Non-neural Machine Learning (ML) and Deep Learning (DL) models are often used to predict system failures in the context of industrial maintenance. However, only a few researches jointly assess the effect of varying the amount of past data used to make a prediction and the extension in the future of the forecast. This study evaluates the impact of the size of the reading window and of the prediction window on the performances of models trained to forecast failures in three data sets concerning the operation of (1) an industrial wrapping machine working in discrete sessions, (2) an industrial blood refrigerator working continuously, and (3) a nitrogen generator working continuously. The problem is formulated as a binary classification task that assigns the positive label to the prediction window based on the probability of a failure to occur in such an interval. Six algorithms (Linear Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Long short-term memory (LSTM), Convolutional Long Short Term Memory (ConvLSTM), and Transformers) are compared using multivariate telemetry time series. The results indicate that, in the considered scenarios, the dimension of the prediction windows plays a crucial role and highlight the effectiveness of DL approaches at classifying data with diverse time-dependent patterns preceding a failure and the effectiveness of ML approaches at classifying similar and repetitive patterns preceding a failure.
I modelli non neurali di Machine Learning e quelli di Deep Learning sono spesso usati per predire dei fallimenti nei sistemi nel contesto di manutenzione industriale. Tuttavia, solo alcune ricerche analizzano l’effetto di variare il numero di dati storici usati per fare le predizioni e l’estendere la predizione nel futuro. Questo studio valuta l’impatto della dimensione della finestra di lettura e della finestra di predizione sulle performance del modello allenato per predire fallimenti in tre datasets che riguardano: (1) una macchina pallettatrice industriale che lavora in sessioni discrete, (2) un frigorifero per il sangue industriale che lavora in modo continuo e (3) un generatore di azoto che lavora in modo continuo. Il problema è formulato come una classificazione binaria che attribuisce delle etichette positive per predire finestre basandosi sulla probabilità che sia presente un fallimento in questo intervallo. Sei algoritmi (LR, RF, SVM, LSTM, ConvLSTM, e Transformers) sono paragonati usando serie temporali multivariate telemetriche. I risultati mostrano che, considerando questi scenari, la dimensione della finestra di predizione ha un ruolo fondamentale e evidenzia l’efficacia degli approcci di Deep Learning nel classificare dati con diversi comportamenti dipendenti dal tempo che precedono un fallimento e l’efficacia degli approcci di Machine Learning nel classificare comportamenti simili tra loro e ripetuti che precedono un fallimento.
Predicting failures in industrial compressor-based machines
Forbicini, Francesca
2022/2023
Abstract
Non-neural Machine Learning (ML) and Deep Learning (DL) models are often used to predict system failures in the context of industrial maintenance. However, only a few researches jointly assess the effect of varying the amount of past data used to make a prediction and the extension in the future of the forecast. This study evaluates the impact of the size of the reading window and of the prediction window on the performances of models trained to forecast failures in three data sets concerning the operation of (1) an industrial wrapping machine working in discrete sessions, (2) an industrial blood refrigerator working continuously, and (3) a nitrogen generator working continuously. The problem is formulated as a binary classification task that assigns the positive label to the prediction window based on the probability of a failure to occur in such an interval. Six algorithms (Linear Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Long short-term memory (LSTM), Convolutional Long Short Term Memory (ConvLSTM), and Transformers) are compared using multivariate telemetry time series. The results indicate that, in the considered scenarios, the dimension of the prediction windows plays a crucial role and highlight the effectiveness of DL approaches at classifying data with diverse time-dependent patterns preceding a failure and the effectiveness of ML approaches at classifying similar and repetitive patterns preceding a failure.File | Dimensione | Formato | |
---|---|---|---|
2023_12_Forbicini_Tesi_01.pdf
accessibile in internet per tutti
Descrizione: testo tesi
Dimensione
1.2 MB
Formato
Adobe PDF
|
1.2 MB | Adobe PDF | Visualizza/Apri |
2023_12_Forbicini_Executive_Summary_02.pdf
accessibile in internet per tutti
Descrizione: testo executive summary
Dimensione
382.13 kB
Formato
Adobe PDF
|
382.13 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/214117