Showing 1–4 of 4 results for author: Pozo, D

Assessment of Uncertainty Quantification in Universal Differential Equations

Authors: Nina Schmid, David Fernandes del Pozo, Willem Waegeman, Jan Hasenauer

Abstract: Scientific Machine Learning is a new class of approaches that integrate physical knowledge and mechanistic models with data-driven techniques for uncovering governing equations of complex processes. Among the available approaches, Universal Differential Equations (UDEs) are used to combine prior knowledge in the form of mechanistic formulations with universal function approximators, like neural ne… ▽ More Scientific Machine Learning is a new class of approaches that integrate physical knowledge and mechanistic models with data-driven techniques for uncovering governing equations of complex processes. Among the available approaches, Universal Differential Equations (UDEs) are used to combine prior knowledge in the form of mechanistic formulations with universal function approximators, like neural networks. Integral to the efficacy of UDEs is the joint estimation of parameters within mechanistic formulations and the universal function approximators using empirical data. The robustness and applicability of resultant models, however, hinge upon the rigorous quantification of uncertainties associated with these parameters, as well as the predictive capabilities of the overall model or its constituent components. With this work, we provide a formalisation of uncertainty quantification (UQ) for UDEs and investigate important frequentist and Bayesian methods. By analysing three synthetic examples of varying complexity, we evaluate the validity and efficiency of ensembles, variational inference and Markov chain Monte Carlo sampling as epistemic UQ methods for UDEs. △ Less

Submitted 13 June, 2024; originally announced June 2024.

Comments: Shared last authorship between W.W. and J.H

Sub-millisecond Video Synchronization of Multiple Android Smartphones

Authors: Azat Akhmetyanov, Anastasiia Kornilova, Marsel Faizullin, David Pozo, Gonzalo Ferrer

Abstract: This paper addresses the problem of building an affordable easy-to-setup synchronized multi-view camera system, which is in demand for many Computer Vision and Robotics applications in high-dynamic environments. In our work, we propose a solution for this problem -- a publicly-available Android application for synchronized video recording on multiple smartphones with sub-millisecond accuracy. We p… ▽ More This paper addresses the problem of building an affordable easy-to-setup synchronized multi-view camera system, which is in demand for many Computer Vision and Robotics applications in high-dynamic environments. In our work, we propose a solution for this problem -- a publicly-available Android application for synchronized video recording on multiple smartphones with sub-millisecond accuracy. We present a generalized mathematical model of timestamping for Android smartphones and prove its applicability on 47 different physical devices. Also, we estimate the time drift parameter for those smartphones, which is less than 1.2 msec per minute for most of the considered devices, that makes smartphones' camera system a worthy analog for professional multi-view systems. Finally, we demonstrate Android-app performance on the camera system built from Android smartphones quantitatively on setup with lights and qualitatively -- on panorama stitching task. △ Less

Submitted 26 August, 2021; v1 submitted 2 July, 2021; originally announced July 2021.

Comments: Accepted to conference IEEE Sensors'2021 as Lecture presentation

Segmentation and Defect Classification of the Power Line Insulators: A Deep Learning-based Approach

Authors: Arman Alahyari, Anton Hinneck, Rahim Tariverdi, David Pozo

Abstract: Power transmission networks physically connect the power generators to the electric consumers. Such systems extend over hundreds of kilometers. There are many components in the transmission infrastructure that require a proper inspection to guarantee flawless performance and reliable delivery, which, if done manually, can be very costly and time consuming. One essential component is the insulator.… ▽ More Power transmission networks physically connect the power generators to the electric consumers. Such systems extend over hundreds of kilometers. There are many components in the transmission infrastructure that require a proper inspection to guarantee flawless performance and reliable delivery, which, if done manually, can be very costly and time consuming. One essential component is the insulator. Its failure can cause an interruption of the entire transmission line or a widespread power failure. Automated fault detection could significantly decrease inspection time and related costs. Recently, several works have been proposed based on convolutional neural networks, which address the issue mentioned above. However, existing studies focus on a specific type of insulator faults. Thus, in this study, we introduce a two-stage model that segments insulators from their background to then classify their states based on four different categories, namely: healthy, broken, burned/corroded and missing cap. The test results show that the proposed approach can realize the effective segmentation of insulators and achieve high accuracy in detecting several types of faults. △ Less

Submitted 15 October, 2020; v1 submitted 21 September, 2020; originally announced September 2020.

Cyberattack Detection in Intelligent Grids Using Non-linear Filtering

Authors: Irina Lukicheva, David Pozo, Alexander Kulikov

Abstract: Electric power grids are evolving towards intellectualization such as Smart Grids or active-adaptive networks. Intelligent power network implies usage of sensors, smart meters, electronic devices and sophisticated communication network. This leads to a strong dependence on information and communication networking that are prone to threats of cyberattacks, which challenges power system reliability… ▽ More Electric power grids are evolving towards intellectualization such as Smart Grids or active-adaptive networks. Intelligent power network implies usage of sensors, smart meters, electronic devices and sophisticated communication network. This leads to a strong dependence on information and communication networking that are prone to threats of cyberattacks, which challenges power system reliability and efficiency. Thus, significant attention should be paid to the Smart Grids security. Recently, it has been proven that False Data Injection Attacks (FDIA) could corrupt results of State Estimation (SE) without noticing, therefore, leading to a possible mis-operation of the whole power system. In this paper, we introduce an algorithm for detecting cyberattacks based on non-linear filtering by using cyber-physical information from Kirchhoff laws. The proposed algorithm only needs data from adjacent nodes, therefore can be locally and distributed implemented. Also, it requires very low computational effort so that it can be run online, and it is suitable for implementation in existing or new ad-hoc low-cost devices. The proposed algorithm could be helpful to increase power system awareness against FDIA complementing the current SE implementations. The efficiency of the proposed algorithm has been proved by mathematical simulations and computer modeling in PSCAD software. Our results show that the proposed methodology can detect cyberattacks to the SE in 99.9% of the cases with very little false alarms on the identification of spoiled measurements (4.6%). △ Less

Submitted 16 July, 2018; v1 submitted 28 June, 2018; originally announced June 2018.