1,721,033 research outputs found
Modeling and Diagnosis of Complex Systems Dynamics by Data-Driven Approaches
No full textI sistemi complessi possono essere rinvenuti in quasi tutti i campi della scienza
contemporanea, e possono avere diversa natura: finanziaria, fisica, biologica,
informativa, sociale, ecc. I sistemi complessi consistono di un gran numero di
componenti che interagiscono non linearmente tra loro e che mostrano un comportamento
collettivo non derivante semplicemente dal comportamento delle
parti individuali. Sebbene tali sistemi godano di numerose proprietà, le più
importanti sono: dimensionalità, incertezza, non linearità e accoppiamento tra
i componenti. Le procedure per ottenere modelli analitici di determinati sistemi
sono solitamente classificate in modellazione fisica e identificazione. Queste procedure
possono essere difficilmente implementabili se applicate a sistemi complessi
perché le loro caratteristiche peculiari rendono difficile la modellazione.
Poiché un modello matematico è una descrizione del comportamento di un sistema,
una modellazione accurata per i sistemi complessi è molto difficile da
ottenere in pratica. Per di più, a volte, potrebbe addirittura essere impossibile
descrivere il sistema attraverso equazioni analitiche.
Alla luce di quanto emerso, la presente trattazione si propone di affrontare
due problemi riguardanti la modellazione e la diagnosi dei sistemi complessi: il
primo riguarda specificamente la modellazione di un sistema complesso, nel caso
in cui il modello analitico non sia ottenibile; il secondo si riferisce alla diagnosi
del comportamento del sistema. Quest’ultima attività dovrebbe rilevare se il
sistema complesso è normale o se sta avvenendo un cambiamento dovuto a
eventi anomali, nonché le cause probabili di tali eventi.
La modellazione dei sistemi complessi viene affrontata sviluppando metodi
data-driven, che sono capaci di apprendere le dinamiche del sistema complesso
direttamente dai dati forniti da sensori installati sul sistema, al fine di monitorarne
le variabili fisiche.
La diagnosi dei sistemi complessi viene invece affrontata sviluppando metodi
di apprendimento automatico in modo da classificare le probabili cause di
scostamento da eventi normali del sistema.
Nella trattazione ampia attenzione è posta al problema di modellazione e
diagnosi di sistemi complessi con riferimento a sistemi reali, presentando diverse
applicazioni pratiche e casi di studio. Il primo caso di studio riguarda la
modellazione e diagnosi di difetti e guasti di motori elettrici in uno scenario di
controllo qualità mentre il secondo si riferisce ad un sistema complesso indusxiii
triale, quale quello di una cartiera. Nel terzo caso viene affrontata la questione
di stimare la vita utile rimasta di un motore turbofan e l’ultimo tratta il problema
di modellare segnali elettroencefalografici attraverso algoritmi basati sui
dati. Dato che il problema di modellazione e diagnosi è affrontato attraverso
procedure basate sui dati, gli algoritmi sviluppati possono essere applicati ad
un’ampia classe di macchine elettriche rotanti e sistemi complessi industriali, e
non solo a quelli riportati
Electric motor defects diagnosis based on kernel density estimation and Kullback-Leibler divergence in quality control scenario
No full textThe present paper deals with the defect detection and diagnosis of induction motor, based on motor current signature analysis in a quality control scenario. In order to develop a monitoring system and improve the reliability of induction motors, Clarke-Concordia transformation and kernel density estimation are employed to estimate the probability density function of data related to healthy and faulty motors. Kullback-Leibler divergence identifies the dissimilarity between two probability distributions and it is used as an index for the automatic defects identification. Kernel density estimation is improved by fast Gaussian transform. Since these techniques achieve a remarkable computational cost reduction respect the standard kernel density estimation, the developed monitoring procedure became applicable on line, as a Quality Control method for the end of production line test. Several simulations and experimentations are carried out in order to verify the proposed methodology effectiveness: broken rotor bars and connectors are simulated, while experimentations are carried out on real motors at the end of production line. Results show that the proposed data-driven diagnosis procedure is able to detect and diagnose different induction motor faults and defects, improving the reliability of induction machines in quality control scenario. © 2015 Elsevier Ltd
Fixed-size LS-SVM LPV System Identification for Large Datasets
Full text linkIn this paper, we propose an efficient method for handling large datasets in linear parameter-varying (LPV) model identification. The method is based on least-squares support vector machine (LS-SVM) identification in the primal space. To make the identification computationally feasible, even for very large datasets, we propose estimating a finite-dimensional feature map. To achieve this, we propose a two-step method to reduce the computational effort. First, we define the training set as a fixed-size subsample of the entire dataset, considering collision entropy for subset selection. The second step involves approximating the feature map through the eigenvalue decomposition of the kernel matrices. This paper considers both autoregressive with exogenous input (ARX) and state-space (SS) model forms. By comparing the problem formulation in the primal and dual spaces in terms of accuracy and computational complexity, the main advantage of the proposed technique is the reduction in space and time complexity during the training stage, making it preferable for handling very large datasets. To validate our proposed primal approach, we apply it to estimate LPV models using provided inputs, outputs, and scheduling signals for two nonlinear benchmarks: the parallel Wiener-Hammerstein system and the Silverbox system. The performances of our proposed approach are compared with the dual LS-SVM approach and the kernel principal component regression
Adaptive Reference Governor for DC–DC Converters Based on Model Predictive Control
Full text linkIn this article, we propose a time-varying model predictive control (MPC)-based scheme to enhance the dynamic performance of dc–dc converters. The proposed approach employs MPC as a reference governor (RG), addressing industrial certification constraints that may limit modifications to the low-level controller. To accommodate the computational limitations of conventional control boards, we introduce a highly efficient real-time optimization algorithm for solving equality-constrained quadratic programming (QP) problems. The algorithm is based on a tailored QR factorization that outperforms well-known linear algebra libraries, and it is shown to be superior to condensing with state elimination. Furthermore, we implement an efficient recursive least-squares (RLS) method to provide a linear-time varying model for the adaptive MPC-based RG. No information regarding the topology of the converter nor the structure of the low-level controller is required for such adaptation, making the proposed method self-tuning and eliminating the need for prior identification steps. The proposed control scheme has been tested on various simulated and real dc–dc converters, demonstrating its computational and memory efficiency, as well as its versatility across different converter topologies
A Kernel-based Learning Approach for Nonlinear MIMO Systems in an Iterative Learning Control Framework
No full textThis paper introduces a kernel-based method for learning feedforward controllers within an Iterative Learning Control (ILC) framework tailored for nonlinear processes. Unlike traditional ILC algorithm that relies on the knowledge of first principle-based models, this approach leverages a data-driven methodology to develop an iterative control update rule using kernel-based training. We compared this method against a traditional ILC scheme and a baseline neural network-based approach. The effectiveness of the proposed method is demonstrated through a unicycle path-following control problem, evaluated across various simulated test scenarios. Performance metrics include vehicle tracking error and ILC convergence speed, confirming the effectiveness of the proposed data-driven approach
Fuzzy logic based economical analysis of photovoltaic energy management
No full textSince 2002 the European Union has seen a rapid growth in the photovoltaic (PV) sector. During the last two years incentives for PV installations were cut almost worldwide slowing the growth of the market. In this scenario the design of a new plant ensuring economic convenience is strongly related to household electricity consumption patterns and energy management actions. This paper presents a high-resolution model of domestic electricity use based on Fuzzy Logic Inference System. Taking into account consumers sensibility concerning the rational use of energy, the model gives as output a 1-min resolution overall electricity usage pattern of the household. The focus of this work is the use of a novel fuzzy model combined with a cost benefits analysis to evaluate the real economic benefits of load shifting actions. A case study is presented to quantify its effectiveness in the new net metering Italian scenario. © 2015 Elsevier B.V
Electric motor fault detection and diagnosis by kernel density estimation and kullback-Leibler divergence based on stator current measurements
No full textThis paper deals with the problem of fault detection and diagnosis of induction motor based on motor current signature analysis. Principal component analysis is used to reduce the three-phase current space to a 2-D space. Kernel density estimation (KDE) is adopted to evaluate the probability density functions of each healthy and faulty motor, which can be used as features in order to identify each fault. Kullback-Leibler divergence is used as an index to identify the dissimilarity between two probability distributions, and it allows automatic fault identification. The aim is also to improve computational performance in order to apply online a monitoring system. KDE is improved by fast Gaussian transform and a points reduction procedure. Since these techniques achieve a remarkable computational cost reduction with respect to the standard KDE, the algorithm can be used online. Experiments are carried out using two alternate current motors: an asynchronous induction machine and a single-phase motor. The faults considered to test the developed algorithm are cracked rotor, out-of-tolerance geometry rotor, and backlash. Tests are carried out at different load and voltage levels to show the proposed method performance
RGBD camera monitoring system for Alzheimer’s disease assessment using Recurrent Neural Networks with Parametric Bias action recognition
No full textThe present paper proposes a computer vision system to diagnose the stage of illness in patients a ected by Alzheimer's disease. In the context of Ambient Assisted Living (AAL), the system monitors people in home environment during daily personal care activities. The aim is to evaluate the dementia stage, observing actions listed in the Direct Assessment of Funcional Status (DAFS) index and detecting anomalies during the performance, in order to assign a score explaining if the action is correct or not. In this work brushing teeth and grooming hair by a hairbrush are analysed. The technology consists of the application of a Recurrent Neural Network with Parametric Bias (RNNPB) that is able to learn movements connected with a speci c action and recognize human activities by parametric bias that work like mirror neurons. This study has been conducted using Microsoft Kinect to collect data about the actions observed and oversee the user tracking and gesture recognition. Experiments prove that the proposed computer vision system can learn and recognize complex human activities and evaluates DAFS score
- …
