The earth is calling for change, every day the effects of global warming are becoming more and more severe, and the role of the construction sector can no longer be ignored. Concrete production is in fact responsible for a large part of CO2 emissions, about 8%, but despite this, it is expected to grow in the coming years. Therefore, modernizing the concrete industry requires the use of advanced technologies and digitalization in order to improve the durability and sustainability of our structures. Two key ideas in this direction are 1) the development of new high-performance concretes (UHDC) which undergo "self-healing" and 2) autonomous systems for monitoring existing structures, based on computer vision. These two technologies, in fact, can potentially lengthen the service life of structures: self-healing offers better performing materials that have greater durability, while, thanks to computer vision, large-scale monitoring by using drones or other automated systems, can be possible. The goal in this thesis was therefore to investigate these two issues and explore the potential of combining self-healing with computer vision and machine learning. All the research done aims to contribute to the construction one day of an automated system that starts from images and estimates structural safety. First of all, experimental data from previous research conducted at Politecnico di Milano on self-healing are collected and analyzed. The results of these analyses are used to derive qualitative empirical laws: they relate the visual crack closure in percentages with an estimation of the recovery of mechanical properties and permeability. The results demonstrate that analyzing the surface is meaningful also in terms of material characteristics. In the second part, the research focus on image processing. This process normally requires three steps: stitching of the microscope photos, recognition of the same crack section at different moments in time, and binarization of the image (to extract the open area). The commercial off-the-shelf softwares do not present problems stitching photos, while they fail to correctly align images taken at different times and require heavy human intervention for the segmentation of the crack. However, by applying Artificial Intelligence, it is possible to greatly improve the performance of these algorithms. This has already been demonstrated for the segmentation of cracks, while "time alignment" is still a novelty. In fact, the crack sections, after a process of self-healing, change some of their fundamental characters such as background and areas filled by self-healing, thus they become hard to be recognized. The last section presents the results of the development of some Artificial Neural Networks for improving "time alignment". The alignment is done by matching some "key-points", and ANNs demonstrate to being successfully able to filter 91.6% of the correct matches compared to the 10.9% of the conventional algorithms.
Il mondo chiede un cambiamento, ogni giorno gli effetti del riscaldamento globale si fanno più gravi e il ruolo del settore delle costruzioni non può più essere ignorato. La produzione di calcestruzzo è infatti responsabile per una gran parte delle emissioni di CO2- circa l'8% -, ma nonostante questo, è destinata a crescere nei prossimi anni. E' quindi arrivato il momento per rivoluzionare anche questa industria ed una delle strade maestre da percorrere è sicuramente quella di utilizzare tecnologie avanzate e digitalizzazione per migliorare durabilità e sostenibilità delle nostre strutture. Due idee chiave in questa direzione sono lo sviluppo 1) di una nuova gamma di calcestruzzi ad alte prestazioni (UHDC) soggetti a "self-healing" - ossia in grado di ripararsi autonomamente - e 2) di sistemi autonomi per il monitoraggio delle opere esistenti, basati su computer vision. Queste due tecnologie, infatti, possono permettere di allungare la vita delle strutture, offrendo materiali più performanti e garantendo un monitoraggio su larga scala altrimenti impossibile. L’obiettivo che mi sono posto nella mia tesi è stato quindi quello di approfondire queste due tematiche, indagando in modo specifico le potenzialità di combinare self-healing con computer vision e machine learning. Tutto il lavoro è stato svolto nell’ottica di contribuire alla costruzione un giorno di un sistema automatizzato, che, partendo da immagini ottenute con droni e telecamere, stimi la variazione di proprietà del materiale e di conseguenza la sicurezza strutturale. Innanzitutto ho raccolto e analizzato i dati sperimentali frutto di precedenti ricerche dal Politecnico di Milano sul self-healing. Da questi ho poi ricavato delle leggi empiriche qualitative con lo scopo di dimostrare che, da un’analisi visiva della superficie si può stimare un recupero delle proprietà meccaniche e di permeabilità. Dimostrato questo mi sono poi concentrato sull'elaborazione delle immagini. Questo processo richiede normalmente tre fasi: collage delle foto del microscopio, riconoscimento della stessa sezione di fessura in momenti temporali diversi e binarizzazione dell'immagine (per estrarre l'area aperta). I comuni software non presentano problemi per collage delle foto, mentre non riescono ad allineare correttamente immagini prese in momenti diversi e richiedono un gravoso intervento umano per il processo di binarizzazione. Applicando però l'Artificial Intelligence è possibile migliorare notevolmente le prestazioni di questi algoritmi. Nello specifico va precisato che questa applicazione è già stata sperimentata per la segmentazione delle fessure, mentre per quanto riguarda “l’allineamento temporale” è una novità. Nell'ultima sezione ho presentato quindi i risultati dello sviluppo di un Artificial Neural Network teso al riconoscimento di una stessa fessura dopo che la medesima ha subito un processo di self-healing e di conseguenza ha mutato alcuni caratteri fondamentali come background e aree riempite dal self-healing. Il riconoscimento è fatto abbinando alcuni "punti chiave" tra le due foto, il nuovo algoritmo ha dimostrato di essere in grado di filtrare con successo il 91,6% degli abbinamenti corretti rispetto al 10,9% degli algoritmi convenzionali.
Computer vision for damage initiation and recovery assessment of self-healing concrete structures
Listrani, Giacomo
2020/2021
Abstract
The earth is calling for change, every day the effects of global warming are becoming more and more severe, and the role of the construction sector can no longer be ignored. Concrete production is in fact responsible for a large part of CO2 emissions, about 8%, but despite this, it is expected to grow in the coming years. Therefore, modernizing the concrete industry requires the use of advanced technologies and digitalization in order to improve the durability and sustainability of our structures. Two key ideas in this direction are 1) the development of new high-performance concretes (UHDC) which undergo "self-healing" and 2) autonomous systems for monitoring existing structures, based on computer vision. These two technologies, in fact, can potentially lengthen the service life of structures: self-healing offers better performing materials that have greater durability, while, thanks to computer vision, large-scale monitoring by using drones or other automated systems, can be possible. The goal in this thesis was therefore to investigate these two issues and explore the potential of combining self-healing with computer vision and machine learning. All the research done aims to contribute to the construction one day of an automated system that starts from images and estimates structural safety. First of all, experimental data from previous research conducted at Politecnico di Milano on self-healing are collected and analyzed. The results of these analyses are used to derive qualitative empirical laws: they relate the visual crack closure in percentages with an estimation of the recovery of mechanical properties and permeability. The results demonstrate that analyzing the surface is meaningful also in terms of material characteristics. In the second part, the research focus on image processing. This process normally requires three steps: stitching of the microscope photos, recognition of the same crack section at different moments in time, and binarization of the image (to extract the open area). The commercial off-the-shelf softwares do not present problems stitching photos, while they fail to correctly align images taken at different times and require heavy human intervention for the segmentation of the crack. However, by applying Artificial Intelligence, it is possible to greatly improve the performance of these algorithms. This has already been demonstrated for the segmentation of cracks, while "time alignment" is still a novelty. In fact, the crack sections, after a process of self-healing, change some of their fundamental characters such as background and areas filled by self-healing, thus they become hard to be recognized. The last section presents the results of the development of some Artificial Neural Networks for improving "time alignment". The alignment is done by matching some "key-points", and ANNs demonstrate to being successfully able to filter 91.6% of the correct matches compared to the 10.9% of the conventional algorithms.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis_Giacomo_Listrani.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Thesis Giacomo Listrani
Dimensione
34.41 MB
Formato
Adobe PDF
|
34.41 MB | 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/10589/182717