Krisztián Horváth

59224275100

Publications - 10

Data-Driven Predictive Modeling for Investigating the Impact of Gear Manufacturing Parameters on Noise Levels in Electric Vehicle Drivetrains

Krisztián Horváth

Publication Name: World Electric Vehicle Journal

Publication Date: 2025-08-01

Volume: 16

Issue: 8

Page Range: Unknown

Description:

Reducing gear noise in electric vehicle (EV) drivetrains is crucial due to the absence of internal combustion engine noise, making even minor acoustic disturbances noticeable. Manufacturing parameters significantly influence gear-generated noise, yet traditional analytical methods often fail to predict these complex relationships accurately. This research addresses this gap by introducing a data-driven approach using machine learning (ML) to predict gear noise levels from manufacturing and sensor-derived data. The presented methodology encompasses systematic data collection from various production stages—including soft and hard machining, heat treatment, honing, rolling tests, and end-of-line (EOL) acoustic measurements. Predictive models employing Random Forest, Gradient Boosting (XGBoost), and Neural Network algorithms were developed and compared to traditional statistical approaches. The analysis identified critical manufacturing parameters, such as surface waviness, profile errors, and tooth geometry deviations, significantly influencing noise generation. Advanced ML models, specifically Random Forest, XGBoost, and deep neural networks, demonstrated superior prediction accuracy, providing early-stage identification of gear units likely to exceed acceptable noise thresholds. Integrating these data-driven models into manufacturing processes enables early detection of potential noise issues, reduces quality assurance costs, and supports sustainable manufacturing by minimizing prototype production and resource consumption. This research enhances the understanding of gear noise formation and offers practical solutions for real-time quality assurance.

Open Access: Yes

DOI: 10.3390/wevj16080426

The Impact of Pitch Error on the Dynamics and Transmission Error of Gear Drives

Dániel Feszty Krisztián Horváth

Publication Name: Applied Sciences Switzerland

Publication Date: 2025-07-01

Volume: 15

Issue: 14

Page Range: Unknown

Description:

Gear whine noise is governed not only by intentional microgeometry modifications but also by unavoidable pitch (indexing) deviation. This study presents a workflow that couples a tooth-resolved surface scan with a calibrated pitch-deviation table, both imported into a multibody dynamics (MBD) model built in MSC Adams View. Three operating scenarios were evaluated—ideal geometry, measured microgeometry without pitch error, and measured microgeometry with pitch error—at a nominal speed of 1000 r min−1. Time domain analysis shows that integrating the pitch table increases the mean transmission error (TE) by almost an order of magnitude and introduces a distinct 16.66 Hz shaft order tone. When the measured tooth topologies are added, peak-to-peak TE nearly doubles, revealing a non-linear interaction between spacing deviation and local flank shape. Frequency domain results reproduce the expected mesh-frequency side bands, validating the mapping of the pitch table into the solver. The combined method therefore provides a more faithful digital twin for predicting tonal noise and demonstrates why indexing tolerances must be considered alongside profile relief during gear design optimization.

Open Access: Yes

DOI: 10.3390/app15147851

Using Machine Learning Models to Predict and Reduce Noise Levels in Gear Systems

Ambrus Zelei Krisztián Horváth

Publication Name: Advances in Science and Technology

Publication Date: 2025-01-01

Volume: 165 AST

Issue: Unknown

Page Range: 215-221

Description:

Machine learning models are effective tools for predicting and reducing noise levels in industrial gear systems. In this study, we compare different machine learning methods to investigate the effects of different gear modification parameters on noise levels. Four different predictive models was used. Random Forest Regressor, XGBoost, Gradient Boosting Machines and neural network. The study concluded that Random Forest and Gradient Boosting Machines models were the most effective. Both models achieved low mean squared error values 6.10 and 6.67. Further tests with synthetic data confirmed the stability of these models. Current sustainability trends show that the integration of machine learning into industrial applications fits well with manufacturers' objectives. However, it is currently challenging to determine which machine learning methods are most effective in optimizing noise reduction. This paper seeks to address this gap by comparing the accuracy and reliability of these models. Based on the results, the use of machine learning models is recommended to reduce noise levels in geared systems.

Open Access: Yes

DOI: 10.4028/p-0GDArj

Development of a multibody model for go-karts considering frame flexibility

Ambrus Zelei Krisztián Horváth

Publication Name: Pollack Periodica

Publication Date: 2024-10-16

Volume: 19

Issue: 3

Page Range: 66-73

Description:

This study focuses on the optimization dynamics of racing go-karts, which is heavily influenced by the frame's stiffness. Lacking suspensions and differentials, go-karts rely on the frame stiffness for wheel balancing and skid prevention by lifting the inner rear wheel during turns. Utilizing a rigid-flexible model in MSC Software ADAMS View, validated by frame deformation measurements, this research integrates finite element analysis with multibody techniques. The model, leverages computer aided design files for frame geometry and employs finite element analysis for frame validation. It facilitates evaluating go-kart dynamics through simulations, aiding in maneuver testing and design optimization. This approach provides a comprehensive framework for advancing go-kart designs.

Open Access: Yes

DOI: 10.1556/606.2024.01050

Simulating Noise, Vibration, and Harshness Advances in Electric Vehicle Powertrains: Strategies and Challenges

Ambrus Zelei Krisztián Horváth

Publication Name: World Electric Vehicle Journal

Publication Date: 2024-08-01

Volume: 15

Issue: 8

Page Range: Unknown

Description:

This study examines the management of noise, vibration, and harshness (NVH) in electric vehicle (EV) powertrains, considering the challenges of the automotive industry’s transition to electric drivetrains. The growing popularity of electric vehicles brings new NVH challenges as the lack of internal combustion engine noise makes drivetrain noise more prominent. The key to managing NVH in electric vehicle powertrains is understanding the noise from electric motors, inverters, and gear systems. Noise from electric motors, mainly resulting from electromagnetic forces and high-frequency noise generated by inverters, significantly impacts overall NVH performance. This article details sources of mechanical noise and vibration, including gear defects in gear systems and shaft imbalances. The methods presented in the publication include simulation and modeling techniques that help identify and solve NVH difficulties. Tools like multi-body dynamics, the finite element method, and multi-domain simulation are crucial for understanding the dynamic behavior of complex systems. With the support of simulations, engineers can predict noise and vibration challenges and develop effective solutions during the design phase. This study emphasizes the importance of a system-level approach in NVH management, where the entire drivetrain is modeled and analyzed together, not just individual components.

Open Access: Yes

DOI: 10.3390/wevj15080367

Noise Reduction Methods in the Vehicle Industry: Using Vibroacoustic Simulation for Sustainability

Ambrus Zelei Krisztián Horváth

Publication Name: Chemical Engineering Transactions

Publication Date: 2024-01-01

Volume: 114

Issue: Unknown

Page Range: 763-768

Description:

To achieve sustainability goals, such as greenhouse gas emissions and environmental noise reduction, continuous innovation plays a key role in the vehicle industry. The noise emitted by vehicles negatively impacts both the environment and public health, making the development of noise reduction strategies crucial. Vibroacoustic simulation methodologies provide an opportunity to optimise vehicle power transmission systems by reducing the emitted noise level. Besides, the energy efficiency and performance of the vehicles can be improved by vibroacoustic simulations. In this research, a vibroacoustic simulation methodology is presented, focusing on the power transmission systems of vehicles. This approach integrates the Finite Element Method and Multibody Dynamics simulations with vibroacoustics to identify and redesign noisy components even during the conceptual design stage. This approach tackles the challenge of high-frequency tonal noise for electric vehicles, using psychoacoustic reviews to enhance passenger comfort. Key tasks involve electromagnetic force analysis in the drivetrain, structural vibration simulations, and noise reduction strategy optimisation using machine learning algorithms to reduce the reliance on physical prototypes. Capturing the current momentum of the industry, machine learning capabilities in vibroacoustic models can help engineers identify sources and eliminate or mitigate noise in the early design phase. Reducing the number of prototypes leads to more sustainable design processes. Our study shows the noise level can be reduced by 3-5 dB. This is particularly important in the context of electric vehicles, where high-frequency tone noise should be reduced, benefiting both passengers and their environment. Improving these factors is in line with the goals of the United Nations and improves the quality of urban life. Our research highlights the importance of vibroacoustic simulation and opens new directions in the field of noise reduction, promoting the spread of sustainable transport solutions.

Open Access: Yes

DOI: 10.3303/CET24114128

Gearbox Fault Diagnosis Using Industrial Machine Learning Techniques †

Ambrus Zelei Krisztián Horváth

Publication Name: Engineering Proceedings

Publication Date: 2024-01-01

Volume: 79

Issue: 1

Page Range: Unknown

Description:

This paper highlights the need for precise and reliable diagnostic methods for early fault detection in gearbox systems, something critical for industrial maintenance. Advances in machine learning (ML) and image processing have opened new avenues for diagnosis. This study explores ML techniques, particularly edge detection and maximized pooling, with the Inverse Distance Weighting method, for diagnosing gearbox faults from vibration signal images. Using the ODYSSEE-A Eye platform, a model was developed that achieved 96% accuracy in identifying faults from a 500-sample dataset. The research results promote further investigation and progress in this area, indicating specific possible directions for further research.

Open Access: Yes

DOI: 10.3390/engproc2024079036

Literature Review of Vibroacoustic Simulation in Geared Vehicle Power Transmission Systems for the Reduction of Radiated Noise

Ambrus Zelei Krisztián Horváth

Publication Name: Advances in Science and Technology

Publication Date: 2024-01-01

Volume: 153

Issue: Unknown

Page Range: 89-98

Description:

The radiated noise reduction of vehicular power transmission systems is one of the most actively researched areas. Noise not only impacts the comfort and safety of the driver and passengers but also regulated by the legislators. The simulation-based prediction of radiated noise of gear-drives is a rapidly evolving area and combines gear meshing models, finite element analysis, multibody dynamics and airborne noise simulation tools. The interfacing of these tools makes virtual noise prediction challenging. In this research, we conducted a literature review on vibroacoustic simulations, with a particular focus on reducing noise in power transmission systems. Based on the reviewed articles, it became evident that, although numerous measurement data are available, the usability of the data is limited. Most research focuses on individual stages of the structure and on smaller-sized powertrains. The measurement methods contain abundant valuable information; however, the literature lack of comprehensive articles that track the simulation process from the inception of excitation to body and air noises. Moreover, the majority of articles investigate the relationship between transmission error and NVH, considering it as a primary source of noise. New methodological approaches, such as the application of FEM meshes on gears, open new horizons in this domain. Throughout the literature review, we compiled potential noise-reduction solutions and highlighted directions for future methodology development research.

Open Access: Yes

DOI: 10.4028/p-Ucpx27

Predicting Natural Frequencies of a Cantilever Using Machine Learning

Ambrus Zelei Krisztián Horváth

Publication Name: Advances in Transdisciplinary Engineering

Publication Date: 2024-01-01

Volume: 59

Issue: Unknown

Page Range: 105-111

Description:

In the context of structural analysis and design, natural frequencies play a vital role, and their prediction is essential in machine and vehicle design processes. The simulations related to the modal parameters are computationally intensive for systems with large complexity. This paper demonstrates on an illustrative academic example that natural frequencies can be successfully predicted using ML models. This paper aims to develop a model based on machine learning (ML) to predict a simple cantilever's natural frequencies based on the physical parameters of the beam. The independent variables X are the geometric parameters including width, length, and thickness, while the dependent variable Y is the natural frequency. The study is framed using a systematic methodology that covers the stages of data collection, ML model selection, model training and validation. The validation process proves the effectiveness of ML as a computationally cheap replacement for traditional methods of prediction. The current research contributes to the investigation of the usage of commercially available ML tools in structural engineering. We report that the ODYSSEE A-Eye software is capable of natural frequency prediction with a varying geometry structure with less than 4% error for an 80-member training set of cantilever beam with various dimensions. Further developments will include considerations of noise, vibration, and harshness (NVH) to enhance system performance and improve user comfort.

Open Access: Yes

DOI: 10.3233/ATDE240533

Surface Waviness of EV Gears and NVH Effects—A Comprehensive Review

Dániel Feszty Krisztián Horváth

Publication Name: World Electric Vehicle Journal

Publication Date: 2025-09-01

Volume: 16

Issue: 9

Page Range: Unknown

Description:

Electric vehicle (EV) drivetrains operate at high rotational speeds, which makes the noise, vibration, and harshness (NVH) performance of gear transmissions a critical design factor. Without the masking effect of an internal combustion engine, gear whine can become a prominent source of passenger discomfort. This paper provides the first comprehensive review focused specifically on gear tooth surface waviness, a subtle manufacturing-induced deviation that can excite tonal noise. Periodic, micron-scale undulations caused by finishing processes such as grinding may generate non-meshing frequency “ghost orders,” leading to tonal complaints even in high-quality gears. The article compares finishing technologies including honing and superfinishing, showing their influence on waviness and acoustic behavior. It also summarizes modern waviness detection techniques, from single-flank rolling tests to optical scanning systems, and highlights data-driven predictive approaches using machine learning. Industrial case studies illustrate the practical challenges of managing waviness, while recent proposals such as controlled surface texturing are also discussed. The review identifies gaps in current research: (i) the lack of standardized waviness metrics for consistent comparison across studies; (ii) the limited validation of digital twin approaches against measured data; and (iii) the insufficient integration of machine learning with physics-based models. Addressing these gaps will be essential for linking surface finish specifications with NVH performance, reducing development costs, and improving passenger comfort in EV transmissions.

Open Access: Yes

DOI: 10.3390/wevj16090540