Optimal Input Design for Subspace-Based Fault Detection and Identification

نویسندگان

  • Alireza Esna Ashari
  • Laurent Mevel
چکیده

This study focuses on input design for subspace based fault detection and identification methods and investigates its possible advantages over using noise inputs. In several real applications the noise available in environment is the only input to the system and in some cases produce low quality output data for subspace identification and fault detection purposes. Therefore, model order may be underestimated. Due to the nature of subspace based methods, some modes of the system may not appear in the response as the input is not strong enough to excite these modes. In order to improve the result, a method is suggested in literature, that is to use “rotated” input. The rotated input design is proposed in several papers to apply to “ill conditioned systems” in which the vector of different outputs are typically close to collinearity if a white noise is used. In this report, we use this technique to verify possible improvement of subspace-based identification method including output-only, and input-output approaches. Then, for the first time we investigate the possible impacts of the rotated input on subspace base fault detection method. Simulations on a high-purity distillation column shows that this auxiliary input can improve subspace-based fault detection and identification. Key-words: Fault Detection, Stochastic System Identification, Subspace Identification, Rotated Input, Ill Conditioned Systems. This work was supported by the European project FP7-NMP CPIP 213968-2 IRIS. ∗ INRIA, Centre Rennes Bretagne Atlantique † INRIA, Centre Rennes Bretagne Atlantique in ria -0 05 89 50 8, v er si on 1 29 A pr 2 01 1 Conception Optimale d’Entrée pour la Détection et l’Identification de défauts par la méthode des sous-espaces Résumé : Cette étude se concentre sur la conception d’entrées pour la détection de défauts par méthodes sous-espace. Nous étudions les avantages possibles d’utiliser une entrée autre que le bruit. Dans plusieurs applications réelles, le bruit produit par l’environnement est la seule entrée du système et dans certains cas, produit des données de qualité faible pour l’identification et la détection de défauts par (mais pas seulement) méthodes sous-espace. Par conséquent, l’ordre du modèle peut être sous-estimé. En raison de la nature et de la construction des méthodes à base de sous-espace, certains modes du système peuvent ne pas apparaître dans la réponse, parce que l’entrée n’est pas assez forte pour exciter ces modes. Afin d’améliorer le résultat, une méthode est suggéré dans la littérature. Elle utilise une entrée “tournée”. Cette méthode-la est proposé dans plusieurs articles pour s’appliquer aux systèmes “mal conditionnés” dans laquelle le vecteur des sorties différentes sont généralement proches de co-linéarité si un bruit blanc est utilisé. Dans ce rapport, nous utilisons cette technique pour vérifier l"amélioration possible apportée aux méthodes d’identification sous-espace, y compris en sortie-seule, mais aussi en entrées-sorties. De plus, pour la première fois, nous étudions l’impact possible de l’entrée tournée sur la méthode de détection de faute par sous-espace . Des simulations sur une colonne de distillation de haute pureté montrent que cette entrée tournée auxiliaire peut améliorer la détection de défauts sous-espace ainsi que l’identification sous-espace. Mots-clés : Détection de pannes, système d’identification stochastique, Identification sous-espaces, entrée tournée, systèmes mal conditionnés. in ria -0 05 89 50 8, v er si on 1 29 A pr 2 01 1 SSI and Rotated Input 3

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Design of nonlinear parity approach to fault detection and identification based on Takagi-Sugeno fuzzy model and unknown input observer in nonlinear systems

In this study, a novel fault detection scheme is developed for a class of nonlinear system in the presence of sensor noise. A nonlinear Takagi-Sugeno fuzzy model is implemented to create multiple models. While the T-S fuzzy model is used for only the nonlinear distribution matrix of the fault and measurement signals, a larger category of nonlinear systems is considered. Next, a mapping to decou...

متن کامل

An Unknown Input Observer for Fault Detection Based on Sliding Mode Observer in Electrical Steering Assist Systems

Steering assist system controls the force transfer behavior of the steering system and improves the steering probability of the vehicle. Moreover, it is an interface between the diver and vehicle. Fault detection in electrical assisted steering systems is a challenging problem due to frequently use of these systems. This paper addresses the fault detection and reconstruction in automotive elect...

متن کامل

Robust Model- Based Fault Detection and Isolation for V47/660kW Wind Turbine

In this paper, in order to increase the efficiency, to reduce the cost and to prevent the failures of wind turbines, which lead to an extensive break down, a robust fault diagnosis system is proposed for V47/660kW wind turbine operated in Manjil wind farm, Gilan province, Iran. According to the acquired data from Iran wind turbine industry, common faults of the wind turbine such as sensor fault...

متن کامل

STRUCTURAL DAMAGE PROGNOSIS BY EVALUATING MODAL DATA ORTHOGONALITY USING CHAOTIC IMPERIALIST COMPETITIVE ALGORITHM

Presenting structural damage detection problem as an inverse model-updating approach is one of the well-known methods which can reach to informative features of damages. This paper proposes a model-based method for fault prognosis in engineering structures. A new damage-sensitive cost function is suggested by employing the main concepts of the Modal Assurance Criterion (MAC) on the first severa...

متن کامل

Accurate Fruits Fault Detection in Agricultural Goods using an Efficient Algorithm

The main purpose of this paper was to introduce an efficient algorithm for fault identification in fruits images. First, input image was de-noised using the combination of Block Matching and 3D filtering (BM3D) and Principle Component Analysis (PCA) model. Afterward, in order to reduce the size of images and increase the execution speed, refined Discrete Cosine Transform (DCT) algorithm was uti...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2011