Periodic Reporting for period 2 - AEMS-IdFit (Aircraft Electrical Model Simulation Identification and Fitting Toolbox)
Reporting period: 2019-01-01 to 2020-06-30
In this context, the development achieved by the AEMS-IdFit, provide model updating methods for complex electrical simulation models as follows:
1. Adaptation, transformation and updating of a data base of measured signals for all hardware components to be validated.
2. Adaptation, updating and development of SaberRD generic simulation models with the capacity of reproducing the real behavior of the hardware components from experimental measurements and design data. The source code of the SaberRD models has been modified to accommodate such models to the requirements of the identification and fitting methods developed and to include some physical effects not included in the original version of such models.
3. Analysis, selection and performance evaluation of accurate parameter identification and data fitting methods based on an optimization approach.
4. Development of a methodology for generalizing the parameter identification from a wide range of real operating conditions, by developing suitable white-box and black-box models.
5. Implementation and integration of the parameter identification, fitting and mapping methods developed in the IdFit-Toolbox.
6. Development of a user-friendly software that includes all the functions developed in the project.
Conclusions:
1.- The main objectives of the project have been accomplished successfully.
2.- Four different identification methods have been applied, which are based on white-box and black-box approaches. The user is able to identify the parameters of the electrical and electronic devices when applying the white-box estimation algorithms. When there is no information about the topology of the circuit, the proposed black-box algorithms train a neural network that reproduce the behavior of the devices.
3.-The library generated in SaberRD considers multiple upgrades to the models of the power converters and the synchronous machine. This enhance the accuracy of the simulations, which reproduce with more fidelity the experimental data.
4.- The algorithms developed in the project have been implemented in four different graphical user interfaces (GUIs). Depending on the type of device (power converter or electrical machine) and the available information of the circuits, the user may select one of the apps.
5.- Most of the developments within the AEMS-IdFit project can be applied to other industrial sectors, such as transportation, automobile or mechatronics sectors among others.
• A detailed bibliographical research on the state-of-the-art of identification procedures for white box, grey box and black box models of electrical and electronics components was presented, as well as a detailed bibliographical research of different parameter fitting methods has also been performed.
• The main parameters for both electrical and electronics components were identified.
• Preliminary selection of the key parameters to take into account in the SaberRD models for power generators, power converters and filters has been carried out.
• Different pre-processing techniques were analyzed and presented. These techniques are based on resizing, filtering and interpolation.
• Improvement of the electrical models of the DC-DC power converters, six-pulse rectifier, EMI filter, and permanent magnet three-phase synchronous machine (PMSM).
• The final version of the AEMS-IdFiT library in SaberRD was finished based on the model improvements, and it was presented to Airbus.
• Two different methods were proposed for the white-box parameter estimation of the power converters and electrical machines. The first one is based on the differential equations of the circuits, while the second one is based on the non-linear least squares (NLS) optimization iterative method. It was proved that the NLS approach has higher accuracy in identifying the power converters parameters.
• Two different black-box identification approaches based on neural networks algorithms (NARX and LSTM) were developed and tested using experimental data for power converters and simulated data for electrical machines. Both approaches show high accuracy and are able to replicate the behavior of the devices.
• The methods explained above were implemented in four different Graphical User Interfaces (GUIs) in order to make it user friendly. The user is able to identify the parameters using the white-box approaches or to train the neural networks using the black-box approach.
Exploitation and Dissemination
The main channel of dissemination of the AEMS-IdFiT project is the webpage: https://aemsidfit.upc.edu/en where all the updates regarding to the project are posted as well as the non-confidential information . Furthermore, the aeronautics schools of different universities, manufacturers and organizations related to the aeronautic sectors were contacted in order to present them an overview of the project. Finally, five research articles in international journals and conferences have been published as a direct result of the AEMS-IdFiT project.
Overview of final period results
A total of four apps have been generated:
• Parameter estimation of power converters
• Parameter estimation of synchronous electrical machine (stage 1)
• Parameter estimation of synchronous electrical machine (stage 2)
• Black-box system identification
• The proposed methods are non-invasive, from measurements at the input and the output of the converters is enough to identify the converter parameters or the neural network matrices.
• The white-box based identification methods offer flexibility because the topology of the model can be changed.
• The black-box approach provides a matrix with different weights that represent the behavior of the devices. Therefore, the prediction of possible outputs of the system is relatively easy because it is just a mathematical operation.
Since the scope of the project involves the equipment present in aircraft electrical systems such as power sources, power converters, power loads, and the development of specific software tools to improve the state of the art models of such devices as well as the development of parameter identification and fitting methods to better reproduce the real behavior of such elements by means of improved models, the AEMS-IdFit project will have a great impact in a large number of SMEs and large enterprises of the aeronautic sector such as designers, manufacturers and software developers as well as research institutions.