithm was proposed. Firstly, the feature extraction network of Mask R-CNN was optimized, and ResNet + Feature Pyramid Networks (FPN) was replaced by MobileNets + FPN as the backbone network, which effectively reduced network parameters, compressed model volume and improved model detection speed. Then,Spatial Transformer Networks (STN) module was added after the ROI Align structure of Mask R-CNN to ensure the detection accuracy of the model. The experimental results show that the size of the model is compressed and the detection speed is doubled. The mean Average Precision (mAP) of the model is also improved. The detection of untrained new samples shows that the model is better than Mask R−CNN in speed, lighter and more accurate, and can quickly and accurately detect and segment stacked automobile parts, which verifies the practical feasibility of the improved model.

Under the tendency of automotive lightweight, the Body in White (BIW) light weighting job is still significantly meaningful. From aspects of the domestic and abroad research status, structure design, structural analysis, mass comparison analysis, the lightweight of the local structure on car upper body was studied. By incorporating ultra-high-strength steel roll forming 3D tube into Honda Civic 2016 A-pillar and roof rail structure, the original hot stamping reinforcements were replaced. Based on the roof crush test referring to GB 26134—2010, the roof crush performance of the optimized structure was need to be equivalent to the original structure. By comparing the quality, the lightweight effect of the new design with 3D tube inside of the body was obtained. The results prove that new structure with 1 mm gauge and 1 700 MPa Martensitic steel 3D tube shows equal bending stiffness and strength with original structure. A significant mass reduction of 3.23 kg per car can be achieved.

Streetcars are easily affected by road environment. The intersection, as the bottleneck of road capacity, restricts the traffic efficiency of the whole road network. An optimization problem model was established, the improved particle swarm algorithm was analyzed, and the particle swarm algorithm based on catastrophic adaptation was used to analyze signal timing of  streetcar intersection. Taking the actual single point intersection of streetcar in Shanghai Songjiang as research object, the optimal signal phase distribution plan of the intersection was obtained through simulation, which can reduce the impact of streetcars on road congestion.

Due to its decentralization, openness, autonomy, information tampering and anonymity, blockchain technology has broad application prospects. Focusing on the application of blockchain in the automotive industry, the application status of blockchain technology and typical framework model were summarized in four aspects: automobile manufacturing, car sharing, automobile insurance, and new energy vehicles. The challenges and future research directions of blockchain application in the automotive industry were analyzed. The result shows that its applications in secure data storage and data control, smart contract and DAPP development technology, network optimization, benefit distribution, policies and regulations still need further breakthroughs .

By theoretical derivation and numerical analysis, the modal parameters of two-dimensional absolute nodal coordinate beam elements under different elastic force modeling methods were studied. The continuum mechanics, enhanced continuum mechanics and strain splitting method were introduced, and the limitations of the application of the strain splitting method were deduced, and the characteristics of the three methods were explained from the theoretical point of view. Based on the generalized characteristic equations, the four natural modes of the absolute nodal coordinate beam element under different elastic force modeling methods were obtained. Taking the simply supported beam structure as an example, the effects of the absolute nodal coordinate beam element elastic force modeling method on the four natural modes were analyzed. Under different elastic force modeling methods, the four natural frequencies of the transverse low-order elements were higher than those of the transverse high-order elements, and the performance was more "rigid". For the continuum mechanics method, there is an error of 20%~30% between the low-order shear natural frequency of each element and the analytical solution, while the strain decomposition method and the enhanced continuum mechanics method can control the error of the shear natural frequency within 4%, and improve the element convergence accuracy.

An assessment model based on the administrative punishment cases of the People’s Bank of China was built to measure the degree of money laundering and conduct an empirical analysis. Based on 1717 administrative punishment cases by the People’s Bank of China, five first-level risk level indexes were constructed. The AHP and entropy method were used to assign the weights of the risk level indexes, and an evaluation model was built based on above methods. The random forest model was used to test the validity of index weights and the accuracy of the model. The results showed that the F−score of the testing set was up to 94%. The research results can provide preferences for finding typical cases and prominent problems from the large number of anti-money laundering administrative punishment cases, and then promote the construction of China's anti-money laundering system.

Count datas with excess zeros and ones arise frequently in the field of public health. In order to fit the kind of data, a zero-and-one-inflated negative binomial (ZOINB) distribution and its regression model were adopted for analysis. Based on data augmentation strategy and Pólya−Gamma latent variables Bayesian inference was used to estimate the parameters of ZOINB regression model. Finally, one corona virus disease 2019 (COVID−19) death data-set from Hubei Province in China was analyzed. The result illustrates that ZOINB regression model can achieve better fitting effect.

In China, the number of enterprises with bad credit is much smaller than that of enterprises with good credit. The extreme imbalance of sample categories results in the traditional credit evaluation model unable to fully learn the characteristics of bad credit enterprises during training. In order to improve the accuracy of extreme gradient boosting (XGBoost) in unbalanced classification problems such as enterprise credit evaluation, an enterprise credit evaluation model based on cost sensitive XGBoost was proposed. In the process of XGBoost algorithm fitting, the cost sensitive loss function was added to force the model to pay more attention to the characteristics of minority classes, and the bayesian optimization was introduced to adjust the hyperparameters of the model. Taking the datas of small and medium-sized enterprises in China's A-share market from 2016 to 2020 as the sample, the experimental results show that the enterprise credit evaluation model based on cost sensitive XGBoost can improve the identification accuracy of bad credit enterprises while ensuring the overall identification accuracy.

Taking the Yangtze River Delta urban agglomeration as an example, the characteristics of the area's surface, climate, and social and economic development were considered, and an evaluation index system for the ecological environment vulnerability of the Yangtze River Delta urban agglomeration from human factors and natural factors was built. At the same time, based on geographic information system (GIS) technology, principal component analysis, entropy weight method, spatial autocorrelation analysis and other methods were used to comprehensively measure and analyze the spatial and temporal characteristics of ecological environmental vulnerability level in the Yangtze River Delta urban agglomeration in 2010, 2015 and 2018, and identify its driving factors. The results show that: 1) in terms of time series distribution, the ecological environment fragility level of the Yangtze River Delta urban agglomeration has increased, and the transition from moderate fragility to severe fragility; 2) in terms of spatial distribution, the vulnerability of ecological environment has spatial autocorrelation, and it is a significant positive correlation; 3) from 2010 to 2018, the degree of land use, GDP per capita, industrial sulfur dioxide emissions, and green coverage of built-up areas are the core driving forces for the fragility of the ecological environment of the Yangtze River Delta urban agglomeration.