Detection of unsupervised anomalies in light sensors

Autor: Gras López, Sergio
Přispěvatelé: Universitat Politècnica de Catalunya. Arquitectura de Computadors, Giesecke & Devrient, Meseguer Pallarès, Roc
Jazyk: angličtina
Rok vydání: 2023
Předmět:
Popis: This project aims to investigate and create a solution to a problem that the company Giesecke & Devrient has in one of the quality control processes. The base of one of its products is made up of a metallic foil of a certain color that has to meet quality standards, but producing it requires a complex process where errors appear. The project aims to use data collected from 11 sensors that measure the foil wavelength to identify these errors. However, there is a challenge as there are no established values for what is considered acceptable, and the sensors used do not work in optimal conditions, which generates measurement errors in the data set. To tackle this challenge, extensive data analysis was performed to understand the system. Different techniques and principles have also been applied for the detection of anomalies, analyzing and comparing the results. Ultimately, a prototype tool was developed to integrate these techniques and improve the quality control process Objectius de Desenvolupament Sostenible::9 - Indústria, Innovació i Infraestructura
Databáze: OpenAIRE