Ultra High Performance Concrete (UHPC) has garnered significant attention in previous research and applications due to its outstanding mechanical properties and durability. Given the inevitability of concrete cracking, enhancing the longevity of UHPC structures and reducing maintenance costs has become a viable solution through the integration of self-healing properties. However, current studies often overlook the evolution of self-healing performance of UHPC under harsh conditions. Therefore, this thesis aims to achieve three main objectives. Firstly, to explore how the simultaneous application of mechanical loads and exposure to harsh environments affect the self-healing capacity of UHPC. Secondly, to investigate the effectiveness and repeatability of self-healing performance of UHPC under multiple cracking/healing cycles. Lastly, to develop a model using machine learning techniques to predict the self-healing performance of the UHPC. The thesis begins with a comprehensive review of existing literature, covering UHPC design and its various properties. It then proceeds to examine the self-healing performance of UHPC under different environmental conditions. To address the first objective, the study compares the long-term development of self-healing in pre-cracked UHPC samples exposed to tap water, salt water, and geothermal water under sustained loading conditions. Various tests, including Ultrasonic Pulse Velocity testing (UPV), microscopic image processing, and double-edge wedge splitting tests, are employed to evaluate the progress of self-healing in UHPC. Additionally, Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray Spectroscopy (EDS) analyses are conducted to assess the impact of different exposure environments on self-healing products. The results indicate that the self-healing of UHPC increases with prolonged exposure time but is notably influenced by environmental factors. Samples exposed to salt water and geothermal water show lower self-healing abilities compared to those exposed to tap water. However, extending the exposure time partially alleviates the inhibitory effects of harsh environments on self-healing. Despite the influence of sustained loading and aggressive environments, the stiffness partially recovers through self-healing. Furthermore, AgNO3 solution spray test shows chloride ion penetration depth increases with crack width and exposure time, while permeability decreases over time due to the crack closure from self-healing effect. SEM and EDS analyses reveal that self-healing products in samples primarily consist of CaCO3, regardless of the environment. For the second objective, it is found in the study that the self-healing ability of UHPC increases after the first cracking/healing cycle if compared to samples without cracking/healing cycle. However, as samples undergo the second cracking/healing cycle, the level of self-healing significantly decreases. Despite multiple cracking/healing cycles being endured, UHPC samples exposed to aggressive environments still exhibit a significant degree of stiffness recovery, emphasizing the repeatability and effectiveness of self-healing mechanisms. Finally, for the third objective, the predictive performance of UHPC self-healing models developed using four traditional machine learning techniques (Support Vector Machine (SVM), Multilayer Perceptron (MLP), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM)) combined with two metaheuristic algorithms (Whale Optimization Algorithm (WOA) and Grey Wolf Optimization (GWO)) is developed and compared. Their effectiveness is validated using various evaluation metrics. The results indicate that the WOA-LGBM model emerges as the best model for predicting UHPC self-healing ability, with R2=0.9756, VAF=97.56, MSE=0.0020, MAE=0.0278 for the training set and R2=0.8899, VAF=89.33, MSE=0.0074, MAE=0.0669 for the test set. Through extensive experimentation and model development, this study aims to promote the use of UHPC materials for infrastructure projects in harsh environmental conditions, where the self-healing properties can enhance durability and reduce maintenance costs.
I calcestruzzi ad altissime prestazioni (UHPC) hanno suscitato notevole attenzione sia nell’ambito della ricerca, sia in applicazioni su varia scala grazie alle loro eccezionali proprietà meccaniche e di durabilità. Considerata l'inevitabilità della fessurazione del calcestruzzo, gli UHPC consentono di migliorare la longevità delle strutture e di ridurne i costi di manutenzione soprattutto grazie alle loro capacità di autoriparazione. Tuttavia, gli studi attualmente disponibili hanno in genere trascurato l'evoluzione delle prestazioni di autoriparazione dell'UHPC in condizioni avverse. Pertanto, questa tesi si propone di raggiungere tre obiettivi principali. In primo luogo, esplorare come l'applicazione simultanea di carichi meccanici e l'esposizione ad ambienti aggressivi influenzino la capacità di autoriparazione dell'UHPC. In secondo luogo, investigare l'efficacia e la ripetibilità delle prestazioni di autoriparazione dell'UHPC soggetto a cicli multipli di fessurazione e riparazione. Infine, sviluppare un modello utilizzando tecniche di machine learning per prevedere le prestazioni di autoriparazione dell'UHPC. La tesi inizia con una revisione esaustiva della letteratura esistente, che include la progettazione delle miscele di UHPC e le loro varie proprietà. Procede quindi ad esaminare le prestazioni di autoriparazione dell’UHPC in diverse condizioni ambientali. Per affrontare il primo obiettivo, lo studio valuta lo sviluppo a lungo termine dell’autoriparazione in campioni di UHPC pre-fessurati esposti ad acqua potabile, acqua salata e acqua di impianto geotermico con un carico costante contestualmente agente. Sono impiegate diverse prove sperimentali, tra cui la misurazione della velocità di transito di impulsi ultrasonici (UPV), l’analisi di immagini al microscopio e una specifica prova di trazione indiretta (DEWS), per valutare l’evoluzione dell’autoriparazione nell’UHPC. Inoltre, vengono condotte analisi al microscopio elettronico a scansione (SEM) e spettroscopia a raggi X a dispersione di energia (EDS) per valutare l’effetto dei diversi ambienti di esposizione sui prodotti di autoriparazione. I risultati indicano che la riparazione dell’UHPC aumenta con il prolungamento del tempo di esposizione ma è notevolmente influenzata dai fattori ambientali. I campioni esposti a acqua salata e acqua geotermica mostrano capacità di autoriparazione inferiori rispetto a quelli esposti ad acqua di rubinetto. Tuttavia, prolungare il tempo di esposizione allevia parzialmente gli effetti inibitori degli ambienti aggressivi sull’autoriparazione. Nonostante l’influenza del carico applicato e degli ambienti aggressivi, la rigidezza si ripristina parzialmente. Inoltre, la profondità di penetrazione dei cloruri, misurata mediante soluzione di nitrato d’argento (AgNO3), registra un incremento in funzione della larghezza della fessura e del tempo di esposizione, mentre la permeabilità diminuisce nel tempo grazie alla richiusura della crepa mediante autoriparazione. Le analisi SEM e EDS rivelano che i prodotti di autoriparazione nei campioni consistono principalmente in carbonato di calcio (CaCO3), indipendentemente dall’ambiente. Per quanto riguarda il secondo obiettivo, nello studio si è riscontrato che la capacità di autoriparazione dell'UHPC aumenta dopo il primo ciclo di ri-fessurazione/riparazione se paragonato alla pre-fessurazione. Tuttavia, la capacità di autoriparazione si riduce significativamente nei successivi cicli. Nonostante l’esecuzione di ripetuti cicli di fessurazione e riparazione, i campioni di UHPC esposti a ambienti aggressivi mostrano comunque un costante ripristino della rigidezza, sottolineando la ripetibilità e l'efficacia dei meccanismi di autoriparazione. Infine, per il terzo obiettivo, è stata valutata la capacità predittiva dei modelli di autoriparazione dell'UHPC sviluppando ed applicando quattro diverse tecniche di machine learning – nello specifico, Support Vector Machine (SVM), Multilayer Perceptron (MLP), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM) – combinate tramite due algoritmi metaeuristici (Whale Optimization Algorithm (WOA) e Grey Wolf Optimization (GWO)). La loro efficacia è stata validata utilizzando molteplici indicatori di accuratezza. I risultati indicano che il modello WOA-LGBM emerge come il miglior modello per la previsione della capacità di autoriparazione dell'UHPC, con R2=0.9756, VAF=97.56, MSE=0.0020, MAE=0.0278 per il set di addestramento e R2=0.8899, VAF=89.33, MSE=0.0074, MAE=0.0669 per il set di test. Attraverso un'ampia sperimentazione e lo sviluppo di modelli, questo studio mira a promuovere l'uso di materiali UHPC per progetti infrastrutturali in condizioni ambientali difficili, dove le proprietà di autoriparazione possono migliorare la durabilità delle strutture e ridurne i costi di manutenzione.
The evolution of self-healing performance of Ultra High Performance Concrete under harsh conditions : experimental investigations and machine learning modeling development
Xi, Bin
2023/2024
Abstract
Ultra High Performance Concrete (UHPC) has garnered significant attention in previous research and applications due to its outstanding mechanical properties and durability. Given the inevitability of concrete cracking, enhancing the longevity of UHPC structures and reducing maintenance costs has become a viable solution through the integration of self-healing properties. However, current studies often overlook the evolution of self-healing performance of UHPC under harsh conditions. Therefore, this thesis aims to achieve three main objectives. Firstly, to explore how the simultaneous application of mechanical loads and exposure to harsh environments affect the self-healing capacity of UHPC. Secondly, to investigate the effectiveness and repeatability of self-healing performance of UHPC under multiple cracking/healing cycles. Lastly, to develop a model using machine learning techniques to predict the self-healing performance of the UHPC. The thesis begins with a comprehensive review of existing literature, covering UHPC design and its various properties. It then proceeds to examine the self-healing performance of UHPC under different environmental conditions. To address the first objective, the study compares the long-term development of self-healing in pre-cracked UHPC samples exposed to tap water, salt water, and geothermal water under sustained loading conditions. Various tests, including Ultrasonic Pulse Velocity testing (UPV), microscopic image processing, and double-edge wedge splitting tests, are employed to evaluate the progress of self-healing in UHPC. Additionally, Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray Spectroscopy (EDS) analyses are conducted to assess the impact of different exposure environments on self-healing products. The results indicate that the self-healing of UHPC increases with prolonged exposure time but is notably influenced by environmental factors. Samples exposed to salt water and geothermal water show lower self-healing abilities compared to those exposed to tap water. However, extending the exposure time partially alleviates the inhibitory effects of harsh environments on self-healing. Despite the influence of sustained loading and aggressive environments, the stiffness partially recovers through self-healing. Furthermore, AgNO3 solution spray test shows chloride ion penetration depth increases with crack width and exposure time, while permeability decreases over time due to the crack closure from self-healing effect. SEM and EDS analyses reveal that self-healing products in samples primarily consist of CaCO3, regardless of the environment. For the second objective, it is found in the study that the self-healing ability of UHPC increases after the first cracking/healing cycle if compared to samples without cracking/healing cycle. However, as samples undergo the second cracking/healing cycle, the level of self-healing significantly decreases. Despite multiple cracking/healing cycles being endured, UHPC samples exposed to aggressive environments still exhibit a significant degree of stiffness recovery, emphasizing the repeatability and effectiveness of self-healing mechanisms. Finally, for the third objective, the predictive performance of UHPC self-healing models developed using four traditional machine learning techniques (Support Vector Machine (SVM), Multilayer Perceptron (MLP), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM)) combined with two metaheuristic algorithms (Whale Optimization Algorithm (WOA) and Grey Wolf Optimization (GWO)) is developed and compared. Their effectiveness is validated using various evaluation metrics. The results indicate that the WOA-LGBM model emerges as the best model for predicting UHPC self-healing ability, with R2=0.9756, VAF=97.56, MSE=0.0020, MAE=0.0278 for the training set and R2=0.8899, VAF=89.33, MSE=0.0074, MAE=0.0669 for the test set. Through extensive experimentation and model development, this study aims to promote the use of UHPC materials for infrastructure projects in harsh environmental conditions, where the self-healing properties can enhance durability and reduce maintenance costs.File | Dimensione | Formato | |
---|---|---|---|
Bin XI phd thesis .pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
10.53 MB
Formato
Adobe PDF
|
10.53 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/10589/220552