Condition monitoring of the cooling of variable speed drives using artificial intelligence
Translated title
Tilstandsovervaagning af koeling af variabel hastighed frekvensomformer med brug af kunstig intelligens
Author
Izquierdo Gomez, Pere
Term
4. term
Education
Publication year
2020
Submitted on
2020-05-29
Abstract
Kunstig intelligens (AI) bruger data til at opbygge modeller, der kan analysere, hvordan systemer fungerer, og forudsige, hvad de vil gøre. Denne afhandling anvender AI til at overvåge tilstanden af kølesystemet i et variabelhastighedsdrev (en enhed, der styrer hastigheden på en elmotor). Vi indsamler målinger fra en fysisk testopstilling ved forskellige driftsbetingelser og bruger disse data til at træne kunstige neurale netværk, altså computermodeller der lærer mønstre ud fra eksempler. Vi undersøger to komplementære tilgange til tilstandsovervågning: (1) modeller, der forudsiger drevets og kølesystemets adfærd på fremtidige tidspunkter, og (2) modeller, der direkte estimerer en tilstandsindikator, dvs. en enkelt værdi, som opsummerer kølesystemets aktuelle tilstand. Derudover beskriver afhandlingen, hvordan drevet og dets kølesystem modelleres, og giver den teoretiske baggrund for kunstige neurale netværk.
Artificial intelligence (AI) uses data to build models that can analyze how systems work and predict what they will do next. This thesis applies AI to monitor the condition of the cooling system in a variable speed drive, a device that controls the speed of an electric motor. We collect measurements from a physical test setup under different operating conditions and use these data to train artificial neural networks—computer models that learn patterns from examples. We study two complementary approaches to condition monitoring: (1) models that forecast the drive’s and cooling system’s behavior at future points in time, and (2) models that estimate a health indicator directly, that is, a single value that summarizes the current state of the cooling system. In addition to the data-driven results, the thesis describes how the drive and its cooling system are modeled and provides the theoretical background needed to understand artificial neural networks.
[This abstract was generated with the help of AI]
Documents