Person:
AKGÜN, GAZİ

Profile Picture

Email Address

Birth Date

Research Projects

Organizational Units

OrgUnit Logo
Organizational Unit

Job Title

Last Name

AKGÜN

First Name

GAZİ

Name

Filters

20232
Reset filters

Settings

Author: AKÜNER, MUSTAFA CANER × Subject: Bilgisayar Bilimleri ×

Search Results

Now showing 1 - 2 of 2
  • Thumbnail Image
    PublicationOpen Access
    An innovative approach to electrical motor geometry generation using machine learning and image processing techniques
    (2023-01-01) DEMİR, UĞUR; AKGÜN, GAZİ; AKÜNER, MUSTAFA CANER; AKGÜN, ÖMER; DEMİR U., AKGÜN G., AKÜNER M. C., Pourkarimi M., AKGÜN Ö., Akıncı T. Ç.
    This paper presents a methodology for generating geometries for interior permanent magnet (IPM) motors in electric vehicles (EVs) through the application of artificial intelligence (AI) and image processing (IP) techniques. Due to the implementation of green agreements and policies aimed at reducing greenhouse gas emissions, EVs have become popularity. As a consequence, the improvement studies on the powertrain and battery system of EVs are focused. Especially for the powertrain, design optimization studies of e-motor have increased in the literature. One of the most widely used e-motor topologies is interior permanent magnet (IPM) motor. However, designing the IPM motor presents a challenge due to the dynamic considerations with geometric constraints. Therefore, e-motor designers encounter challenges related to determining initial geometry and the long time of the optimization process. To address these challenges, a novel approach is proposed that leverages machine learning (ML) techniques in combination with IP to generate initial geometries and design parameters for IPM motors. The proposed approach generates images of the motor geometry and extract dimensional features from the resulting images by using artificial neural networks (ANNs). The proposed method performs an analysis of the input vectors to reduce their size using techniques such as Histogram, 2D Maximum, 2D Mean, 2D Minimum, 2D Standard Deviation, and 2D Variance to enhance feature extraction. Additionally, FFT (Fast Fourier Transform) and IFFT (Inverse Fast Fourier Transform) are used to improve the neural network process in generating the image geometry. Further, the generated image geometry is improved by applying digital filtering techniques such as Log, FFT, Log+FFT, Laplacian, Sobel, and Histogram Equalization. Finally, the trained ANNs are tested to validate the results by using Ansys RMXprt and Maxwell. Eventually, the proposed method represents an innovative solution to generating initial geometries for IPM motors in EVs that satisfies desired design requirements. This approach leverages the power of AI and image processing techniques, which could lead to significant improvements in the optimization process for IPM motors, accelerate the designer’s analysis process, and enhance the performance of EVs.
  • Thumbnail Image
    PublicationOpen Access
    Feature extraction and NN-based enhanced test maneuver deployment for 2 DoF vehicle simulator
    (2023-01-01) DEMİR, UĞUR; AKGÜN, GAZİ; AKÜNER, MUSTAFA CANER; AKGÜN, ÖMER; DEMİR U., AKGÜN G., AKÜNER M. C., Demirci B., AKGÜN Ö., Akıncı T. Ç.
    This paper presents a deployment method of various test maneuver scenarios for 2 degree of freedom (2 DoF) vehicle simulator by using feature extraction and neural networks (NN). A prototype version has been set up for the 2 DoF vehicle simulator. Then, a hardware in the loop (HIL) model with 2 inputs (torque, τ1-τ2) and 3 outputs (acceleration, ax-ay-az) is created. System identification is performed to obtain the training data of NNs to be used for the deployment of test maneuvers. In the system identification process, 2 arbitrary sinusoidal torque signals (τ1-τ2) are generated by using the actuator specs of the 2 DoF vehicle simulator. By applying the generated torque signals to the actuators, acceleration (ax-ay-az) data are collected from the inertial measurement sensor (IMU) on the 2 DoF vehicle simulator. It is determined to create 3 different NN models for the obtained data. The 1st NN model is trained with 3 inputs (ax-ay-az) and 2 targets (τ1-τ2) training data. The 2nd NN model is trained with 6 inputs (amplitudes and phases of ax-ay-az) and 2 targets (τ1-τ2) training data. The input data features for the 2nd NN model is extracted by using the Fast Fourier Transform (FFT). The 3rd NN model is trained with 6 inputs (amplitudes and phases of ax-ay-az) and 4 targets (amplitudes and phases of τ1-τ2) training data. For the 3rd NN model, the features of input and target data are extracted by using the FFT. The NN training process continues until acceptable performance criteria are reached. Then, 3 NN models are run and analyzed under various test scenarios such as Double Lane Change, Constant Radius, Increase Steer, Fish Hook, Sine with Dwell and Swept Sine. Only for the 3rd NN, the actuator signals (τ1-τ2) are recomposed by applying an inverse FFT process to the 4 targets (amplitudes and phases of τ1-τ2). Finally, the reference trajectory tracking performances are evaluated by comparing the NN models that are run under the test scenarios.