First Name
Peng
Last Name
Liu
Dataset Entries from this Author
MagNet is a large-scale dataset designed to enable researchers modeling magnetic core loss using machine learning to accelerate the design process of power electronics. The dataset contains a large amount of voltage and current data of different magnetic components with different shapes of waveforms and different properties measured in the real world. Researchers may use these data as pairs of excitations and responses to build up dynamic magnetic models or calculate the core loss to derive static models.
- Categories:
Data Management Plan Entries from this Author
Multi-State-Space Modeling for Magnetic Core Loss Prediction Using Empirical Approach
- Citation Author(s):
-
Peng Liu
- Created by:
- Peng Liu
- Last updated:
3 views
Project Description
准确预测磁芯损耗 仍然是电磁工程中的关键挑战。协定的 经验公式（例如，修正的 Steinmetz 方程、广义的 Steinmetz equation） 受到过于简化的线性假设的影响，限制了它们的 适用于多材料和多作场景。虽然神经 网络展示了非线性建模功能及其工程效用 由于忽视了多因素耦合效应和对 单变量输入。为了解决这些限制，本研究提出了一种 经验状态空间模型 （ESSM） 将状态空间模型与 Agent-Attention 机制，建立多元协同预测 框架。通过合并磁通密度 （Bt）、磁场 强度 （Ht）、温度和频率组成一个统一的状态空间矩阵， 并采用基于 Manba 的选择性状态转换机制，我们的方法 动态求解交叉因子耦合的非线性特性。 对 MagNet 数据集中的 10 种材料的实验评估表明 RMS 误差为 2.13%，分别比 iGSE 和 FANN 提高了 87% 和 47% 模型。所提出的 ESSM 明显优于神经网络 所有运行条件下的基准（Transformer、LSTM），克服 传统模型的强独立性假设，同时保留物理 可解释性。此外，我们系统地进行多因素趋势分析 揭示了与温度和频率相关的损耗演变模式， 定量确定最佳作范围，以提供通用的 高频电力电子设计中的多耦合建模解决方案。
Data Access and Data Sharing
- Data for Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution will be made accessible to the global technical community on an Open Access basis through IEEE DataPort™, a secure web-based data repository that enables large amounts of data to be stored and enables data to be made publicly accessible to all without cost (Open Access data). The goal is to share Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution data openly with the research community to enable this research to have the maximum impact and to bring exposure to this valuable research.
- Users will not have to make special requests of the researcher to access the data because it will be searchable and available at all times through IEEE DataPort. In IEEE DataPort, data is stored in the Amazon Web Services S3 Cloud and can either be downloaded by users or accessed directly in the Amazon Web Services S3 Cloud to facilitate data analysis. The anticipated size of the dataset files made available on IEEE DataPort is 500M.
- The data from Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution will be stored and made accessible through IEEE DataPort for 10 years. IEEE maintains IEEE DataPort and is committed to ensuring data is stored and accessible for duration of 10 years after the conclusion of the project.
- Original data from Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution project will be posted on IEEE DataPort. Original data may be posted to IEEE DataPort as it becomes available through the project term to enable the scientific community to begin accessing the valuable data even during the project term. Metadata will also be publicly available through IEEE DataPort and the research team may also post processed data on IEEE DataPort.
- Data will be shared with the general public through IEEE DataPort and those who specifically ask for the data will be advised to access it directly through IEEE DataPort. The research team will retain ownership of the data, but will provide a Creative Commons Attribution 4.0 International to IEEE DataPort users so they can download and utilize it to accelerate their research or supplement their research efforts. Any use of the data will require the user to provide the proper citation for the data and IEEE DataPort provides data citations in eight different formats to make it easy for the user to cite the data properly.
- Through IEEE DataPort, it will be possible to track users who access Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution data if necessary.
- Re-use, Re-distribution and Derivatives: As noted above, the research team will retain ownership of the data collected in Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution, but will provide a Creative Commons Attribution 4.0 International license to IEEE DataPort users so they can download and utilize it to accelerate their research or supplement their research efforts. Any use or redistribution of the data will have a requirement of attribution and the users must provide the proper citation for the data. IEEE DataPort provides data citations in eight different formats to make it easy for the user to cite the data properly. The eight different formats are as follows: IEEE, BibTeX, RIS, APA, Harvard, MLA, Vancouver, and Chicago. If NIST requires a specific citation format, that can be included in IEEE DataPort. Matters of appropriate protection of privacy, confidentiality, security, intellectual property, or other rights or requirements are managed through the IEEE DataPort Terms of Use and each IEEE DataPort user that wants to upload a dataset is required to agree to the IEEE DataPort Terms of Use before uploading.
Metadata Requirements on IEEE DataPort
IEEE DataPort is designed to store metadata for each dataset submitted on the platform. All metadata is indexed and accessible in human and machine readable formats. The following metadata is supported by IEEE DataPort:
- Metadata - Required Fields
- Title of Dataset
- Citation: name(s) of data author(s)
- Category (subject matter or topic)
- Abstract
- Instructions
- DOI
- Metadata - Optional Fields
- Keywords
- Data Format
- Related Dataset(s)
- Scripts
- Dataset Image
- Documentation file(s)
Cost of Data Retention and Accessibility
The estimated cost to utilize IEEE DataPort for Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution as described in this Data Management Plan is $1950. The $1950 cost value represents an anticipated data volume of at approximately 500M.
Print Data Management PlanMulti-State-Space Modeling for Magnetic Core Loss Prediction Using Empirical Approach
- Citation Author(s):
-
Peng Liu
- Created by:
- Peng Liu
- Last updated:
1 view
Project Description
准确预测磁芯损耗 仍然是电磁工程中的关键挑战。协定的 经验公式（例如，修正的 Steinmetz 方程、广义的 Steinmetz equation） 受到过于简化的线性假设的影响，限制了它们的 适用于多材料和多作场景。虽然神经 网络展示了非线性建模功能及其工程效用 由于忽视了多因素耦合效应和对 单变量输入。为了解决这些限制，本研究提出了一种 经验状态空间模型 （ESSM） 将状态空间模型与 Agent-Attention 机制，建立多元协同预测 框架。通过合并磁通密度 （Bt）、磁场 强度 （Ht）、温度和频率组成一个统一的状态空间矩阵， 并采用基于 Manba 的选择性状态转换机制，我们的方法 动态求解交叉因子耦合的非线性特性。 对 MagNet 数据集中的 10 种材料的实验评估表明 RMS 误差为 2.13%，分别比 iGSE 和 FANN 提高了 87% 和 47% 模型。所提出的 ESSM 明显优于神经网络 所有运行条件下的基准（Transformer、LSTM），克服 传统模型的强独立性假设，同时保留物理 可解释性。此外，我们系统地进行多因素趋势分析 揭示了与温度和频率相关的损耗演变模式， 定量确定最佳作范围，以提供通用的 高频电力电子设计中的多耦合建模解决方案。
Data Access and Data Sharing
- Data for Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution will be made accessible to the global technical community on an Open Access basis through IEEE DataPort™, a secure web-based data repository that enables large amounts of data to be stored and enables data to be made publicly accessible to all without cost (Open Access data). The goal is to share Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution data openly with the research community to enable this research to have the maximum impact and to bring exposure to this valuable research.
- Users will not have to make special requests of the researcher to access the data because it will be searchable and available at all times through IEEE DataPort. In IEEE DataPort, data is stored in the Amazon Web Services S3 Cloud and can either be downloaded by users or accessed directly in the Amazon Web Services S3 Cloud to facilitate data analysis. The anticipated size of the dataset files made available on IEEE DataPort is 500M.
- The data from Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution will be stored and made accessible through IEEE DataPort for 10 years. IEEE maintains IEEE DataPort and is committed to ensuring data is stored and accessible for duration of 10 years after the conclusion of the project.
- Original data from Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution project will be posted on IEEE DataPort. Original data may be posted to IEEE DataPort as it becomes available through the project term to enable the scientific community to begin accessing the valuable data even during the project term. Metadata will also be publicly available through IEEE DataPort and the research team may also post processed data on IEEE DataPort.
- Data will be shared with the general public through IEEE DataPort and those who specifically ask for the data will be advised to access it directly through IEEE DataPort. The research team will retain ownership of the data, but will provide a Creative Commons Attribution 4.0 International to IEEE DataPort users so they can download and utilize it to accelerate their research or supplement their research efforts. Any use of the data will require the user to provide the proper citation for the data and IEEE DataPort provides data citations in eight different formats to make it easy for the user to cite the data properly.
- Through IEEE DataPort, it will be possible to track users who access Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution data if necessary.
- Re-use, Re-distribution and Derivatives: As noted above, the research team will retain ownership of the data collected in Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution, but will provide a Creative Commons Attribution 4.0 International license to IEEE DataPort users so they can download and utilize it to accelerate their research or supplement their research efforts. Any use or redistribution of the data will have a requirement of attribution and the users must provide the proper citation for the data. IEEE DataPort provides data citations in eight different formats to make it easy for the user to cite the data properly. The eight different formats are as follows: IEEE, BibTeX, RIS, APA, Harvard, MLA, Vancouver, and Chicago. If NIST requires a specific citation format, that can be included in IEEE DataPort. Matters of appropriate protection of privacy, confidentiality, security, intellectual property, or other rights or requirements are managed through the IEEE DataPort Terms of Use and each IEEE DataPort user that wants to upload a dataset is required to agree to the IEEE DataPort Terms of Use before uploading.
Metadata Requirements on IEEE DataPort
IEEE DataPort is designed to store metadata for each dataset submitted on the platform. All metadata is indexed and accessible in human and machine readable formats. The following metadata is supported by IEEE DataPort:
- Metadata - Required Fields
- Title of Dataset
- Citation: name(s) of data author(s)
- Category (subject matter or topic)
- Abstract
- Instructions
- DOI
- Metadata - Optional Fields
- Keywords
- Data Format
- Related Dataset(s)
- Scripts
- Dataset Image
- Documentation file(s)
Cost of Data Retention and Accessibility
The estimated cost to utilize IEEE DataPort for Spatial and Channel Two-step Processing Lightweight Modules to Replace Convolution as described in this Data Management Plan is $1950. The $1950 cost value represents an anticipated data volume of at approximately 500M.
Print Data Management Plan
