The rapid development of autonomous driving technology has increased the demand for the authentic and diverse simulation test scenarios. However, traditional methods for constructing autonomous driving simulation scenarios heavily rely on manual editing, which is not only costly but also limited by the combination and complexity of scene elements, making it difficult to meet the comprehensive testing and validation needs of autonomous driving systems. To address this issue, this paper proposes a method for generating autonomous driving simulation test scenarios based on Large Language Models (LLMs). This approach utilizes a pre-trained LLM, enhanced through LoRA fine-tuning, and integrates a scenario language parser to produce a structured interpretive language, which is used to generate scenario files. The generated text is processed by a parser to convert it into usable scenario files, effectively addressing the issues of overly long texts and model hallucinations, while also achieving the specialization of a general model's capabilities.

Taking a domestic fuel commercial vehicle as an example, an energy consumption optimization prediction model suitable for commercial vehicles was constructed using the Internet of Vehicles big data platform and a neural network model. Firstly, the historical vehicle operation data was preprocessed to analyze the correlation between different vehicle operation characteristic data. Secondly, an adaptive weight attention mechanism was introduced based on Bi-directional Long Short-Term Memory (BiLSTM) and the characteristics of vehicle data. The Improved Whale Optimization Algorithm (IWOA) was used to optimize the network hyperparameters of the model, leading to the construction of the IWOA-BilSTM-Attention commercial vehicle energy consumption optimization prediction model. Finally, the prediction performance of multiple models under different driving conditions were compared and analyzed. The results show that under actual driving conditions, the root mean square error and the mean absolute error of the optimized model are reduced by approximately 26.73% and 20.0%, respectively, compared with the original model. This verifies the feasibility of the optimized model for predicting the energy consumption of commercial vehicles.

To address the heat dissipation issue of hub motor in a specialized vehicle, a serpentine flow channel was chosen as the cooling structure. CFD simulations were performed to analyze the temperature rise, temperature distribution, and radial temperature variation of the hub motor. Subsequently, the effects of the number and width of flow channels on the motor's temperature and flow fields were investigated. The appropriate number of flow channels was determined and an initial selection of channel width was decided. Furthermore, taking the flow channel height and width as optimization variables, a multi-objective optimization was conducted, with the maximum motor temperature and flow channel pressure drop as the optimization targets. The results show that after optimization, the highest motor temperature increased by 0.22℃, while the flow channel pressure drop decreased by 904.19 Pa, effectively reducing energy loss and improving the heat dissipation efficiency of the cooling system.

The rapid rise of new energy vehicles under the“dual carbon” goals has shifted the focus of vehicle lightweighting from traditional structural and process innovations to the substitution and optimization of materials. Aluminum alloy materials highly match the requirements for materials used in automotive lightweighting, and are currently the most preferred materials. This paper reviews and discusses the research and development, application and new directions of aluminum alloy materials in automobile lightweighting.Firstly, the paper introduces the main brands and the application status of die casting aluminum, which accounts for more than 70% of the aluminum used in automobile drive systems, chassis systems and body structure parts. The focus is placed on analyzing the research status of integrated die casting technology and its necessary non-heat treatment aluminum alloy. The paper summarizes the use of wrought aluminum alloy in the automotive field, including the stamped parts, profiled components and forgings. It also discusses the research status of traditional automotive forged aluminum and high-strength aluminum, guided by the development trend towards high-strength and ductile aluminum alloys. Finally, the existing bottlenecks of aluminum alloy materials in the automotive field are analyzed and prospected.

A multi-objective optimized automatic design process is developed based on the airflow velocity required for cooling performance in the passenger cabin. This process considers the airflow performance on the driver's side and the passenger's side, focusing on the positioning of the grille vent blades. Based on CFD simulations and a multi-disciplinary optimization design platform, the Latin hypercube sampling method is used to generate sample points and construct the DOE matrix. A neural network-based proxy model is then built to predict the blow-face airflow velocity performance parameters. The NSGA-Ⅲ algorithm is used to obtain the Pareto frontier diagram for the multi-objective optimization problem. The optimized grille blade position increases the airflow speed by 109.1% on the driver's side and by 137.5% on the front passenger's side. The reliability of the optimization results is verified through unsteady CFD simulations and cooling performance tests before and after the optimization.

The slide movement pattern of traditional mechanical presses is relatively fixed, and quasi-static stamping simulations usually ignore the effect of strain rate. In contrast, servo presses feature a flexible slide stroke and adjustable stamping speed; thus, a material constitutive model that includes the strain rate effect is necessary to achieve accurate servo stamping simulations. To address the limitation of obtaining the stress-strain curves at only a finite number of strain rates through tensile tests, this paper analyzed and evaluated the strain rate-sensitive models, including the power law model, linear power law model, Johnson-Cook model and Cowper-Symonds model, using DDQ steel test data for model fitting as an example. Curve fitting is performed for segmented strain range and grouped strain rates, the parameter identification methods in each model are constructed and the applicability of each model in finite element software is discussed. The results provide guidance for selecting an appropriate constitutive model for servo stamping simulations.

Lithium-ion power batteries are currently the most widely used energy storage devices in electric vehicles. Rapid and accurate battery fault diagnosis is crucial for ensuring safe vehicle operation. This paper proposes a method for diagnosing self-discharge faults in power batteries based on adaptive voltage thresholds for individual cells and Particle Swarm Optimization-Support Vector Machine (PSO-SVM). This study focuses on the voltage signals of power batteries, and combines the boxplot method with expert review to label self-discharge fault samples. A sliding window method is used to extract 16 features from both the time and frequency domains. To further reduce the dimensionality of voltage features, principal component analysis is applied, selecting the top five principal components with a 95% cumulative variance contribution as inputs for the PSO-SVM model. This method aims to improve the accuracy of self-discharge fault detection in batteries. The results show that the proposed method achieves high detection accuracy, strong reliability, and promising potential for practical applications in electric vehicles. Additionally, it provides theoretical support for enhancing the safety performance of electric vehicles.

The paper examines the door sealing cavities near the B-pillar, C-pillar and the top of the tailgate of a SUV. The geometric features of these cavities are extracted and represented using three equivalent regular cavity models. The experimental studies on the sound phenomena and formation mechanism of these cavities were carried out in a small acoustic wind tunnel. The results indicate that the three cavities exhibit two different sound phenomena, i.e. strong resonance and weak resonance. Further numerical simulations are performed to analyze the acoustic characteristics of the cavities. The results show that self-excited oscillations are difficult to form in the small opening cavities, such as those near the B-pillar and C-pillar. Instead, pulsating excitation at the opening induces weak resonance in the cavity mode. In contrast, the cavities with wider openings near the tailgate can form self-excited oscillations, which resonate with cavity acoustic modes or Helmholtz resonance modes to produce strong resonance with whistling noise. The differences in phenomena and mechanisms between automobile door sealed cavities and large cavities are revealed. A method for determining the frequency of self-excited oscillations is proposed by characterizing the vortex motion in the Q-factor cloud diagram. Additionally, the contribution of self-excited oscillations to the peak values in the cavity sound pressure spectrum is clearly explained, while effectively defining their frequency.

To extend the range of rear-drive electric buses, a brake force distribution control strategy based on intention recognition is proposed. Firstly, 400 sets of real-vehicle braking data were collected and analyzed. The brake pedal opening and its rate of change were used to calculate the braking force applied to the front and rear axles. Considering the battery constraints and the regenerative braking constraints of the rear axle motor, the braking control strategy is formulated and validated using Simulink-Trucksim joint simulation. The results show that the composite braking control strategy based on intention recognition achieves an accuracy of 95.7% in detecting the driver's braking intention. Under typical urban driving cycle in China, the fuzzy neural network-based energy recovery strategy, the fuzzy control recovery strategy, and the conventional recovery strategy increased the final state of charge (SOC) by 2.69%, 2.09% and 1.83%, respectively, compared with the non-recovery control strategy.

This paper discusses the challenges of applying virtual reality (VR) in visual simulation and real-time rendering for automotive research and development. Focusing on the application scenarios and practical needs of automotive virtual development, a high-performance VR computing system based on real-time ray tracing and parallel rendering has been established for the first time among domestic automotive companies. By using the Bounding Volume Hierarchy (BVH) acceleration structure while optimizing algorithm parameters and network configuration, high-precision real-time visualization of large-scale vehicle models has been achieved. The offline rendering process, which previously took several hours, has been optimized to real-time rendering in just 1～2 seconds. The system effectively simulates ambient light refraction and reflection on parts, light uniformity and leakage, and physical occlusion between parts. Compared with traditional rasterization rendering, this approach has made a significant breakthrough, achieving over 90% physical realism. Through full-process implementation in several vehicle projects, it has replaced many physical models and reduced sample costs in the R&D process.