Karakterizacija in avtomatsko zaznavanje kakovosti elektromotorjev pralnih strojev

Valenčič, David (author), Govekar, Edvard (mentor), Potočnik, Primož (co-mentor)

V okviru dela smo analizirali možnost avtomatske zaznave kakovosti elektromotorjev pralnih strojev na osnovi signalov zajetih s pospeškomerom in mikrofonom. Meritve smo izvajali ločeno na realnem pralnem stroju in izdelanem eksperimentalnem modelu pralnega stroja. Na osnovi analize smo izpeljali več kot 50 različnih značilk v časovni in frekvenčni domeni ter izvedli redukcijo dimenzionalnosti z metodo glavnih osi. Na eksperimentalnem modelu pralnega stroja nam je največjo ločljivost med dobrimi in slabimi elektromotorji zagotovila značilka moči iz signala mikrofona v frekvenčnem pasu 60 - 70 Hz, medtem ko je na realnem pralnem stroju izstopala značilka 50. percentil iz signala enoosnega pospeškomera pritrjenega na stranici pralnega stroja. Z uporabo večdimenzijskih grafov značilk in uporabo diskriminantne analize smo uspešno ločevali med dobrimi in slabimi elektromotorji, skladno s predhodno predvideno kakovostjo določeno s strani sodelujočega podjetja.

• Language: Slovenian
• Keywords: karakterizacija komponent, zaznava kakovosti, analiza signalov, pralni stroj, elektromotor
• Work type: Master's thesis/paper
• Organization: FS - Faculty of Mechanical Engineering
• Year: 2017
• PID: 20.500.12556/RUL-92959
• Publication date in RUL: 11.07.2017

Secondary language

• Language: English
• Title: Characterisation and automatic quality detection of electromotors in washing machines

Abstract

The work studies a possibility of characterisation and quality detection of electromotors in washing machine from signals of accelerometers and microphone. For this purpose we developed experimental model of a washing machine with the same dynamic and structural properties. We have separately observed and measured received motors on a built test machine and on a real washing machine provided by the cooperating company. During signal processing we developed more than 50 features in time and frequency domain and reduced the dimension with a principal component analysis. Feature carrying most quality dependent information on experimental model of a washing machine was power in a frequency band from 60 Hz to 70 Hz on a microphone and on a real washing machine, it was 50. percentile from a single axis accelerometer mounted on a housing of the machine. With the use of multidimension feature chart and discriminant analysis we successfully draw the classification line between higher and lower quality motors in accordance with quality defined by a participating company.

• Keywords: characterisation of components, quality detection, signal processing, washing machine, electric motor