Browse Topic: Machine learning

Items (549)

Prediction of Trimmed BIW Noise transfer function based on BIW response using Machine Learning

2026-26-0640To be published on 01/16/2026

This research focuses on the prediction of the Noise Transfer Function (NTF) in a trimmed Body-in-White (BIW) structure by leveraging machine learning technique considering BIW-level NTF, dynamic stiffness at engine mount attachment points, and Vibration Transfer Function (VTF). The study employs an iterative methodology, adjusting the thickness of various BIW panels to analyse their impact on noise transfer characteristics. Frequency response functions are generated at the BIW attachment points to assess the effectiveness of these modifications. By systematically increasing and decreasing panel thicknesses, and stiffness at the attachment point and trimmed BIW level responses the study aims to optimize the acoustic performance of the BIW structure. The findings provide critical insights into the relationship between structural modifications and noise transfer, offering valuable guidance for automotive design and engineering during early design stage to reduce the development time.

Kulkarni, Prasad Ramesh, Bijwe, Vilas, Kulkarni, Shirish, Sahu, Dilip, Inamdar, Pushpak

Application of DOE Techniques to Identify Factors Influencing Gearbox Synchronizer Wear and Life.

2026-26-0489To be published on 01/16/2026

Gearbox synchronizers are essential components for ensuring smooth gear shifting in both manual and automated manual transmissions. The lifespan of a synchronizer is influenced by a complex interplay of several factors, including synchro geometry, lubrication, friction material properties, clutch design parameters, powertrain vibrations, and various operating conditions. These factors contribute to the wear and degradation of synchronizers over time, affecting the overall performance and durability of the transmission system. This paper investigates the influence of various design and operational parameters on the synchronizer's life using a Design of Experiments (DOE) approach. A comprehensive experimental plan was meticulously developed to systematically vary key factors such as synchronizer friction material, the number of cones, lubricant viscosity, and clutch stiffness, among others. The resulting experimental data was analyzed through statistical methods, particularly regression

Jagtap, Amol Shivaji, Koulage, Dasharath, Rudramath, Sagar

Advanced Indirect Tire Health Monitoring System Using Transfer Learning for Predictive Maintenance

2026-26-0670To be published on 01/16/2026

Tire wear is a critical factor in vehicle safety and operational reliability, with worn-out tires contributing to nearly 30% of highway blowouts and a significant portion of tire-related accidents. Traditional tire wear indicators—such as tread wear bars or mileage-based replacement schedules—lack precision, do not offer real-time feedback, and often fail to prevent hazardous situations caused by uneven or rapid tire degradation. These shortcomings underscore the urgent need for intelligent, data-driven solutions in predictive tire maintenance. While earlier efforts have focused on physical models or basic machine learning algorithms, their limited adaptability and poor generalization across platforms reduce effectiveness in real-world deployments. This paper presents a Machine Learning-based Tire Wear Prediction System, designed for both cloud-based analytics and on-vehicle edge deployment. The system predicts individual tire wear using CAN-based inputs, supported by a transfer

Imteyaz, Shahma, Iqbal, Shoaib

Chest Deflection mapping between HBM Connect® 50M and Humanetics ATD models for frontal impact

2026-26-0010To be published on 01/16/2026

Objective Human Body Models (HBMs) are increasingly recognized by consumer protection agencies as essential tools for evaluating vehicle safety, complementing the use of traditional Anthropomorphic Test Devices (ATDs). However, HBMs are new for product development and pose challenges in connecting them with the ATDs. These challenges can be overcome if a link is established for the injury metrics between HBMs and ATDs. The study aims to develop a chest deflection mapping function between the HBM Connect® 50M (HBMC-50M) and two ATD models: THOR-50M and Hybrid III 50M for thoracic assessment in frontal impact. Method Several frontal chest loading scenarios were selected from the HBM4VT qualification catalogue (Euro NCAP technical bulletin CP 550), including hub impacts and table-top tests. Matched-pair LS-DYNA simulations were conducted with HBMC-50M and the ATD models recording peak chest deflections for developing the mapping function. In the HBMC-50M model, deflection outputs were

Velmurugan, Gopinath, Kumar PhD, Devendra, R, Udhaya Kumar, Kulavi, Pradeep, Soni, Anurag, Tejero de la Piedra, Ricardo, Fu, Stephen, Long, Teng, Arora, Tushar, Jagadish, Renuka, Shah, Chirag

Real-time Intrusion Detection System (IDS) for Vehicle Networks Using Machine Learning for Predictive Threat Analysis

2026-26-0617To be published on 01/16/2026

The rapid adoption of connected vehicle technologies and advanced driver assistance systems (ADAS) necessitates robust security mechanisms capable of identifying and mitigating sophisticated cyber threats in real-time. Traditional signature-based intrusion detection systems (IDS) are often inadequate in addressing the dynamic and evolving nature of automotive cybersecurity threats, particularly in modern vehicle networks like Controller Area Network (CAN), CAN with Flexible Data-Rate (CAN-FD), and Automotive Ethernet. This research introduces a novel Real-time Intrusion Detection System utilizing advanced Machine Learning (ML) techniques designed specifically for automotive network environments. The proposed IDS framework employs supervised and unsupervised ML algorithms, including anomaly detection, behavioral analytics, and predictive threat modeling, to achieve high accuracy and rapid threat identification capabilities. Through extensive testing in simulated and actual vehicle

chaudhary lng, Vikash, Desai, Manoj, Chatterjee, Avik

Proactive AI-Enhanced Cybersecurity for Next-Gen Automotive Infotainment ECU

2026-26-0619To be published on 01/16/2026

Abstract With the rapid advancement of connected vehicle technologies, infotainment Electronic Control Units (ECUs) have become central to user interaction and connectivity within modern vehicles. However, this enhanced functionality has introduced new vulnerabilities to cyberattacks. This paper explores the application of Artificial Intelligence (AI) in enhancing the cybersecurity framework of infotainment ECUs. The study introduces AI-powered modules for threat detection and response, presents an integrated architecture, and validates performance through simulation using MATLAB, CANoe, and NS-3. This approach addresses real-time intrusion detection, anomaly analysis, and voice command security. Key benefits include zero-day exploit resistance, scalability, and continuous protection via OTA updates. The paper references real-world automotive cyberattack cases such as Tesla's OTA vulnerability patches, BMW ConnectedDrive exploits, and Jeep Cherokee's Uconnect hack, emphasizing the

More, Shweta, Kulkarni, Shraddha, Kumar, Priyanshu, GHANWAT, HEMANT, Joshi, Vivek

AI-Driven Stress and Fatigue Life Prediction of Engine Connecting Rods using Geometric Deep Learning

2026-26-0633To be published on 01/16/2026

The connecting rod is one of the critical components of the engine which is subjected to complex cyclic loading, demanding precise evaluation of stress distribution and fatigue life. Conventional Finite Element Method (FEM)-based simulations, while accurate, are computationally expensive and time-intensive—especially when iterating through multiple design variations. To address this, we present an AI/ML-based predictive framework using Graph Neural Networks (GNNs) and Geometric Deep Learning (GDL) techniques to efficiently predict stress and fatigue life of the connecting rod. The methodology involves representing FEM mesh data as graph structures, where nodes and edges capture geometric and physical relationships. GNN architectures, including MeshCNN and PointNet++, are trained on historical simulation data to learn spatial dependencies and nonlinear mappings between geometry, load conditions, and resulting stress/life outputs. The framework significantly reduces computation time

Pathan, Mohammed Shakil, K, Karthikeyan, pilla, sashanka, S Kangde, Suhas

Identification of productive and non-productive activities in agricultural machinery using ECU and GPS parameters through the application of machine learning techniques

2026-26-0102To be published on 01/16/2026

Any agricultural operation (such as cultivation, rotavation, ploughing, and harrowing) includes both effective work and non-productive activities (like transportation, stops, and idling) in the field. Non-productive work can distort the actual load profile, fuel consumption, and emissions. In this project, a machine learning-based methodology has been developed to differentiate between effective operations and non-productive activities, utilizing data collected in the field from data loggers installed on the machinery. Measurements were conducted on various machines across the country in all major applications to minimize the influence of any individual sample's deviation and to account for variability in customer operating practices. Nine critical parameters (Engine Speed, Exhaust Gas Temperature, THECU2, VPEXGAS, FULRAT, Actual Engine Percentage Torque, LBEGR, POIL & GPS Speed) were selected after screening & analyzing more than 100 CAN and GPS parameters. The critical parameters

Maharana, Devi prasad, Gangsar, Purushottam, gokhale, Varun, Pandey, Anand Kumar

AI-Driven Digital Twin for Enhanced Suspension Assembly Testing in Automotive Vehicles

2026-26-0445To be published on 01/16/2026

This paper presents the development and application of a digital twin for the suspension assembly of Automotive vehicle, a critical component in evaluating the performance and durability of automotive systems. The digital twin, a virtual replica of the physical suspension assembly, is created using advanced simulation techniques and physical test data integration. By leveraging machine learning algorithms, the digital twin provides a comprehensive and dynamic model that accurately reflects the behavior and performance of the physical system under various conditions. The primary objective of this study is to enhance the accuracy and efficiency of suspension assembly testing by enabling predictive maintenance, real-time monitoring, and optimization of test parameters. The digital twin allows for the identification of potential issues before they occur, reducing downtime and maintenance costs. Additionally, it facilitates the exploration of different test scenarios and conditions without

Sonavane, Pravinkumar, Patil, Amol

Machine Learning based Engine Mount NVH Life Health Monitoring System

2026-26-0359To be published on 01/16/2026

Refinement for NVH (Noise, Vibration, and Harshness) performance, and long-term vehicle reliability are rapidly evolving in today’s automotive industry. Engine mounts play a central role in isolating powertrain-induced vibrations; their deterioration can significantly affect cabin comfort, powertrain integrity, and customer satisfaction. Prior work in this area has primarily focused on direct mount sensors and physical inspection at service centre after failure. While effective in controlled environments, such methods are not scalable, add system complexity, and increase vehicle cost due to sudden breakdowns—challenges that restrict practical OEM deployment. This paper introduces a novel indirect health monitoring method that leverages a driver seat rail-mounted accelerometer to capture driver specific vibrational responses. By analyzing these signals using machine learning models and domain-specific analytical features, engine mount health is inferred without requiring sensors on the

Iqbal, Shoaib, Dusane, Mangesh

A Novel Wide Range Machine Learning Polynomial Regression Formulation for Enhanced Load Cell Calibration Accuracy

2026-26-0567To be published on 01/16/2026

Calibration of measuring instruments is of utmost importance in the field of metrology. It is a mandatory pre-requisite for establishing the fidelity of the measurements as well as to lend confidence. Even more critical is the requirement for the master equipment deployed to calibrate the devices in use. This entails that high accuracy needs to be guaranteed in the calibration process, and that the uncertainty be quantified precisely. The widely used conventional least squares polynomial regression formulation for load cell calibration is based on the non-normalized residual, which is the difference between the measured and master values. The nature of this formulation is such that it imparts more weightage on measured values at higher ranges resulting in good accuracy. However, there is a limitation of this same formulation that results in lesser accurate fit at lower values especially if the instrument is to be used in operation over a wide range including lower ranges of the

S Thipse, Yogesh

GTSRB-Shield: A Novel Machine Learning Framework for Intrusion Detection in Autonomous Vehicles Leveraging the German Traffic Sign Recognition Benchmark

2026-26-0611To be published on 01/16/2026

The escalating dependence of Autonomous Vehicles on Intelligent Transportation Systems (ITS) has highlighted the imperative for comprehensive security protocols to safeguard such vehicles against cyber threats. Intrusion Detection Systems (IDS’s) are pivotal in ensuring the protection of these systems by detecting and alleviating unauthorized access and nefarious activities. The German Traffic Sign Recognition Benchmark (GTSRB) database, which encompasses an extensive compilation of traffic sign imagery, functions as a vital asset for the advancement of machine learning-based IDS. This research elucidates an intrusion detection system (IDS) that employs machine learning algorithms to scrutinize the GTSRB database. The proposed IDS emphasize the preprocessing of the GTSRB dataset to extricate pertinent features that can be employed for the training of machine learning models. Research also focuses on model development with machine learning algorithms to classify traffic signs and

Patil, Kamalesh, Akbar Badusha, A., Jadhav, Savitri, Gunale, Kishanprasad

Advanced Sensor less Machine Learning Technique for Early Detection of Engine Oil Degradation

2026-26-0656To be published on 01/16/2026

Maximizing vehicle uptime and reducing maintenance costs are critical objectives in modern automotive systems, making efficient resource utilization a top priority. One key factor is engine oil degradation, which directly affects engine performance, longevity, and overall vehicle efficiency. Current Conventional oil monitoring approaches are real time based on dedicated sensors—can be inefficient, costly, or poorly adapted to varying usage conditions. While earlier studies have explored sensor-based or chemical analysis methods for oil health monitoring, these solutions often lack scalability and cost-effectiveness for widespread adoption. This paper introduces a Machine Learning-based Engine Oil Degradation Detection System that operates non-intrusively using existing vehicle CAN signals, requiring no additional sensors. Key engine parameters such as RPM, Torque, Gear Position, Engine Power, Coolant Temperature, and Odometer readings are trained with tri-bological properties

Dusane, Mangesh, Tade, Vilas, Iqbal, Shoaib

Data analytics methodology for vehicle quality complaint analysis

2026-26-0654To be published on 01/16/2026

Automobiles have become a part of people's everyday life. Customers base their trust on a brand's reputation while making a buying decision. With the evolving trends, feature advancements, automobiles have become a complex system. Ensuring product quality amidst system complexities is necessary to maintain the brand value & ensure customer satisfaction. Efficient addressal of customer quality complaints is pivotal in maintaining product quality. Until recent times, manual analysis using traditional data processing tools has been used for vehicle complaint analysis. However, considering the large quantum of data in new age vehicles with more number of control units, it becomes time-consuming, expertise dependent and susceptible to human errors. Given this & the advent of new age machine learning techniques & data analysis tools that can handle large amount of data, it is imperative to adopt them for vehicle complaint analysis. This research paper puts forth a method to analyze vehicle

Venkatesan, Prasanna, Raju, Vivek

Representative Drive cycle velocity profiles Prediction using ML Algorithms for Energy Efficiency

2026-26-0638To be published on 01/16/2026

Maximizing vehicle energy efficiency and performance is a high priority for the automotive industry as customer expectations rise. Engineers constantly face the challenge of balancing the conflicting goals of achieving superior performance and maximizing energy efficiency, all while meeting increasingly tight development timelines. Leveraging digital methods can potentially enable a considerable reduction in development lead times. Driving cycles act as standardized measurement procedures for certifying a vehicle's fuel efficiency and driving range. Representative velocity profiles condense numerous real-life driving cycles to enable quicker energy analysis and driver feedback evaluation. This paper introduces a novel methodology for generating synthetic drive cycles, such as average velocity cycles and ideal consumption velocity cycles, based on real-life driving scenarios. In this study, the importance of creating representative drive cycles to enhance vehicle performance and energy

Kanakannavar, Rohit, Kelkar, Kshitij, Sadalge, Anand

Realistic LiDAR Data Simulation for Autonomous Systems using Physics-Informed Learning

2026-26-0138To be published on 01/16/2026

Accurate and realistic simulation of LiDAR data is critical for the development and validation of autonomous driving systems. However, existing simulation approaches often suffer from a significant sim-to-real gap due to oversimplified modelling of physical interactions and environmental factors. In this work, we present a physics-informed deep learning framework that bridges this gap by enhancing the realism of simulated LiDAR data using generative adversarial networks guided by domain-specific physical constraints for LiDAR intensity. Our method incorporates key physical factors such as range, surface material properties, angle of incidence, and environmental conditions along with their underlying physical relationships as constraints into the Cycle-Consistent GAN architecture, enabling it to learn realistic transformations from synthetic to real-world LiDAR intensity data without requiring paired samples. We demonstrate the effectiveness of our approach across multiple datasets

Anand, Vivek, Yadav, Sourav, Limba, Mohit, Pandey, Gaurav, Lohani, Bharat

Data-Driven Prediction of Damper durability Cycles Using AIML Techniques

2026-26-0550To be published on 01/16/2026

The suspension dampers play a pivotal role in controlling suspension dynamics, especially during rapid suspension movements such as pothole impacts, speed bumps, and off-road conditions. The dampers directly influence the ride comfort, vehicle stability. Ensuring the durability of these valves is essential to maintain consistent performance throughout the vehicle's lifecycle. To validate the valve configuration at high speed, in current development practices, durability cycles are typically derived through physical data collection on actual vehicle driven on proven ground tracks. While this method provides accurate real world validation, it is highly time consuming and resource intensive. The dependency on full vehicle testing also limits early phase of validation, creating bottlenecks in the product development cycle. An AIML based methodology is presented in this paper to predict the HSVD cycles for damper validation. The AIML algorithm was trained using the historical test data and

Kabbin, Chetan D., Suryawanshi, Vishal, Daphal, Pratap, Kulkarni, Prasad

AI-Powered Innovation: Harnessing Data Analytics and AI in Modern Powertrain Development

2026-26-0641To be published on 01/16/2026

The proposed presentation will delve into the transformative role of artificial intelligence (AI) in the automotive industry, specifically focusing on product development from concept to series support. The presented effort is driven by the need to enhance development efficiency, reduce time to market, improve product quality, and lower production costs. AI was explored as a solution to address challenges such as late availability of hardware components, issues during software integration, and high warranty costs. The AI applications discussed are primarily in the automotive industry, covering areas such as vehicle concept development, usage space analysis, root cause analysis, battery health monitoring, and failure prediction and health monitoring. The methodologies used include advanced AI methods to shorten the time from customer expectations to target agreements, statistical driving data analysis to understand user behaviour, machine learning (ML) surrogate models to identify key

Mayr, Christian, Keuth, Nikolaus

Software-Defined Vehicles: Architecting the Future of Intelligent and Connected Mobility

2026-26-0691To be published on 01/16/2026

The rise of Software-Defined Vehicles (SDVs) marks a fundamental shift in automotive design, where software, rather than hardware, defines vehicle capabilities. This review explores the SDV paradigm within Intelligent Transportation Systems (ITS), emphasizing centralized computing, over-the-air (OTA) updates, and service-oriented architectures (SOA). Key enablers such as high-performance computing units, middleware platforms (e.g., AUTOSAR Adaptive), and digital twins support scalable, flexible software deployment and real-time functionality. Artificial Intelligence (AI) and Machine Learning (ML) enhance features like predictive maintenance, driver personalization, and autonomous decision-making. However, SDVs also introduce complex challenges, including software validation, real-time performance, and system interoperability. Cybersecurity becomes critical, especially as vehicles interface with cloud platforms, edge computing, and Vehicle-to-Everything (V2X) networks. The paper also

Ahmad, Aqueel, Hemanth, Khimavath, Kumar, Om, Kumar, Rajiv, Haregaonkar, Rushikesh Sambhaji

Real-time Helmet Detection and Closed-Loop Enforcement System for Two-Wheelers

2026-26-0650To be published on 01/16/2026

Need: In India, two-wheelers account for a significant portion of road traffic and contribute heavily to the national burden of road fatalities. Despite regulatory mandates, helmet non-compliance remains widespread due to limited enforcement reach and behavioral inertia. Currently there is no intelligent system that enforces helmet usage autonomously in a closed loop directly at the vehicle level. Motivation: The existing enforcement strategies such as traffic policing or external camera-based surveillance are reactive, infrastructure-dependent, and ineffective at scale. To overcome these limitations, we propose a real-time helmet detection and closed-loop system that is integrated into the two-wheeler itself. This approach promotes proactive safety compliance without relying on external monitoring infrastructure. Methodology: The proposed model is AI-powered, vision-based system that leverages deep learning techniques for helmet detection, supported by a custom-built dataset

Kandimalla, Om Mahalakshmi, Shah, Ravindra, Karle, Ujjwala

Lithium-ion Battery’s State of Health Prediction: A Deep Learning Framework

2026-26-0162To be published on 01/16/2026

The growing adoption of electric vehicles (EVs) has increased the need for reliable and efficient battery health monitoring systems. Lithium-ion batteries, widely used in EVs, undergo performance degradation over time, making the prediction of their health status a crucial task for ensuring safety, longevity, and operational efficiency. Traditional model-based approaches often face challenges due to their reliance on complex physical models and sensitivity to varying operational conditions. To overcome these limitations, data-driven methods are increasingly preferred for predictive maintenance. In this study, a data-driven framework is proposed for estimating the State of Health (SoH) of lithium-ion batteries. The framework is developed using data collected from multiple battery cells under diverse operating scenarios. Key steps such as data preprocessing, feature engineering, and model development are carried out to build robust machine learning and deep learning models. A focus is

Suryawanshi, Chaitanya Balasaheb, Nangare, Kapilraj, Gaikwad, Pooja

AI based models as Virtual sensors to replace real sensors

2026-26-0671To be published on 01/16/2026

Artificial Intelligence and Machine learning models have a large scope and application in Automotive embedded system. These models are used in automotive world for various applications like calibration, simulation, predictions etc. These models are generally very accurate and play the role of a virtual sensor. However, the AI/ML models are resource intensive which makes them difficult to execute on a largely optimized automotive embedded systems. The models also need to follow safety standards like ASIL-D. The current work involves creating a Global DoE with ASCMO to generate data from a 125cc single to create AI/ML model for the engine outputs like Torque, T3, Midcat temperatures etc. The created models were validated across the operating space of the engine and found to have good accuracies. With AI Coder, the torque and T3 prediction AI models were converted to embedded code which can be easily used as a virtual sensor in real time. Using these AI models, accurate predictions can be

Chouhan, Vineet Singh, Bulandani, Saurabh, Kumar, Alok, Varsha, Anuroopa, P R, Renjith

Leveraging Advanced AI/ML Algorithms to Derive Actionable Insights from Fleet Vehicle In-Use Performance Ratio Monitoring (IUPRm) Data

2026-26-0637To be published on 01/16/2026

In recent years, growing concerns over environmental pollution have led to the implementation of increasingly stringent emission regulations worldwide. These regulations require not only the reduction of pollutant emissions but also robust monitoring of critical emission control components under real-world driving conditions. The In-Use Performance Ratio Monitoring (IUPRm) framework quantifies how frequently onboard diagnostic systems evaluate these components within defined operational boundaries for each vehicle. IUPRm is organized into several monitoring groups, including catalyst monitoring, oxygen sensor monitoring, exhaust gas recirculation (EGR) monitoring, and gasoline particulate filter monitoring. The specific groups and monitoring requirements vary based on fuel type (gasoline or diesel), engine technologies (such as variable valve timing or EGR), and exhaust system configurations (single or dual bank). For automakers, analyzing these parameters across large vehicle fleets

Ghadge, Ganesh Narayan, Jadhav, Marisha, Hosur, Viswanatha

Generative AI based methodology for Road Profile Mix Creation

2026-26-0674To be published on 01/16/2026

The acquisition of road profile data is crucial for various automotive testing applications, including vehicle dynamics analysis, chassis endurance tests, and simulation of vehicle-road interactions. This is necessary for conducting virtual tests to accelerate research and development processes and can significantly reduce testing costs. However, most of the on-road measurements lack comprehensive and relevant road profile data. Conducting on-road trials to acquire this data is a laborious and time-consuming process, often impeded by logistical and environmental challenges. This research proposes a generative AI-based methodology for creating diverse and realistic road profile mixes from the existing on-road dataset of front axle displacement and road profile measured with a laser sensor. By leveraging advanced machine learning techniques, the proposed approach seeks to generate synthetic road profiles that accurately reflect real-world conditions, thereby reducing the dependency on

Rajappan, Dinesh Kumar, Venkatesh, Anirudh Anand, R, Suresh, N, Gopi Kannan

Vision Driven Testing Framework for Human-Machine Interfaces: An Automated Approach Using Computer Vision / Machine Learning

2026-26-0571To be published on 01/16/2026

Modern automotive infotainment systems and digital instrument clusters require rigorous testing to ensure reliability and compliance with industry standards. Traditional testing methods are labor-intensive and lack scalability. This research proposes a phased, vision-based approach using image processing, machine learning (ML), and computer vision (CV) to enhance testing automation and accuracy. Phase 1 employs open-loop testing using an NVIDIA Jetson Nano with camera modules to capture visual data from infotainment and instrument cluster displays. This phase focuses on the static analysis of visual elements, screen transitions, animations, and error messages. Feature extraction and pattern recognition techniques identify interface behaviors and detect anomalies without a feedback loop. This research will focus in depth on Phase 1, providing a detailed evaluation of its implementation and effectiveness. Phase 2 extends the testing environment with a hardware-in-the-loop (HIL) system

Lad, Rakesh Pramod, Mehrotra, Soumya, Mishra, Arvind

A Data-Centric Approach to Identify Optimal NVH simulation Tools and Solvers

2026-26-0356To be published on 01/16/2026

The world is rapidly evolving towards a data-driven paradigm, with widespread adoption of tools and techniques such as Artificial Intelligence, Machine Learning, Digital Twins, and Cloud Computing. In the automotive sector, vast amounts of data are being generated through both physical and digital test evaluations. Among these, Computer-Aided Engineering (CAE) stands out as one of the largest contributors to data generation, as physical testing is often cost-prohibitive due to the need for prototypes and complex test setups. The field of Automotive Noise, Vibration, and Harshness (NVH) is advancing exponentially, driven by stringent regulatory norms and growing customer demand for comfort. Digitally advanced techniques are playing a pivotal role in revolutionizing NVH processes. Data generation via CAE tools has become an integral part of engineers' daily operations, with the selection of appropriate CAE software and solvers being critical. This choice impacts factors such as user

Hipparge, Vinod, masurkar, Nikita, Arabale, Vinand, Billade, Dayanand

A Machine Learning Approach to Early-Stage Failure Prediction in Traction Motor Manufacturing

2026-26-0281To be published on 01/16/2026

Traction motors technology has, driving the EV industry forward with more efficient, lightweight, and durable solutions. However, despite these advancements, noise testing at the end of the production line remains a critical stage for identifying manufacturing defects in traction motors. Hence early fault detection in traction motors is crucial to ensure safety and reliability of EV. This research contributes a solution that predicts early-fault detection, supporting improved reliability, reduced material cost and minimizing process time in the series production line. To identify the root cause of this problem, historical quality data has been acquired from manufacturing plants to enable efficient analysis. Feature selection was then carried out using embedded and wrapper methods to identify the most important features. These selected features were subsequently used as input for ML models. The best accuracy was achieved using SVC model for early-stage motor failure prediction.

Gaikwad, Pooja, Nangare, Kapilraj, Suryawanshi, Chaitanya

Extreme Learning Machines (ELM) for Fast Instantaneous Prediction of Emission Level of an Automobile

2026-26-0226To be published on 01/16/2026

Addressing climate issues is a key aspect of good global governance today. A key aspect of managing the threats caused to the environment around is to ensure a sustainable transportation system so that humans exist in peace with nature. According to sources, in 2020 alone, cars accounted for approximately 23% of global CO2 emissions. In addition, they also emit nitrogen oxides and other pollutants thus damaging the ecosystem. In line with this objective, there are emission level testing strategies in place in each country. However, we can do better for a sustainable future. On one hand, the huge volume of vehicles around the world makes it an excellent choice and source for a vast emission level dataset comprising of features such as vehicle type, fuel type, driving conditions, road type etc.. as well as the target variable representing the emission band of the vehicle. In addition to the big data available as mentioned above, there is a marked progress in the machine learning

Sridhar, Sriram, Aswani, Shelendra

Machine Learning Approach to ISO Road Class Prediction Using Multiple Linear Regression

2026-26-0667To be published on 01/16/2026

The inertial profiler methodology is traditionally employed in RLDA (Road Load Data Acquisition) to measure road profiles and classify test routes into ISO road classes. However, this approach demands significant time and effort during instrumentation. Also, during data acquisition, laser height sensor data is affected especially during adverse conditions such as rainy seasons or on surfaces with improper reflectivity. Additionally, substantial resources are required for data processing to convert raw measurements into road classifications. To address these challenges, an initial attempt was made to establish a relationship between axle acceleration responses and road profiles, enabling axle acceleration measurements during RLDA to predict ISO road classes. However, this approach relied on a simple linear model that considered only axle acceleration responses, rendering the predictions susceptible to inaccuracies due to varying parameters such as vehicle speed. To overcome these

P, Praveen Kumar, P, Dayalan, Sriramulu, Yoganandam

Emerging Trends in Emission Optimization

2026-26-0229To be published on 01/16/2026

The exhaust emission control is a vital aspect of automotive development, aimed at ensuring effective control of pollutants such as NOx, CO, and HC. The traditional method of calibrating emission control strategies, that requires extensive vehicle testing with both the fresh and aged catalysts under a variety of operating conditions, is a time-consuming process. The frequent changes in emission legislation creates new challenges for achieving a faster time to market and effective utilization of resources. So, the use of automation and machine learning (ML) in the domain of emission control is the need of the hour to proactively tackle these challenges. This paper attempts to explore emerging trends in emission optimization, such as the automated measurement process and ML/AI based data analysis, that can improve the overall process efficiency in the calibration of emission control strategies. The automated measurement process is achieved using the automation software. The data

Dhayanidhi, Hukumdeen, Balasubramanian, Karthick, A, Akash

Artificial Intelligence in Passenger Cars: Unlocking Intelligent Mobility and Human-Centric Vehicle Innovation

2026-26-0672To be published on 01/16/2026

Artificial Intelligence (AI) is fundamentally reshaping the landscape of automotive engineering, particularly in passenger vehicles where comfort, safety, and customization are paramount. This paper presents a comprehensive review of current and emerging AI-driven applications in passenger cars, including intelligent vehicle insurance, real-time traffic optimization, personalized driving experiences, emotion recognition, and advanced EV battery management. The integration of AI technologies—ranging from machine learning and computer vision to natural language processing and deep neural networks—enables predictive maintenance, adaptive driver-assist systems, and next-generation human-machine interfaces. AI’s role in vehicle insurance is also explored through behaviour-based risk scoring and claims automation. Furthermore, emotion-aware cabin systems are discussed, which aim to detect driver fatigue, stress, or distraction to improve road safety and mental wellness. The paper also

Hazra, Sandip, Tangadpalliwar, Sonali, Khan, Arkadip

Transforming Gear Manufacturing_ A Deep Learning Approach to Quality Prediction

2026-26-0646To be published on 01/16/2026

The automotive industry is rapidly transitioning towards Industry 4.0, transforming vehicle manufacturing. To achieve a lower carbon footprint, it is crucial to minimize raw material wastage and energy consumption. Reducing component wastage, lead time, and automating gear manufacturing are key areas. Gear micro-geometry inspection is vital, as variations affect service life and NVH (Noise, Vibration, Harshness). Despite standards for permissible errors, manual evaluation is often needed.This subjective evaluation approach will have a possibility that a gear with undesired variations gets assembled into the product. These issues can be detected during NVH testing, leading to replacement of part and re-assembly thus increasing lead time. This generates a need for an automated system which could reduce the human intervention and perform the activity. This paper discusses methodologies to meet these requirements and achieve desired results. The research aims to develop a deep learning

Ramakrishnan, Gowtham Raj, Baheti, Palash, PR, Vaidyanathan, Durgude, Ranjit, Bathla, Archana, R, Greeshmita, V, Rangarajan

Real-Time In-Cabin Vital Sensing and Emergency Response Using Ultra-Wideband Radar in Automotive Environments

2026-26-0012To be published on 01/16/2026

This study introduces a novel in-cabin health monitoring system leveraging Ultra-Wideband (UWB) radar technology for real-time, contactless detection of occupants' vital signs within automotive environments. By capturing micro-movements associated with cardiac and respiratory activities, the system enables continuous monitoring without physical contact, addressing the need for unobtrusive vehicle health assessment. The system architecture integrates edge computing capabilities within the vehicle's head unit, facilitating immediate data processing and reducing latency. Processed data is securely transmitted via HTTPS to a cloud-based backend through an API Gateway, which orchestrates data validation and routing to a machine learning pipeline. This pipeline employs supervised classifiers—Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF)—to analyze features such as temporal heartbeat variability, respiration rate stability, and energy distribution patterns

Singh, Samagra, Pandya, Kavita, Jituri, Keerti

Artificial Neural Network for predicting thermal cool-down behavior of a School Bus Cabin

2026-26-0653To be published on 01/16/2026

Passenger vehicles, such as buses, exhibit significant heat absorption when parked under direct sunlight, leading to elevated interior temperatures during daytime and a consequent reduction in thermal comfort. In our previous study, we developed and validated a transient numerical method to predict the thermal behavior of the cool-down process in a school bus cabin. Despite its accuracy, modeling such complex simulations using Computational Fluid Dynamics (CFD) remains time-intensive, with pre-processing taking up to a week and simulations requiring an additional 8-10 hours. While physics-based simulations are known for their accuracy, they are inherently slower. Conversely, data-based meta models offer faster results, but their outputs must be carefully correlated with test data or physics-based model data. This study aims to leverage existing CFD data to train and develop an Artificial Neural Network (ANN) based meta model, which is then benchmarked against test results for

Suryavanshi, Apoorva, sharma, shantanu

Reverse Engineering and Digital Twins in Powertrain Design: Enhancing Performance and Sustainability

2026-26-0065To be published on 01/16/2026

This study explores the application of reverse engineering (RE) and digital twin (DT) technology in the design and optimization of advanced powertrain systems. Traditional approaches to powertrain development often rely on legacy designs with limited adaptability to modern efficiency and emission standards. In this work, we present a methodology combining 3D scanning, computational modeling, and machine learning to reconstruct, analyze, and enhance internal combustion engines (ICEs) and electric vehicle (EV) drivetrains. By digitizing physical components through RE, we generate high-fidelity DT models that enable virtual testing, performance prediction, and iterative improvement without costly physical prototyping. Key innovations include a novel mesh refinement technique for scanned geometries and a hybrid simulation framework integrating finite element analysis (FEA) and multi-body dynamics (MBD). Our case study demonstrates a 12% increase in thermal efficiency for a retrofitted ICE

Bernikov, Mark Alexandrovich, Kurmaev, Rinat

NN FMU: Deep Learning Models for Next-Gen xiL Vehicle Simulation

2026-26-0452To be published on 01/16/2026

Functional Mock-up Units (FMUs) have become a standard for enabling co-simulation and model exchange in vehicle system development. However, traditional FMUs derived from physics-based models can be computationally intensive, especially in scenarios requiring real-time performance. This paper presents a Python-based approach for developing a Neural Network (NN) based FMU using deep learning techniques, aimed at accelerating vehicle simulation while ensuring high fidelity. The neural network was trained on vehicle simulation data and developed using Python frameworks such as TensorFlow. The trained model was then exported into FMU, enabling seamless integration with FMI-compliant platforms. The NN FMU replicates thermal behavior of vehicle with high accuracy while offering a significant reduction in computational load. Benchmark comparisons with Physical Thermal Model demonstrate that the proposed solution provides both efficiency and reliability across various driving conditions

Srinivasan, Rangarajan, ASHOK BHARDE, Pooja, MHETRAS, Mayur, CHEHIRE, Marc

Accelerating engine auxiliaries’ development by deploying ML techniques in CFD virtual verification

2026-26-0374To be published on 01/16/2026

The automotive and off-road industries are heavily investing in research and development (R&D) to improve both physical and virtual verification and validation techniques. Recent software and hardware advancements have extended these techniques from simple component evaluations to complex system assessments such as involving multi-physics scenarios. Despite the benefits of virtual validation tools like structural analysis and computational fluid dynamics (CFD), they often come with high development costs, particularly in CFD applications. Virtual verification methodology, especially when combined with data science, offer significant advantages over traditional physical methods by enhancing computer-aided engineering (CAE) efficiency and reducing resource consumption which can greatly improve product design and validation efficiency across many industries. The success of machine learning applications depends on effective data processing, adequate computational resources, and the right

Jadhav, Mitali, Kumbhar, Appaso, Tirumala, Bhaskar, Nisha, Kumari

Automated Parameterization of CAE Model Based on Test Data Response Bands

2026-26-0436To be published on 01/16/2026

This study presents a data-driven approach aimed at enhancing the correlation between physical test data and Computer-Aided Engineering (CAE) simulations, with an emphasis on adapting the standard CAE model's response to minimize any gaps relative to the response of a given test specimen. Leveraging historical test data, machine learning techniques are used to categorize responses into distinct bands, effectively capturing the inherent variability observed in real-world scenarios. This categorization step recognizes patterns across a wide range of test data, forming the foundation for closely matching and adapting CAE models to new, unseen hardware data. In typical automotive simulation workflows, tuning a standard CAE model to match new hardware test data involves iterative parameter adjustments and simulations. This process can be time-consuming and often lacks predictive insight into the necessary modifications. The approach developed in this study addresses this challenge by

Khopekar, Maria, Arya, Bibhu, Sridhar, Raam, Mohan, Pradeep, Kurkuri, Mahendra

Multi-Model Analysis of O-Ring Compression and Relaxation Force Predictions Using Machine Learning and CAE Simulations

2026-26-0466To be published on 01/16/2026

O-rings play a critical role in ensuring leak-proof seals in a wide range of engineering systems. Accurate prediction of their compression and relaxation behavior under various material and geometric configurations is essential for optimal design and reliability. This study presents an analysis of machine learning techniques to predict two key performance outputs, compression force and relaxation force (after 10 minutes) trained on computer-aided engineering (CAE) simulation data. The experimental setup was represented in CAE simulation and the results were compared with experimental data conducted at ZF test facilities. Simulation results correlated well with the experimental data (deviation was less than the 5%). To create a dataset for training machine learning (ML) models, realistic ranges for the input parameters such as hardness and geometrical parameters were determined, and simulation data were generated using design of experiments (DOE). Multiple ML models were developed and

Kosgi, Durgaprasad, Alva, P Pancham, Dangeti, VenkataKrishna Pavan

Machine learning model to predict Chest Injury during Thorax Impact on Human Body Models

2026-26-0665To be published on 01/16/2026

In recent years, virtual models have been handy in predicting potential injury risk to occupants in vehicle crashes. Virtual models offer detailing of occupant anthropometry and closest possible bio-fidelity over existing test devices. This study focuses on the assessment of chest deflections in frontal thorax impacts using virtual human body models of a few anthropometries and transforming the assessment of injuries of broader range of anthropometries (population). The study utilizes machine learning models to enable injury assessment across a broader range of body types. The study has taken a standard test scenario (Kroell load case) having frontal blunt thoracic impact. The load case is replicated in LS-DYNA and finite element human body models (HBMC) of 05th Female and 50th Male are used. The simulation setup replicates the boundary conditions as per literature, the responses necessary to measure chest deflection were validated & recorded. The outputs from the simulations and data

Sridhar, Raam, Arya, Bibhu, Divakar, Prajwal, R, Udhaya Kumar, Bhutki, Prasad, Kumar PhD, Devendra, Kurkuri, Mahendra, Mohan PhD, Pradeep

Video 2 Scenario: A Novel Method to Generate Edge Case Traffic Scenarios from Dashcam Videos

2026-26-0666To be published on 01/16/2026

This paper presents Video 2 Scenario, a systematic and cost-effective method to generate edge case scenarios derived from road events captured in dash camera videos. Current methods for obtaining real-world scenarios rely on supplementary sensor data like LiDAR, GPS and radar, which require expensive sensor setups for data collection. Other scenario generation methods rely on manual effort by human experts, which can be laborious and time-consuming. Our proposed methodology utilizes state-of-the-art deep learning models to extract vehicle trajectories, road layouts and other scenario-specific elements and reconstructs a scenario in the format of simulation scripts. The scripts are then refined by replaying the scenario multiple times in a simulation with attached virtual sensors to minimize the discrepancy between the subsequently obtained simulated video data and the original dash camera video data. The refined simulation scripts ensure that critical edge case scenarios are replicated

Ramalingam, Surya Prasath, Priyadarshi, Marut, Khan, Parvej, Kumar, Vaibhav, Pandey, Gaurav, Lohani, Bharat

Self-supervised diagnostics & prognostics framework for predictive maintenance in electric vehicles

2026-26-0673To be published on 01/16/2026

Predictive maintenance is essential for enhancing the reliability, safety, and operational efficiency of connected vehicles. However, traditional supervised learning approaches for fault prediction depend heavily on large, labeled datasets of failure events, which are often scarce and expensive to obtain in real-world automotive environments. This paper proposes an original self-supervised anomaly detection framework that predicts emerging failures using only healthy operational data, eliminating the need for explicit failure labels. The approach employs self-supervised pretext tasks, such as masked signal reconstruction and future telemetry prediction, to learn robust representations of normal multi-sensor behavior (e.g., temperature, pressure, current, vibration). An unsupervised anomaly detection model then identifies deviations from these learned patterns. This integrated with data-informed predictive models enable early detection of faults across critical subsystems such as

Kumar, Pankaj, Deole, Kaushik, Hivarkar, Umesh

Transforming Headlamp Levelling Compliance Testing Through AI-Powered Predictive Modelling

2026-26-0660To be published on 01/16/2026

The process of validating headlamp levelling compliance as per regulatory standards involves physical testing with various vehicle loading conditions. This traditional method is labour-intensive, time-consuming, and consumes significant resources. There is a need for a predictive solution that can simulate and validate headlamp levelling tests virtually, thereby reducing dependency on physical trials. Headlamp levelling compliance is a critical regulatory requirement to ensure optimal visibility and safety under varying vehicle loading conditions. This paper presents an Artificial Intelligence and machine learning-based (AI / ML) solution to simulate headlamp levelling tests virtually/Digitally by using Historical Headlamp designs from past vehicles. By leveraging historical test data and developing regression models, the system predicts headlamp height and dipped beam height for both unladen and laden conditions. Key vehicle parameters such as tire load, overhang, deflection, and

Mandloi, Prince, Joshi, Vivek S., GHANWAT, HEMANT, Ugale, Anand, Munda, Rohit, GHAN, PRAVIN

Digital Twin for Motor & MCU incorporating Deep learning AI-ML algorithms for intelligent prognostics for Electric vehicles

2026-26-0479To be published on 01/16/2026

As electric vehicles (EVs) become more advanced, so ensuring the reliability of critical components like the motor and Motor Control Unit (MCU) is essential. This paper presents a digital twin model designed to predict failures in motor and MCU components using machine learning. The approach focuses on detecting early signs of failure through real-world data and advanced analytics. We collected thermal and performance data from field vehicles, capturing both normal (healthy) and abnormal (faulty) operating conditions. Using this dataset, we developed and trained an Auto Encoder-based machine learning model that learns what “normal” looks like and flags deviations as potential issues. One key outcome of this study is the successful early prediction of Insulated Gate Bipolar Transistor (IGBT) degradation, where the system identified subtle behavioral changes long before any visible failure symptoms appeared. This digital twin acts as a virtual replica of the physical components

Joshi, Pawan, Pandey, Suchit, KONDHARE, Manish, UPADHYAY, ABHAY, JaganMoahanarao, Vana, Tank, Prabhu

Tractor Cabin Structure Borne Noise Reduction in the Virtual Environment

2026-26-0101To be published on 01/16/2026

In recent days, cabin variants in the tractor are preferred by the farmers for the Coziness and longer field hour operation with less fatigue. Noise perceived by customer is the most important factor taken into account during the design stage, as it’s directly linked with operator’s comfort. Observed noise levels has to be within the defined limits as per national/international standards Overall cabin noise levels is contributed by the structure borne noise below 630 Hz. Structure borne noise is the noise typically radiated by the door, roof, windshield, floor, fender and structure assembly due to the engine excitation through the transmission housings and backstories. This paper depicts the process of tractor cabin structure borne noise prediction in the virtual environment. Firstly, Engine bearing loads and axle bearings has been extracted in the virtual stage from the vehicle level driveline model using commercially available MBD software. The finite element (FE) model of the cabin

gunasekaran, pandiyanayagam, K, Somasundaram, Chavan, Amit

A Deep Learning Approach for Dynamic Range Estimation in Electric Fleet Vehicles.

2026-26-0675To be published on 01/16/2026

Keywords: Real-Time Range Prediction, Long Short-Term Memory (LSTM), Time Series Modelling, Deep Learning, Vehicular Telemetry, CAN Data, Range Anxiety Accurate and real-time range estimation is a critical enabler for efficient energy management and operational planning in Commercial Electric Vehicles (CVEVs). This project presents a robust deep learning-driven framework for vehicle range prediction, leveraging a Long Short-Term Memory (LSTM) network trained on high-frequency powertrain CAN data obtained from CVEVs operating across diverse driving environments. The dataset comprises key dynamic and environmental parameters—including State of Charge (SOC), vehicle speed, acceleration, and temperature—captured during real-world operations. Extensive data pre-processing techniques were applied to ensure signal integrity, including high-resolution timestamp synchronization, noise filtering, and imputation of missing values, thereby establishing a reliable foundation for time-series

surlekar, Samarth, Ghorpade, Swapnali, Kalghatgi, Bhakti

Data-Driven Prediction of Global Automotive Trends: Forecasting Fuel Economy and CO₂ Emissions Using Machine Learning

2026-26-0643To be published on 01/16/2026

In the pursuit of cleaner transportation and environmental sustainability, the global automotive industry is undergoing rapid transformation. This study leverages historical multi-decade vehicle data to develop predictive models for fuel economy and CO₂ emissions across a wide range of vehicle technologies. By utilizing advanced machine learning algorithms in Python and integrating insights through Power BI visualizations, the project identifies key correlations between vehicle attributes—such as weight, powertrain, and footprint—and their environmental performance. Results highlight the increasing impact of electric vehicle adoption, hybridization, and light weighting on overall emissions reduction. These insights help forecast the direction of fuel economy standards, emission patterns, and technology shifts across manufacturers and vehicle types. Beyond technical predictions, the study offers a decision-support framework for global policymakers, automotive designers, and

Hazra, Sandip, Tangadpalliwar, Sonali, Hazra, Sanjana

Conceptual Framework for Real-Time Battery Health Estimation in Automotive Digital Twins Using Reinforcement Learning

2026-26-0658To be published on 01/16/2026

The explosive growth of electric vehicles (EVs) calls forth the need for smart battery management systems that can perform health monitoring and predictive diagnostics in real-time. The conventional battery modeling methods mostly do not cover the complicated, dynamic behaviors coming from different usage patterns. The study outlines a structure that would use Reinforcement Learning (RL)-based AI agent as a part of the Battery Electrical Analogy (BEA) simulation platform. With the help of the AI agent, different health parameters such as State of Health (SOH), State of Charge (SOC), and the signs of early thermal runaway can be predicted in real-time. The suggested design takes advantage of the simulation-based approach to have the agent learn and utilizes a decentralized cloud architecture suitable for scaling and reducing the response time. The RL agent performs an essential role in the process by tagging along with the continuous learning and the adjustment of the battery conditions

Pardeshi, Rutuja Rahul, KONDHARE, MANISH, Sasi Kiran, Talabhaktula

Algorithm development for AI driven Path planning using Augmented reality for navigating in Indian complex road scenarios

2026-26-0038To be published on 01/16/2026

The road infrastructure in India has complex navigational challenges, with 82% roads being unstructured according to Ministry of Roads Transports and Highways (MoRTH). Decision-making becomes a challenge for drivers in unpredictable environments such as narrow roads or sometimes flooded roads and heavy traffic. In this paper, an augmented reality (AR) driver assistance algorithm has been proposed that improves awareness to safely maneuver in these conditions.The methodology for development and validation of this AR algorithm contains multiple steps. Firstly, extensive data collection is conducted using real time capture and benchmark datasets like Berkeley Deep Drive (BDD) and Indian Driving Dataset (IDD). Secondly, A machine learning model consisting of the collected data are labelled and trained to rightly identify the complex scenario. In third step, an AR navigation algorithm is developed, integrating these models to accurately detect floods and estimate road widths and provide

Anandaraj, Prem Raj, Sivakumar, Vishnu, Thanikachalam, Ganesh, L, Radhakrishnan, Motoki, Yaginuma, Selvam, Dinesh Kumar

Optimizing System Design performance and its accuracy using Feature Importance

2026-26-0421To be published on 01/16/2026

With continuous improvements in model accuracy and computational performance, system simulation can be seamless integrated into development tools and methodologies as a "virtual test bed" in synergy with experimental tests. However, virtual vehicle and powertrain thermal models still face significant obstacles in realizing their full benefits. One common challenge is modeling accuracy and its anomalies, particularly in models provided by internal automotive companies and external suppliers. The models developed for vehicle thermal management and its systems have been created using specific software, making it difficult to verify the accuracy of the models across different systems and their performance. Integrating different systems in a Co-simulation environment has affected the overall accuracy of the Vehicle Simulation Platform model. This situation significantly reduces the return on investment in model development. To address these issues, we propose deploying feature importance

Srinivasan, Rangarajan, SARAPALLI RAMACHANDRAN, Raghuveeran, ASHOK BHARDE, Pooja, SARAVANAN, Vivek

Items per page:

1 – 50 of 549