Browse Topic: Neural networks

Items (1,328)

Prediction of hazardous gaseous emissions from a gasoline engine during cold starts using machine learning methods

2025-01-0321To be published on 07/02/2025

Internal combustion engines generate higher exhaust emissions of hazardous gases during the initial minutes after engine start. Experimental data from a state-of-the-art turbo-charged 3-cylinder, 999 cc gasoline engine are used to predict cold start emissions using two Machine Learning (ML) models: a Multilayer Perceptron (MLP) which is a fully connected neural network and an Encoder-Decoder Recurrent Neural Network (ED-RNN). Engine parameters and various temperatures are used as input for the models and NOx (Nitrogen Oxides), CO (Carbon monoxide) and unburned hydrocarbon (UHC) emissions are predicted. The dataset includes time series recordings from the Worldwide harmonized Light-duty vehicles Test Cycle (WLTC) and four Real Diving Emissions (RDE) cycles at ambient and initial engine temperatures ranging from -20 °C to +23 °C. In total, 21 cases are considered, consisting of eight different ambient temperatures and five distinct driving cycles. Each case consists of a sequence of 2500

Mangipudi, Manoj, Denev, Jordan A., Bockhorn, Henning, Trimis, Dimosthenis, Koch, Thomas, Debus, Charlotte, Götz, Markus, Zirwes, Thorsten, Hagen, Fabian P., Tofighian, Hesam, Wagner, Uwe, Braun, Samuel, Lanzer, Theodor, Knapp, Sebastian M.

Design Space Exploration of a Continuous Rubber Track System via Surrogate Modeling

02-18-04-0023

05/26/2025

Continuous rubber track systems for heavy applications are typically designed using multiple iterations of full-scale physical prototypes. This costly and time-consuming approach limits the possibility of exploring the design space and understanding how the design space of that kind of system is governed. A multibody dynamic simulation has recently been developed, but its complexity (due to the number of model’s inputs) makes it difficult to understand and too expensive to be used with multi-objective optimization algorithms (approximately 3 h on a desktop computer). This article aims to propose a first design space exploration of continuous rubber track systems via multi-objective optimization methods. Using an existing expensive multibody dynamic model as original function, surrogate models (artificial neural networks) have been trained to predict the simulation responses. These artificial neural networks are then used to explore the design space efficiently by using optimization

Faivre, Antoine, Rancourt, David, Plante, Jean-Sébastien

Hybrid Approaches to Software Reliability: Evaluating and Enhancing Prediction Models

2025-01-5024

05/06/2025

Software reliability prediction involves predicting future failure rates or expected number of failures that can happen in the operational timeline of the software. The time-domain approach of software reliability modeling has received great emphasis and there exists numerous software reliability models that aim to capture the underlying failure process by using the relationship between time and software failures. These models work well for one-step prediction of time between failures or failure count per unit time. But for forecasting the expected number of failures, no single model will be able to perform the best on all datasets. For making accurate predictions, two hybrid approaches have been developed—minimization and neural network—to give importance to only those models that are able to model the failure process with good accuracy and then combine the predictions of them to get good results in forecasting failures across all datasets. These models once trained on the dataset are

Mahdev, Akash Ravishankar, Lal, Vinayak, Muralimohan, Pramod, Reddy, Hemanjaneya, Mathur, Rachit

A Comparative Analysis of Fault Diagnosis by Vibration Signals for Critical Gear Components in Electric Vehicle Motor Testing Machines Using Machine Learning Algorithms

2025-01-0040

05/05/2025

Electric vehicles (EVs) are shaping the future of mobility, with drive motors serving as a cornerstone of their efficiency and performance. Motor testing machines are essential for verifying the functionality of EV motors; however, flaws in testing equipment, such as gear-related issues, frequently cause operational challenges. This study focuses on improving motor testing processes by leveraging machine learning and vibration signal analysis for early detection of gear faults. Through statistical feature extraction and the application of classifiers like Wide Naive Bayes and Coarse Tree, the collected vibration signals were categorized as normal or faulty under both loaded (0.275 kW) and no-load conditions. A performance comparison demonstrated the superior accuracy of the wide neural networks algorithm, achieving 95.3%. This methodology provides an intelligent, preventive maintenance solution, significantly enhancing the reliability of motor testing benches.

S, Ravikumar, Sharik, N, Syed, Shaul, V, Muralidharan, D, Pradeep Kumar

Mode Shape Identification Using Graph Neural Networks for Vehicle Structure Design

2025-01-0131

05/05/2025

This paper introduces a novel, automated approach for identifying and classifying full vehicle mode shapes using Graph Neural Networks (GNNs), a deep learning model for graph-structured data. Mode shape identification and naming refers to classifying deformation patterns in structures vibrating at natural frequencies with systematic naming based on the movement or deformation type. Many times, these mode shapes are named based on the type of movement or deformation involved. The systematic naming of mode shapes and their frequencies is essential for understanding structural dynamics and “Modal Alignment” or “Modal Separation” charts used in Noise, Vibration and Harshness (NVH) analysis. Current methods are manual, time-consuming, and rely on expert judgment. The integration of GNNs into mode shape classification represents a significant advancement in vehicle modal identification and structure design. Results demonstrate that GNNs offer superior accuracy and efficiency compared to

Tohmuang, Sitthichart, Swayze, James L., Fard, Mohammad, Fayek, Haytham, Marzocca, Piergiovanni, Bhide, Sanjay, Huber, John

Application of AI/ML in Hydrobush Tuning to Enhance Overall Value Proposition

2025-01-0132

05/05/2025

In the highly competitive automotive industry, optimizing vehicle components for superior performance and customer satisfaction is paramount. Hydrobushes play an integral role within vehicle suspension systems by absorbing vibrations and improving ride comfort. However, the traditional methods for tuning these components are time-consuming and heavily reliant on extensive empirical testing. This paper explores the advancing field of artificial intelligence (AI) and machine learning (ML) in the hydrobush tuning process, utilizing algorithms such as random forest, artificial neural networks, and logistic regression to efficiently analyze large datasets, uncover patterns, and predict optimal configurations. The study focuses on comparing these three AI/ML-based approaches to assess their effectiveness in improving the tuning process. A case study is presented, evaluating their performance and validating the most effective method through physical application, highlighting the potential

Hazra, Sandip, Khan, Arkadip Amitava

A Comprehensive Review of Vehicle and Road Condition Estimation Techniques

10-09-02-0014

04/30/2025

This article reviews the key physical parameters that need to be estimated and identified during vehicle operation, focusing on two key areas: vehicle state estimation and road condition identification. In the vehicle state estimation section, parameters such as longitudinal vehicle speed, sideslip angle, and roll angle are discussed, which are critical for accurately monitoring road conditions and implementing advanced vehicle control systems. On the other hand, the road condition identification section focuses on methods for estimating the tire–road friction coefficient (TRFC), road roughness, and road gradient. The article first reviews a variety of methods for estimating TRFC, ranging from direct sensor measurements to complex models based on vehicle dynamics. Regarding road roughness estimation, the article analyzes traditional methods and emerging data-driven approaches, focusing on their impact on vehicle performance and passenger comfort. In the section on road gradient

Chen, Zixuan, Duan, Yupeng, Wu, Jinglai, Zhang, Yunqing

Road Surface Adaptive Stationary Steering Coordinated Control of Four-Wheel Hub Motor Vehicles

2025-01-5031

04/29/2025

The wheel hub motor–driven electric vehicle, characterized by its independently controllable wheels, exhibits high torque output at low speeds and superior dynamic response performance, enabling in-place steering capabilities. This study focuses on the control mechanism and dynamic model of the wheel hub motor vehicle’s in-place steering. By employing differential torque control, it generates the yaw moment needed to overcome steering resistance and produce yaw motion around the steering center. First, the dynamic model for in-place steering is established, exploring the various stages of tire motion and the steering process, including the start-up, elastic deformation, lateral slip, and steady-state yaw. In terms of control strategy, an adaptive in-place steering control method is designed, utilizing a BP neural network combined with a PID control algorithm to track the desired yaw rate. Additionally, a control strategy based on tire/road adhesion ellipse theory is developed to

Huang, Bin, Cui, Kangyu, Zhang, Zeyang, Ma, Minrui

Study on the Optimal Pre-Chamber Geometry for Active Pre-Chamber Gas Engines

2024-32-0023

04/18/2025

In a pre-chamber engine, fuel in the main-chamber is ignited and combusted by the combustion gas injected from the pre-chamber. Therefore, further fuel dilution is possible and thermal efficiency can be also improved. However, adding a pre-chamber to an engine increases the number of design parameters which have a significant impact on the main combustion and the exhaust gas. Then, in this study, the optimum geometry of the pre-chamber in an active pre-chamber gas engine was investigated. The considered parameters were the volume of pre-chamber, the diameter of a nozzle hole, and the number of nozzle holes. 18 types of pre-chambers with different geometries were prepared. Using these pre-chambers, engine experiments under steady conditions were conducted while changing the conditions such as engine speeds, mean indicated pressure and air excess ratio. Based on the experimental data, neural network models were constructed that predict thermal efficiency, NOx and CO emissions from the

Yasuda, Kotaro, Yamasaki, Yudai, Sako, Takahiro, Takashima, Yoshitane, Suzuki, Kenta

Development of a Robust Framework for Fault Detection and Quantification in Automotive Drum Brakes Using Vibration Measurements and ANN

10-09-03-0024

04/15/2025

Passenger safety is of utmost importance in the automotive industry. Hence, the health of the components, especially the brake system, should be effectively monitored. On account of the significance of artificial intelligence in recent times, any brake fault resulting during operation can be accurately detected using a combination of advanced measurement techniques and machine learning algorithms. The current study focuses on developing and evaluating a robust framework to quantify and classify the faults of a general automotive drum brake. For this purpose, a new experiment for a drum brake, which can be operated under a controlled environment with known levels of faults, is developed. The experiment is instrumented to measure the fundamental dynamic signals (such as brake torque, the angular velocity of the brake drum, and brake shoe accelerations) during a braking event. The response signals from several experiments with various faults and operating conditions serve as the input

Yella, Akash, Bharinikala, Yuva Venkat Ajay, Sundar, Sriram

Elongation Prediction of Die-Cast Aluminum Alloy Based on 3D Convolutional Neural Network Model

05-18-04-0032

04/09/2025

This study aims to predict the impact of porosities on the variability of elongation in the casting Al-10Si-0.3Mg alloy using machine learning methods. Based on the dataset provided by finite element method (FEM) modeling, two machine learning algorithms including artificial neural network (ANN) and 3D convolutional neural network (3D CNN) were trained and compared to determine the optimal model. The results showed that the mean squared error (MSE) and determination coefficient (R2) of 3D CNN on the validation set were 0.01258/0.80, while those of ANN model were 0.28951/0.46. After obtaining the optimal prediction model, 3D CNN model was used to predict the elongation of experimental specimens. The elongation values obtained by experiments and FEM simulation were compared with that of 3D CNN model. The results showed that for samples with elongation smaller than 9.5%, both the prediction accuracy and efficiency of 3D CNN model surpassed those of FEM simulation.

Zhang, Jin-sheng, Zheng, Zhen, Zhao, Xing-zhi, Gong, Fu-jian, Huang, Guang-sheng, Xu, Xiao-min, Wang, Zhi-bai, Yang, Yutong

Research on Dangerous Driving Behavior Recognition Algorithm Based on LSTM

2025-01-8204

04/01/2025

To address the issue of high accident rates in road traffic due to dangerous driving behaviors, this paper proposes a recognition algorithm for dangerous driving behaviors based on Long Short-Term Memory (LSTM) networks. Compared with traditional methods, this algorithm innovatively integrates high-frequency trajectory data, historical accident data, weather data, and features of the road network to accurately extract key temporal features that influence driving behavior. By modeling the behavioral data of high-accident-prone road sections, a comprehensive risk factor is consistent with historical accident-related driving conditions, and assess risks of current driving state. The study indicates that the model, in the conditions of movement track, weather, road network and conditions with other features, can accurately predict the consistent driving states in current and historical with accidents, to achieve an accuracy rate of 85% and F1 score of 0.82. It means the model can

Huang, Yinuo, Zhang, Miaomiao, Xue, Ming, Jin, Xin

A Suspension Model Based on Semi-Recursive Method and Data-Driven Compliance Characteristics

2025-01-8282

04/01/2025

Many methods have been proposed to accurately compute a vehicle’s dynamic response in real-time. The semi-recursive method, which models using relative coordinates rather than dependent coordinates, has been proven to be real-time capable and sufficiently accurate for kinematics. However, not only kinematics but also the compliance characteristics of the suspension significantly impact a vehicle’s dynamic response. These compliance characteristics are mainly caused by bushings, which are installed at joints to reduce vibration and wear. As a result, using relative or joint coordinates fails to account for the effects of bushings, leading to a lack of compliance characteristics in suspension and vehicle models developed with the semi-recursive method. In this research, we propose a data-driven approach to model the compliance characteristics of a double wishbone suspension using the semi-recursive method. First, we create a kinematic double wishbone suspension model using both the semi

Zhang, Hanwen, Duan, Yupeng, Zhang, Yunqing, Wu, Jinglai

An Integrated Data-Driven and Physics-Based Approach for Dynamic Operation Simulation of Electric Vehicles

2025-01-8604

04/01/2025

To address the challenges of complex operational simulation for Electric Vehicles (EVs) caused by spatial-temporal variations and driver behavior heterogeneity, this study introduces a dynamic operation simulation model that integrates both data-driven and physics-based principles, referred to as the Electric Vehicle-Dynamic Operation Simulation (EV-DOS) model. The physics-based component encompasses critical aspects such as the powertrain energy transfer module, heat transfer module, charge/discharge module, and battery state estimation module. The data-driven component derives key features and labels from second-by-second real-world vehicle driving status data and incorporates a Long Short-Term Memory (LSTM) network to develop a State-of-Health (SOH) prediction model for the EV power pack. This model framework combines the interpretability of physical modeling with the rapid simulation capabilities of data-driven techniques under dynamic operating conditions. Finally, this study

Jing, Hao, HU, Jianyao, Ouyang, Jianheng, Ou, Shiqi(Shawn)

Active and Passive Security Generalization Algorithm and Typical Case Analysis

2025-01-8744

04/01/2025

In order to effectively predict the vehicle safety performance and reduce the cost of enterprise safety tests, a generalized simulation model for active and passive vehicle safety was proposed. The frontal driver-side collision model under the intervention of the Autonomous Emergency Braking (AEB) was created by using the MADYMO software. The collision acceleration obtained from the sled test was taken as the original input of the model to conduct simulation for the working conditions under different sitting postures of the human body. The injury values of various parts of the Hybrid III 50th dummy were read. Based on the correlation between the two, an active and passive simulation model was established through the Back Propagation (BP) neural network. The input of the model was the inclination angle centered on the dummy's waist, and the output was the acceleration of the dummy's head. The results showed that the comprehensive prediction accuracy rate exceeded 80%. Therefore, the

Ge, Wang, fengyao, LV

Real-Time PMV Thermal Comfort Index Observer Based on Artificial Neural Networks for Infrared Heating Panel Control

2025-01-8139

04/01/2025

Improving electric vehicles’ range can be achieved by integrating infrared heating panels (IRPs) into the existing Heating Ventilation and Air-Conditioning system to reduce battery energy consumption while maintaining thermal comfort. Localized comfort control enabled by IRPs is facilitated by thermal comfort index feedback to the control strategy, such as the well-known Predicted Mean Vote (PMV). PMV is obtained by solving nonlinear equations iteratively, which is computationally expensive for vehicle control units and may not be feasible for real-time control. This paper presents the design of real-time capable thermal comfort observer based on feedforward artificial neural network (ANN), utilized for estimating the local PMV extended with IRP radiative heating effects. The vehicle under consideration is equipped with 12 heating panels (zones) organized into six controller clusters that rely on the average PMV feedback from its respective zone provided by a dedicated ANN. Each of six

Cvok, Ivan, Yerramilli-Rao, Isha, Miklauzic, Filip

Role of AI in an Autonomous Vehicle Perception Systems: Use of Convolutional Neural Network Approach to Design the Vehicle Perception System

2025-01-8012

04/01/2025

Artificial Intelligence has gained lot of traction and importance in the 21st century with use cases ranging from speech recognition, learning, planning, problem solving to search engines etc. Artificial Intelligence also has played a key role in the development of autonomous vehicles and robots ranging from perception, localization, decision to controls. Within the big AI umbrella there is machine learning which is all about using your computer to "learn" how to deal with problems without “programming". Deep learning is a branch of machine learning based on a set of algorithms that learn to represent the data directly from the input such as an image, text, Sound, etc. Within deep learning there are Convolutional Neural Networks and Recurrent Neural Networks (CNN/RNN). The study here used convolutional neural network approach to perform image/object recognition. Given that the objective of the autonomous or semi-autonomous vehicle is to promote safety and reduce number of accidents, it

Mansour, Iyad, Singh, Sanjay

Combine HD Map Information to Enhance Road Topology Reasoning by Offline Supervised Training and on-Line Ensemble Learning

2025-01-8021

04/01/2025

Topology reasoning plays a crucial role in understanding complex driving scenarios and facilitating downstream planning, yet the process of perception is inevitably affected by weather, traffic obstacles and worn lane markings on road surface. Combine pre-produced High-definition maps (HDMaps), and other type of map information to the perception network can effectively enhance perception robustness, but this on-line fused information often requires a real-time connection to website servers. We are exploring the possibility to compress the information of offline maps into a network model and integrate it with the existing perception model. We designed a topology prediction module based on graph attention neural network and an information fusion module based on ensemble learning. The module, which was pre-trained on offline high-precision map data, when used online, inputs the structured road element information output by the existing perception module to output the road topology, and

Kuang, Quanyu, Rui, Zhang, Zhang, Song, Yixuan, Gao

Avoiding the Crash: A Vision-Language Model Evaluation of Critical Traffic Scenarios

2025-01-8213

04/01/2025

Autonomous Vehicles (AVs) have transformed transportation by reducing human error and enhancing traffic efficiency, driven by deep neural network (DNN) models that power image classification and object detection. However, to maintain optimal performance, these models require periodic re-training; failure to do so can result in malfunctions that may lead to accidents. Recently, Vision-Language Models (VLMs), such as LLaVA-7B and MoE-LLaVA, have emerged as powerful alternatives, capable of correlating visual and textual data with a high degree of accuracy. These models’ robustness and ability to generalize across diverse environments make them especially suited to analyzing complex driving scenarios like crashes. To evaluate the decision-making capabilities of these models across common crash scenarios, a set of real-world crash incident videos was collected. By decomposing these videos into frame-by-frame images, we task the VLMs to determine the appropriate driving action at each frame

Fernandez, David, MohajerAnsari, Pedram, Salarpour, Amir, Pesé, Mert D.

Bi-Level Control Co-Design for Parallel Electric-Hydraulic Hybrid Vehicles

2025-01-8582

04/01/2025

This study presents a control co-design method that utilizes a bi-level optimization framework for parallel electric-hydraulic hybrid powertrains, specifically targeting heavy-duty vehicles like class 8 semi-trailer trucks. The primary objective is to minimize battery energy consumption, particularly under high torque demand at low speed, thereby extending both battery lifespan and vehicle driving range. The proposed method formulates a bi-level optimization problem to ensure global optimality in hydraulic energy storage sizing and the development of a high-level energy management strategy. Two nested loops are used: the outer loop applies a Genetic Algorithm (GA) to optimize key design parameters such as accumulator volume and pre-charged pressure, while the inner loop leverages Dynamic Programming (DP) to optimize the energy control strategy in an open-loop format without predefined structural constraints. Both loops use a single objective function, i.e. battery energy consumption

Taaghi, Amirhossein, Yoon, Yongsoon

Comparative Analysis of Mass Estimation Methods for Heavy-Duty Vehicles under Varying Payload Conditions

2025-01-8209

04/01/2025

Accurate mass estimation is essential for commercial heavy-duty vehicles (HDVs) because fluctuating payloads significantly impact energy consumption. Precise vehicle mass estimates enhance the accuracy of energy consumption models, leading to more effective energy management systems and performance optimization strategies. For example, improved energy estimates can lead to more optimized routing and refueling schedules, improving operational efficiency and reducing costs. For electric HDVs, accurate mass estimates are crucial for battery sizing, range prediction, and optimized charge scheduling. While direct mass measurements may be obtained through external weight-in-motion or specialized onboard weighing systems, this paper focuses on methods that use data from Controller Area Network systems for alternative real-time predictions. The challenge lies in identifying a method that performs well under the highly variable and often sparse data conditions typical of HDV driving datasets

Jayaprakash, Bharat, Eagon, Matthew, Fakhimi, Setayesh, Kotz, Andrew, Northrop, William

Prediction of Health Status and Remaining Life of Lithium Batteries Based on Deep Learning

2025-01-8126

04/01/2025

Rechargeable lithium batteries are widely used in the electric vehicle industry due to their long lifespan and high energy density. However, after long-term repeated charging and discharging, various electrochemical reactions inside lithium batteries can lead to performance degradation and even cause battery fires. Estimating the health status and predicting the remaining life of lithium batteries can provide insights into their future operating conditions, which is crucial for achieving fault warnings and ensuring the safe operation of battery-related equipment. In terms of predicting the health status of lithium batteries, this paper proposes a method based on an improved Long Short-Term Memory (LSTM) for health status estimation. This method first employs nearest neighbor component analysis to eliminate feature redundancy among the multidimensional health factors of the battery. Then, a differential grey wolf optimization algorithm (DEGWO) is used to globally optimize the

K, Meng Zi

Hydro Bushing Model Identification Using Physics-Informed Neural Networks

2025-01-8263

04/01/2025

Advances in computer aided engineering and numerical methods have made modeling and analyzing vehicle dynamics a key part of vehicle design. Over time, many tools have been developed to model different vehicle components and subsystems, enabling faster and more efficient simulations. Some of these tools use simplified mathematical models to achieve the desired performance. These models depend on model identification methods to determine the parameters and structure that best represent a system based on observed data. This work focuses on the development of a model identification for hydro bushings, a crucial component in nearly all ground vehicles. It introduces an innovative approach to identifying the dynamic properties of hydro bushings using the rapidly evolving physics-informed neural networks. The developed physics-informed network incorporates physical laws into its training process, allowing for an improved mapping of a hydro bushing’s excitation to its dynamic response. The

Koutsoupakis, Josef, Ribaric, Adrijan, Nolden, Ingo, Karyofyllas, George, Giagopoulos, Dimitrios

Development of Ultra-Fast AI-Driven Diesel Engine Model for Real-Time Optimization

2025-01-8397

04/01/2025

Combustion engines and hybrid systems remain important in sectors like light- and heavy-duty vehicles, where performance, range, or cost limitations play a major role. Optimizing diesel engine efficiency and reducing emissions is critical. However, classical physics-based 0D/1D models are computationally demanding and are hardly applicable for real-time purposes. In this study, a calibrated 1D diesel engine model is suggested for transformation into a neural network architecture to enable real-time optimization. The model divides the engine into intake, exhaust, and combustion sections, each modeled by different neural networks. One of the advantages of this modular and layered approach is the flexibility to change individual components without needing to retrain every single model. Long Short-Term Memory (LSTM) networks are used to capture transient phenomena, such as thermal inertias that arise in the combustion process and gas flow dynamics. The training data was generated by

Frey, Markus, Itzen, Dirk, Sautter, Johannes, Weller, Louis, Hagenbucher, Timo, Yang, Qirui, Grill, Michael, Kulzer, Andre Casal

Integrated Nonlinear Observer and Prescribed Performance Control for a Four In-Wheel Motor-Driven Electric Vehicles Based on Stability Boundary and Road Information

2025-01-8298

04/01/2025

To effectively improve the performance of chassis control of a four in-wheel motor (IWM)-driven electric vehicles (EVs), especially in combing nonlinear observer and chassis control for improving road handling and ride comfort, is a challenging task for the IWM-driven EVs. Simultaneously, inaccurate state-based control and uncertainty with system input, are always existing, e.g., variable control boundary, varying road input or control parameters. Due to the higher fatality rate caused by variable factors, how to precisely chose and enforce the reasonable chassis prescribed performance control strategy of IWM-driven EVs become a hot topic in both academia and industry. To issue the above mentioned, the paper proposes a novel observer-based prescribed performance control to improve IWM-driven EVs chassis performance under the double lane change steering. Firstly, a nonlinear nine degree-of-freedom of full-car model is developed to describe vehicle chassis dynamics, and the proposed

Wang, Zhenfeng, Long, Jiarong, Li, Shengchong, Zhang, Xiaoyang, Zhao, Binggen

Multi-Objective Optimization of the Electric Bus Body Frame Based on the PSO-BP-MOMVO Approach

2025-01-8652

04/01/2025

This paper focuses on the design optimization of a commercial electric bus body frame with steel-aluminum heterogeneous material orienting the performances of strength, crashworthiness and body lightweight. First, the finite element (FE) model of the body frame is established for static and side impact analysis, and the body frame is partitioned into several regions according to the thickness distribution of the components. The thicknesses of each region are regarded as the variables for the sensitivity analysis by combining the relative sensitivity method and the Sobol index method, and nine variables to which the performance indexes are more sensitive are selected as the final design variables for design optimization. Then the surrogate models are developed, and in order to improve the accuracy of the surrogate models, a model-constructing method called the particle swarm optimization BP neural network (PSO-BP) data regression prediction is proposed and formulated. In this method

Yang, Xiujian, Tian, Dekuan, Cui, Yan, Lin, Qiang, Song, Yi

Highly Accurate Machine Learning Models for Automotive Crash Applications Using CAE Centric AI/ML Platform

2025-01-8719

04/01/2025

In the automotive industry, there have been many efforts of late in using Machine Learning tools to aid crash virtual simulations and further decrease product development time and cost. As the simulation world grapples with how best to incorporate ML techniques, two main challenges are evident. There is the risk of giving flawed recommendations to the design engineer if the training data has some suspect data. In addition, the complexity of porting simulation data back and forth to a Machine Learning software can make the process cumbersome for the average CAE engineer to set up and execute a ML project. We would like to put forth a ML workflow/platform that a typical CAE engineer can use to create training data, train a PINN (Physics Informed Neural Network) ML model and use it to predict, optimize and even synthesize for any given crash problem. The key enabler is the use of an industry first data structure named mwplot that can store diverse types of training data - scalars, vectors

Krishnan, Radha

Driver Distraction Recognition Based on the Information Bottleneck Theory and Graph Convolutional Networks

2025-01-8112

04/01/2025

Driver distraction remains a leading cause of traffic accidents, making its recognition critical for enhancing road safety. In this paper, we propose a novel method that combines the Information Bottleneck (IB) theory with Graph Convolutional Networks (GCNs) to address the challenge of driver distraction recognition. Our approach introduces a 2D pose estimation-based action recognition network that effectively enhances the retention of relevant information within neural networks, compensating for the limited data typically available in real-world driving scenarios. The network is further refined by integrating the CTR-GCN (Channel-wise Topology Refinement Graph Convolutional Network), which models the dynamic spatial-temporal relationships of human skeletal data. This enables precise detection of distraction behaviors, such as using a mobile phone, drinking water, or adjusting in-vehicle controls, even under constrained input conditions. The IB theory is applied to optimize the trade

Zhang, Ji, Bai, Yakun

A Lateral Motion Controller for Smart Chassis Addressing Structural Uncertainty

2025-01-8197

04/01/2025

In order to make lateral motion robust and stable for smart chassis, a lateral motion controller is proposed in this paper taking structural uncertainty into account. Firstly, a new lateral error model is developed to describe the lateral motion problem. Secondly, the kernel of the lateral motion controller is active disturbance rejection control method, that is, a second order linear tracking differentiator is embedded between lateral error model and first order linear active disturbance rejection control (FO-LADRC). And every module of the lateral motion controller has been designed in detail, which contains first-order linear tracking differentiator (LTD1), second-order linear tracking differentiator (LTD2), linear extended state observer (LESO), and linear state error feedback (LSEF). Finally, six typical scenarios with two conditions are designed to validate the controller referencing to the test standard, considering the structural uncertainty, including wheelbase length and

Li, Yishuai, Lu, Jun, Xu, Shuiling, Ming, Jinghong, Yu, Qin, Wang, Xiaoliang, Zeng, Dequan, Hu, Yiming, Yu, Yinquan, Yang, Jinwen, Jiang, Zhiqiang

Hyperparameter-Optimized Neural Network for Precise Battery State of Charge Estimation

2025-01-8210

04/01/2025

Accurate estimation of the state of charge (SoC) of battery cells is crucial for the efficient management and longevity of battery systems, particularly in electric vehicles and renewable energy storage. This paper presents an approach utilizing a nonlinear autoregressive exogenous (NARX) model to estimate the SoC of battery cells. The proposed method leverages hyperparameter optimization to determine the optimal configuration of the neural network, including the number of neurons, the number of hidden layers, the number of feedback loops, the best activation function, and the most effective learning rate. The primary objective of this research is to minimize the estimation error of the SOC to within 2%, thereby enhancing the reliability and performance of battery management systems. The hyperparameter optimization process involves a systematic search and evaluation of various configurations to identify the most effective neural network architecture. This process is critical as it

Saini, Sandeep, Admane, Chinmay

Load-Deflection Evaluation of Bump Stoppers Using a Machine Learning Approach

2025-01-8208

04/01/2025

The accurate prediction and evaluation of load-deflection behavior in bump stoppers are critical for optimizing driving performance and durability in the automotive industry. Traditional methods, such as extensive experimental testing and finite element analysis (FEA), are often time-consuming and costly. This paper introduces a machine learning-based approach to efficiently evaluate load-deflection curves for bump stoppers, thereby streamlining the design and testing process. By leveraging a comprehensive dataset that includes historical test results, material properties, and geometrical dimensions, various machine learning methods, including Gradient boosting, Random forest & XG Boost were trained to predict load-deflection behavior with high accuracy. This approach reduces the reliance on extensive physical testing and simulations, significantly enhancing the design optimization process, leading to faster development cycles and more precise performance predictions. A case study is

Hazra, Sandip, Tangadpalliwar, Sonali

Study on an Urban Ramp Driving Cycle Using Self-Organizing Map Neural Network

2025-01-8259

04/01/2025

To tackle the issue of lacking slope information in urban driving cycles used for vehicle performance evaluation, a construction method for urban ramp driving cycle (URDC) is formulated based on self-organizing map (SOM) neural network. The fundamental data regarding vehicles driving on typical roads with urban ramp characteristics and road slopes were collected using the method of average traffic flow, which were then pre-processed and divided into short-range segments; and twenty parameters that can represent the operation characteristics of vehicle driving on urban ramp were selected as the feature parameters of short-range segments. Dimension of the selected feature parameters was then reduced by means of principal component analysis. And a SOM neural network was applied in cluster analysis to classify the short-range segments. An URDC with velocity and slope information were constructed by combination of short-range segments with highly relevant coefficients according to the

Yin, Xiaofeng, Wu, Zhimin, Liang, Yiming, Wang, Peng, Xie, Yu

Image-Based Machine Learning Methods in Materials Microstructure and Failure Analysis

2025-01-8324

04/01/2025

Image-based machine learning (ML) methods are increasingly transforming the field of materials science, offering powerful tools for automatic analysis of microstructures and failure mechanisms. This paper provides an overview of the latest advancements in ML techniques applied to materials microstructure and failure analysis, with a particular focus on the automatic detection of porosity and oxide defects and microstructure features such as dendritic arms and eutectic phase in aluminum casting. By leveraging image-based data, such as metallographic and fractographic images, ML models can identify patterns that are difficult to detect through conventional methods. The integration of convolutional neural networks (CNNs) and advanced image processing algorithms not only accelerates the analysis process but also improves accuracy by reducing subjectivity in interpretation. Key studies and applications are further reviewed to highlight the benefits, challenges, and future directions of

Akbari, Meysam, Wang, Andy, Wang, Qigui, Yan, Cuifen

Course of Action (COA) Generation for Robotic Military Ground Vehicles

2025-01-8338

04/01/2025

Course of action (COA) generation for robotic military ground vehicles is required to support autonomous operations in well-structured and non-structured environments. Traditional pathing algorithms such as Dijkstra, A*, Hybrid A*, or D* are exhaustive and well structured, and as a result, a single COA may be derived if one exists. Traditional path-planning algorithms have been optimized to identify paths that achieve a single scalar objective (duration, distance, energy, etc.). The algorithms are not natively able to account for multi-objective cost considerations. Military operations represent multi-objective optimization problems, impacted by time, space, and atmospherics. The battlefield is dynamic and ever-changing, thus pathing algorithms must incorporate multi-objective costs and constraints and be provided in near-real-time or real-time. For this reason, the use of a genetic algorithm (GA) and Artificial Intelligence/ Machine Learning (AI/ML) were investigated for COA

Jane, Robert, Ferrying, Zane, Janat, Tsegaye, James, Corey

Neural Network Based Modeling and Model Predictive Control for Reduction in Diesel Emissions

2025-01-8369

04/01/2025

The paper illustrates the process and steps in the development of a neural network-based economic Model Predictive Control (MPC) strategy for reducing diesel engine feed gas emissions. This MPC controller performs fuel limiting and modifies intake manifold pressure and exhaust gas recirculation (EGR) rate set-points to the inner loop air path controller to reduce engine-out oxides of nitrogen (NOx) and Soot emissions. We examine two Recurrent Neural Network (RNN) options for a control-oriented emissions model which are based on a multi-layer perception (MLP) architecture and a long short-term memory (LSTM) architecture. These RNN models are trained for use as prediction models in MPC. Both models are defined in input-output form, assuming that measurements/estimates of current values of NOx and Soot are available. We discuss and compare their training using PyTorch. The formulation of economic MPC is detailed, including the definition of the cost function and soft constraints

Zhang, Jiadi, Li, Xiao, Kolmanovsky, Ilya, Tsutsumi, Munecika, Nakada, Hayato

Dynamic Performance Optimization of Double-Isolation Rubber Mounts Based on Neural Network and Genetic Algorithm Collaboration

2025-01-8266

04/01/2025

As a crucial connecting component between the powertrain and the chassis, the performance of rubber mounts is directly related to the NVH (Noise, Vibration, and Harshness) characteristics of electric vehicles. This paper proposes a double-isolation rubber mount, which, compared to traditional rubber mounts, incorporates an intermediate skeleton and features inner and outer layers of “cross-ribs”. The design parameters can be simplified to: skeleton diameter, skeleton thickness, main rib width, and main rib thickness. To comprehensively evaluate its performance, a finite element analysis (FEA) model of the proposed double-isolation rubber mount was first established in Abaqus, with static stiffness and dynamic performance analyzed separately. The results indicate that, compared to traditional rubber mounts with similar static stiffness, this design effectively controls dynamic stiffness in the high-frequency range. To expand the effective vibration isolation frequency range of the

Xu, Che, Kang, Yingzi, Tu, Xiaofeng, Shen, Dongming

WITHDRAWAL NOTICE

2025-01-8042.01

04/01/2025

Fuel Consumption Estimation Using Spatio-Temporal Modeling and Traffic Flow Predictions: A Comparative Analysis

2025-01-8101

04/01/2025

Effective traffic management and energy-saving techniques are increasingly needed as metropolitan areas grow and traffic volumes rise. This work estimates fuel consumption over three selected routes in an urban context using spatio-temporal modeling essentially building on a previously developed approach in traffic prediction and forecasting. A weighted adjacency matrix for a Graph Neural Network (GNN) is constructed in the original approach which combines graph theory frameworks with travel times obtained from average speeds and distances between traffic count stations. Next, the traffic flow estimate uncertainty is measured using Adaptive Conformal Prediction (ACP) to provide a more reliable forecast. This work predicts fuel consumption under different scenarios by utilizing Monte Carlo simulations based on the expected traffic flows providing insights into energy efficiency and the best routes to take. The study compares passenger vehicles' and heavy-duty trucks' mean fuel

Patil, Mayur, Moon, Joon, Hanif, Athar, Ahmed, Qadeer

Machine Learning–Based Modeling for Predicting the Lap Shear Strength of High-Strength Steel Laser Lap Welds

05-18-04-0029

03/29/2025

This study investigates the nonlinear correlation between laser welding parameters and weld quality, employing machine learning techniques to enhance the predictive accuracy of tensile lap shear strength (TLS) in automotive QP1180 high-strength steel joints. By incorporating three algorithms: random forest (RF), backpropagation neural network (BPNN), and K-nearest neighbors regression (KNN), with Bayesian optimization (BO), an efficient predictive model has been developed. The results demonstrated that the RF model optimized by the BO algorithm performed best in predicting the strength of high-strength steel plate-welded joints, with an R 2 of 0.961. Furthermore, the trained RF model was applied to identify the parameter combination for the maximum TLS value within the selected parameter range through grid search, and its effectiveness was experimentally verified. The model predictions were accurate, with errors controlled within 6.73%. The TLS obtained from the reverse-selected

Han, Jinbang, Ji, Yuxiang, Liu, Yong, Liu, Zhao, Wang, Xianhui, Han, Weijian, Wu, Kun

GA-BP Neural Network-Based Prediction of Impact Resistance in Electric Vehicle Charging Piles

05-18-04-0028

03/27/2025

Impact resistance is crucial for assessing charging pile safety and reliability. This study proposes a prediction model, called GA-BP neural network, which achieved prediction errors below 5% and reduced computation time by over 95% in comparison to finite element analysis (FEA). Initially, the charging pile impact test platform is constructed, and a matching finite element simulation model is developed. The correctness of the simulation model is then verified by integrating the experimental findings. Furthermore, the Latin hypercube approach is used to create 200 sets of simulation schemes, and using the Python programming language, the impact resistance performance indicators of charging piles are automatically collected. Next, a genetic algorithm is used to optimize the initial weight and bias of the BP neural network, lastly, fine-tune the hyperparameters in the neural network to develop a prediction model for the impact resistance performance of the charging pile. The GA-BP model

Jiang, Bingyun, Hu, Peng, Liu, Zhenyu, Yuan, Pengfei, Liu, Hui

Multi-objective Optimization of Process Parameters for Dual-Color Injection Molding of the Middle Cover of Electric Vehicle Charging Pile

05-18-04-0031

03/26/2025

This article takes the cover of the AC charging pile as the research object and studies the process parameters of dual-color injection molding. First, the optimal Latin hypercube experimental design is carried out by using optimization software by taking the melt temperature and mold temperature of the first shot and the second shot and the holding pressure as the influencing factors. Injection simulation is carried out based on mold flow software. A high-precision neural network model RBF is constructed according to the test factors and results. Second, based on the obtained RBF prediction model, the multi-objective NSGA-II algorithm is used for optimization. The obtained optimal combination of molding process parameters is: the melt temperature of the first shot is 266.8°C, the mold temperature is 107°C, the melt temperature of the second shot is 230.3°C, the mold temperature is 59.5°C, the holding pressure of the first shot is 95 MPa, the holding pressure of the second shot is 89.9

Liu, Hao, Jiang, Bingyun, Jiang, Hong, Hu, Peng, Cheng, Shan

Fault-Tolerant Control of a Quadcopter Using Reinforcement Learning

01-18-01-0006

03/03/2025

This study presents a novel reinforcement learning (RL)-based control framework aimed at enhancing the safety and robustness of the quadcopter, with a specific focus on resilience to in-flight one propeller failure. This study addresses the critical need of a robust control strategy for maintaining a desired altitude for the quadcopter to save the hardware and the payload in physical applications. The proposed framework investigates two RL methodologies, dynamic programming (DP) and deep deterministic policy gradient (DDPG), to overcome the challenges posed by the rotor failure mechanism of the quadcopter. DP, a model-based approach, is leveraged for its convergence guarantees, despite high computational demands, whereas DDPG, a model-free technique, facilitates rapid computation but with constraints on solution duration. The research challenge arises from training RL algorithms on large dimension and action domains. With modifications to the existing DP and DDPG algorithms, the

Qureshi, Muzaffar Habib, Maqsood, Adnan, Fayyaz ud Din, Adnan

When Mushrooms Control Robots

TBMG-52640

03/01/2025

In creating a pair of new robots, Cornell researchers cultivated an unlikely component: fungal mycelia. By harnessing mycelia’s innate electrical signals, the researchers discovered a new way of controlling “biohybrid” robots that can potentially react to their environment better than their purely synthetic counterparts.

Methods for Selecting Temporal Granularity in Passenger Flow Forecasting

2025-01-7173

02/21/2025

With the increasing improvement of urban rail transit network, the passenger flow of rail transit is also increasing year by year, and the management methods for passenger flow are becoming more and more intelligent and informative. To grasp the changing pattern of passenger flow has become the focus of research on rail transit operation. Previous studies have been conducted in a single time granularity for prediction, ignoring the influence of different time granularities on the prediction results. Furthermore, selecting time granularity solely based on prediction accuracy is inadequate, as larger time granularities result in a loss of detail regarding the specific changes in passenger flow, which can be referred to as passenger flow characteristics. In order to address the above issues, this study use Person correlation analysis and systematic clustering method to select the best time granularity. Then the BP neural network is employed to verify the prediction effect of different

Zhang, Zhenyu, Wang, Jian, Huang, Lianbo

Remaining Useful Life Prediction Method for Highway Electromechanical Equipment Based on Bayesian Algorithm-Optimized CNN-LSTM Model

2025-01-7149

02/21/2025

Efficient maintenance of highway electromechanical equipment is crucial for ensuring reliability within intelligent highway infrastructure and optimizing the allocation of limited maintenance resources. Traditional Remaining Useful Life (RUL) prediction models frequently face limitations due to the complex and dynamic operating conditions of such systems, which often hinder their predictive accuracy and adaptability. To overcome these persistent challenges, this study introduces an advanced RUL prediction model that integrates a Bayesian-optimized Convolutional Neural Network (CNN) with a Long Short-Term Memory (LSTM) network. Initially, the study identifies key health indicators that effectively represent the degradation of equipment performance over time. These indicators undergo Spearman correlation analysis to determine their relevance to equipment capacity, ensuring that only the most pertinent features are used for model input. The CNN-LSTM model leverages CNN’s spatial pattern

Wang, Leyan, Zhang, Jian, Yao, Xuejian, Ping, Hao

A Traffic Conflict Risk Assessment Model for Highway Construction Zones Based on Trajectory Data

2025-01-7217

02/21/2025

Highway construction zones present substantial safety challenges due to their dynamic and unpredictable traffic conditions. With the rising number of highway projects, limited accident data during brief construction phases underscores the need for alternative safety evaluation methods, such as traffic conflict analysis. This study addresses vehicular safety issues within the Kunshan section of the Shanghai-Nanjing Expressway, focusing on conflict risk assessment through a spatio-temporal analysis of a construction zone. Using drone-captured video, vehicle trajectories were extracted to derive key operational indicators, including speed and acceleration, providing a spatio-temporal foundation for analyzing traffic flow and conflict dynamics. A novel **Comprehensive Collision Risk Index (CCRI)** was introduced, integrating Time-to-Distance-to-Collision (TDTC) and Enhanced Time-to-Collision (ETTC) metrics to enable a multidimensional assessment of conflict risk. The CCRI captures both

Zhang, Yuwen, Guo, Xiucheng, Ma, Yuheng

Vehicle Trajectory and Traffic Conflict Prediction at Urban Expressway Confluence

2025-01-7128

02/21/2025

The urban expressway ramp entrance has always been one of the traffic accident prone areas. At present, most of the traffic safety research focuses on the intersection traffic conflict prediction analysis, so it is necessary to conduct related research on the urban expressway ramp entrance traffic safety. In this paper, we focus on trajectory prediction and traffic conflict analysis between straight vehicles and ramp vehicles on main roads for small data sets. Firstly, the transformer model and LSTM model were used to predict the trajectories of the straight vehicle on the main road and the vehicle at the entrance of the ramp respectively. TTC is calculated from the predicted trajectories of straight vehicles on the main road and confluence vehicles on the ramp. Then, the historical features are used to predict the future TTC by using LSTM plus self-attention mechanism, and the two conflict prediction methods are compared. The results show that in the case of small data sets, the

Liu, Tiansheng, Xiang, QiaoJun

Uncertainty Quantification for Predicting Low-Temperature Crack Resistance of Asphalt Mixture Base on Bayesian Neural Network

2025-01-7143

02/21/2025

Developing models for predicting the low-temperature cracking resistance of asphalt mixtures is a complex process with a wide variety and complex influence mechanisms of variables, leading to higher uncertainty in the prediction results. Several models have been developed in this regard. This study developed a Bayesian neural network (BNN) model for predicting the fracture energy of low-temperature semi-circular bending (SCB) tests based on pavement condition measurements, traffic, climate, and basic parameters of the material. The model was trained and evaluated using low-temperature SCB test data from in-situ pavement core samples, and the results showed that the coefficient of determination (R2) of the BNN model was greater than 0.8 for both the training and testing sets. The variable importance scores showed that the decrease of transverse crack rating index (TCEI) and gradation were the most important factor affecting low-temperature fracture energy and that the ambient

Song, Ziyu, Ni, Fujian, Huang, Jiaqi, Jiang, Jiwang

Classification of Pavement Distress Images Using Fusion Convolutional Neural Network of Dual Branch

2025-01-7131

02/21/2025

In recent years, the issue of highway maintenance has become increasingly prominent. How to precisely detect and classify fine cracks and various types of pavement defects on highways through technical means is an essential foundation for achieving intelligent road maintenance. This paper first constructs the DenseNet201-PDC and MobileNetV2-PDC sub-classification networks that incorporate the three-channel attention judgment mechanism MCA. Secondly, based on the principle of parallel connection, a brand-new dual-branch fusion convolutional neural network DBF-PDC capable of classifying pavement defects in highway scenarios is proposed. Finally, this paper builds the Pavement Distress Datasets of Southeast University and conducts relevant ablation experiments. The experimental results demonstrate that both the attention mechanism module and the feature fusion strategy can significantly enhance the network's ability to classify pavement defects in highway scenarios. The average

Zhang, Ziyi, Zhao, Chihang, Shao, Yongjun, Wang, Junjun

Matching and Fusing Multi-Source Vehicle Information in Highway Gantry Scenarios Using Convolutional Neural Networks

2025-01-7136

02/21/2025

The performance differences of multiple sensors lead to inconsistencies, incompleteness, and distortion in the perception data of multi-source vehicle information in highway scenarios. Optimizing data fusion methods is important for intelligent toll collection systems on highways. First, this paper constructs a dataset for matching and fusing multi-source vehicle information in highway gantry scenarios. Second, it develops convolutional neural network models, Match-Pyramid-MVIMF-EGS and CDSSM-MVIMF-EGS, for this purpose. Finally, comparative experiments are conducted based on the constructed dataset to assess the performance of the Match-Pyramid-MVIMF-EGS and CDSSM-MVIMF-EGS models. The experimental results indicate that the Match-Pyramid-MVIMF-EGS model performs better than the CDSSM-MVIMF-EGS model, achieving matching and fusion accuracy of 93.07%, precision of 95.71%, recall of 89.17%, F1 scores of 92.32%, and 186 of training throughput respectively.

Wang, Junjun, Zhao, Chihang

Items per page:

1 – 50 of 1328