Sposito, Gianluca (2024) Artificial neural networks and statistics for quality technology in railway industry. [Tesi di dottorato]
Preview |
Text
Sposito_Gianluca_36.pdf Download (2MB) | Preview |
| Item Type: | Tesi di dottorato |
|---|---|
| Resource language: | English |
| Title: | Artificial neural networks and statistics for quality technology in railway industry |
| Creators: | Creators Email Sposito, Gianluca gianluca.sposito@unina.it |
| Date: | 20 March 2024 |
| Number of Pages: | 125 |
| Institution: | Università degli Studi di Napoli Federico II |
| Department: | Ingegneria Industriale |
| Dottorato: | Ingegneria industriale |
| Ciclo di dottorato: | 36 |
| Coordinatore del Corso di dottorato: | nome email Grassi, Michele michele.grassi@unina.it |
| Tutor: | nome email Lepore, Antonio UNSPECIFIED Palumbo, Biagio UNSPECIFIED |
| Date: | 20 March 2024 |
| Number of Pages: | 125 |
| Keywords: | statistical process control; artificial neural networks; railway industry |
| Settori scientifico-disciplinari del MIUR: | Area 13 - Scienze economiche e statistiche > SECS-S/02 - Statistica per la ricerca sperimentale e tecnologica |
| Date Deposited: | 21 Mar 2024 13:58 |
| Last Modified: | 10 Mar 2026 08:21 |
| URI: | http://www.fedoa.unina.it/id/eprint/15399 |
Collection description
In the modern Industry 4.0 and 5.0 framework, advancements in technology have enabled an unprecedented collection of a huge amount of data, which is parallelly becoming more complex and typically demands the implementation of novel statistical methodologies. In particular, this thesis aims to develop methods for Statistical Process Control (SPC) through the integration of artificial neural network, or simply Neural Network (NN), algorithms to tackle modern industrial problems in the digitalization era. The industrial scenario that motivates the research work is the monitoring of the Heating, Ventilation and Air Conditioning (HVAC) systems installed onboard modern passenger trains. The proposed methodologies are developed into two main parts in this dissertation. In the first part, the simultaneous SPC of the sensor signals coming from each HVAC system of the same train is regarded as produced by a Multiple Stream Process (MSP). In particular, this part presents a NN-based approach designed to enhance the monitoring of a MSP and the identification of the streams responsible for the out-of-control alarm. The second part of this dissertation explores the possibility of using a nonparametric SPC approach based on autoencoders, which are a type of NN used to learn an efficient representation of the input data, and a novel control chart for functional data based on a NN specifically trained to handle data in the form of profiles. The common ground of all the proposed methodologies is the need to provide tools to industrial practitioners that give clear indications if an anomaly occurs and are easy to implement through the development of open source packages, and the use of operational data in the railway sector to demonstrate their practical applicability.
Downloads
Downloads per month over past year
Actions (login required)
 |
View Item |
