Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Moisture- and mould-resistance: multi-modal modelling leveraging X-ray tomography in edge-sealed cross-laminated timber
Luleå University of Technology, Department of Engineering Sciences and Mathematics, Wood Science and Engineering.ORCID iD: 0000-0001-7091-6696
Division of Building Physics, Department of Building and Environmental Technology, Lund University, Lund, Sweden.
Luleå University of Technology, Department of Engineering Sciences and Mathematics, Wood Science and Engineering.ORCID iD: 0000-0002-3544-8716
Luleå University of Technology, Department of Engineering Sciences and Mathematics, Wood Science and Engineering.ORCID iD: 0000-0003-2247-674x
2023 (English)In: Materials & design, ISSN 0264-1275, E-ISSN 1873-4197, Vol. 230, article id 111967Article in journal (Refereed) Published
Abstract [en]

Edge-sealing, which involves treating the edges of wood products, improves water resistance. This study investigated the feasibility of edge-sealed cross-laminated timber (CLT) panels to reduce capillary water uptake, thereby resisting mould formation. The water and vapour permeabilities of ten characteristically different single-layer sealant coating systems were systematically determined. Multi-modal assessment leveraged by computed tomography (CT) scanning methodology was used to enhance detection of material characteristics beyond the standard coating permeability assessment. Moisture content was observed to change during the specimens’ absorption and desorption depending on the sealant system applied. The results revealed different characteristics of coatings during the water absorption and desorption stages. Findings from this study were used to develop recommendations regarding the water resistance of coating systems, curing time, susceptibility to mould formation, and industrial applicability. Results suggest that edge-sealed CLT could minimise the risk of mould formation, which can occur at worksites with minimal weather protection. The method developed in this study provides a basis to evaluate new coating systems and determine which use case is the best for a particular coating type. This study also incorporates insights from industry to identify future research orientations, which may pave the way for new designs and assessment techniques.

Place, publisher, year, edition, pages
Elsevier, 2023. Vol. 230, article id 111967
Keywords [en]
CT scan, Full-field data, Image processing, Moisture simulation, Mould estimation, Multivariate modelling
National Category
Wood Science
Research subject
Wood Science and Engineering
Identifiers
URN: urn:nbn:se:ltu:diva-97626DOI: 10.1016/j.matdes.2023.111967ISI: 001041731500001Scopus ID: 2-s2.0-85159149308OAI: oai:DiVA.org:ltu-97626DiVA, id: diva2:1759893
Note

Validerad;2023;Nivå 2;2023-05-29 (joosat);

Funder: TräCentrum Norr (TCN), [grant number 239268, 239278]; FORMAS project: Experimental Studies of Capillary Phenomena in Bio-based Materials [grant number 942-2016-64]

Licens fulltext: CC BY License

Available from: 2023-05-29 Created: 2023-05-29 Last updated: 2025-10-21Bibliographically approved
In thesis
1. Data-driven Full-field Correlated Mechanics: Advancing Multi-modal Assessment Beyond Imaging Toward Artificial Intelligence Applied to Cross-laminated Timber Integrated with Material Processing
Open this publication in new window or tab >>Data-driven Full-field Correlated Mechanics: Advancing Multi-modal Assessment Beyond Imaging Toward Artificial Intelligence Applied to Cross-laminated Timber Integrated with Material Processing
2024 (English)Doctoral thesis, comprehensive summary (Other academic)
Alternative title[sv]
Data-driven heltäckande korrelerad mekanik : Främjande av multimodal bedömning bortom avbildning mot artificiell intelligens tillämpad på korslimmat trä integrerat med materialbearbetning
Abstract [en]

Innovations in materials science techniques are essential for advancing sustainable construction materials and achieving resource efficiency. Ideally, these novel approaches should replace traditional single-localised measurements with multidimensional, full-field data-fusion methods. A multi-sensor and data-driven approach can reveal hidden patterns and correlations in datasets that conventional methods might overlook. Guided by a vision to create a framework for material- and application-independent assessment, the work described here demonstrated a comprehensive approach. By prioritising the extension of measurement technologies, efforts were directed towards applications for cross-laminated timber (CLT), combined with material processing.

Hypothesis: An industrially feasible, performance-driven assembly strategy for CLT layers can be developed by integrating correlated full-field mechanics, with non-contact multi-modal methods based on artificial intelligence (AI) tools driven through machine learning (ML), to extract insights into variables that govern material mechanics. Basing this work on multidimensional image-based techniques will lead to a more complete understanding of the mechanical behaviour of composite materials such as CLT.

Novel applications of measurement techniques, together with innovative combinations of both measurement techniques and data processing methods, have been developed, to use less resources and provide added functionality to CLT. For this, various techniques for evaluating material properties have been investigated and compared. Digital image correlation (DIC) was used for full-field mechanics assessment and was applied in various formats, including static, quasi-static, cyclic, and high-speed experiments. DIC improved displacement and strain measurements across different scales, from full-scale sections down to tenths of millimetres, and made it possible to differentiate finer wood features with greater spatial and temporal resolution. CLT was subjected to loading conditions including in-plane compression and out-of-plane bending, which were investigated. Displacement and strain were assessed concerning longitudinal and transverse interlayer interactions, taking into account bonded or non-bonded adhesive interfaces, as well as natural features like knots, fibre deviations, heartwood, and sapwood. CLT with a ±45°-layered configuration provided greater stiffness and strength than conventional configurations, particularly in regions where greater shear resistance was required. Knots, traditionally seen as structural weaknesses in CLT, have here been shown to improve the shear resistance in transverse layers. A reduction in shear propagation in knotty timber means that it may be possible to revise timber standards to support the use of materials previously considered unsuitable for structural applications.

Multi-sensor data fusion was applied in material processing such as cutting, and DIC merged with chemical analysis revealed surface-densified set-recovery differences between earlywood and latewood in growth rings. Three different techniques – thermoelastic stress analysis (TSA) using thermal imaging data, the grid technique for tracking finer deformation patterns, and DIC – were augmented with finite element (FE) methods and compared with respect to their effectiveness in stress estimation. Multispectral and hyperspectral imaging were used to capture chemical and physical properties, localised measurements were made with a point-based near-infrared (NIR) probe, and full-field imaging was used to capture pixel-level data. Four-dimensional X-ray computed tomography (CT) was used on CLT to monitor internal moisture absorption and desorption. Hyperspectral NIR imaging in combination with X-ray imaging was used to classify the material properties of CLT. NIR was also used for mould detection and to compare the efficiency of CLT edge-sealing compounds. Moisture and its role in material processing were explored, and absorption and desorption experiments revealed that the edge-sealing of CLT could reduce the risk of mould formation.

Experimental procedures were optimised with factorial designs. MATLAB was integrated by generative AI-supported code generation and decentralised high-performance data processing. To enhance material assessment, multivariate data analysis (MVDA) and multivariate image analysis (MIA) were applied for multi-modal analysis by means of principal component analysis (PCA) and projection to latent structures – partial least squares discriminant analysis (PLS-DA). Hierarchical clustering analysis (HCA) was used to classify the impact of different clusters on modelling. Deep learning with a convolutional neural network (CNN) extracted additional refined features from the wood structure.

Mechanical properties were derived using multi-modal integration of image-based data, demonstrating the potential of combining traditional principles with modern technology to uncover previously unknown material characteristics. By integrating measurement techniques and incorporating AI, the work described in this thesis enhances material performance and resource efficiency in composite materials, in particular CLT.

Abstract [sv]

Innovationer inom materialvetenskapliga tekniker är essentiella för att främja hållbara byggmaterial och resurseffektivitet. Idealiskt bör dessa nya tillvägagångssätt ersätta traditionella enskilda lokaliserade mätningar med multidimensionella, heltäckande datafusionstekniker. En multisensorisk och datadriven ansats kan avslöja dolda mönster och korrelationer i datamängder som konventionella metoder kan förbise. Med en vision att skapa ett ramverk för material- och applikationsoberoende bedömning, demonstrerade detta arbete en omfattande metod. Genom att prioritera utökningen av mätteknologier riktades insatser mot applikationer för korslimmat trä (KL-trä), integrerat med materialbearbetning.

Hypotes: En industriellt genomförbar, prestandadriven sammanfogningsstrategi för KL-trä-lager kan utvecklas genom att integrera korrelerad heltäckande mekanik med beröringsfria multimodala metoder baserade på verktyg inom artificiell intelligens (AI) drivna av maskininlärning (ML), för att extrahera insikter om variabler som styr materialmekanik. Att basera detta arbete på multidimensionella, bildbaserade tekniker kommer att leda till en mer fullständig förståelse av det mekaniska beteendet hos kompositmaterial såsom KL-trä.

Nya tillämpningar av mättekniker, tillsammans med innovativa kombinationer av både mättekniker och databehandlingsmetoder, har utvecklats för att använda färre resurser och ge ökad funktionalitet till KL-trä. För detta ändamål har olika tekniker för att utvärdera materialegenskaper undersökts och jämförts. Digital bildkorrelation (DIC) användes för bedömning av heltäckande mekanik och tillämpades i olika format, inklusive statiska, kvasi-statiska, cykliska och höghastighetsexperiment. DIC förbättrade mätningar av förskjutning och töjning över olika skalor, från fullskaliga sektioner ner till tiondels millimetrar, och möjliggjorde att särskilja detaljerade trästrukturer med förbättrad rumslig och tidsmässig upplösning. KL-trä utsattes för lastförhållanden som inkluderade kompression i planet och böjning vinkelrätt mot planet, vilka undersöktes. Förskjutning och töjning bedömdes i longitudinella och tvärgående interlagerinteraktioner, inklusive limmade eller olimmade limfogar, samt naturliga egenskaper som kvistar, fiberavvikelser, kärnved och splintved. KL-trä med en ±45°-lagerkonfiguration uppvisade större styvhet och styrka än konventionella konfigurationer, särskilt i områden där större skjuvmotstånd krävdes. Kvistar, som traditionellt betraktas som strukturella svagheter i KL-trä, har här visat sig förbättra skjuvmotståndet i tvärgående lager. En minskning av skjuvutbredning i kvistigt trä antyder att det kan vara möjligt att revidera trästandarder för att stödja användningen av material som tidigare ansetts olämpliga för strukturella applikationer.

Multisensorisk datafusion tillämpades i materialbearbetning såsom skärning, och DIC kombinerat med kemisk analys avslöjade skillnader i återgång efter ytdensifiering mellan vårved och sommarved i årsringar. Tre olika tekniker – termoelastisk stressanalys (TSA) med hjälp av termisk bilddata, grid-tekniken för att spåra finare deformationsmönster, och DIC – förstärktes med finita elementmetoder (FE) och jämfördes avseende deras effektivitet i stressestimering. Multispektral och hyperspektral avbildning användes för att fånga kemiska och fysiska egenskaper; lokaliserade mätningar utfördes med en punktbaserad nära-infraröd (NIR) sond, och heltäckande avbildning användes för att fånga pixelnivådata. Fyrdimensionell röntgendatortomografi (CT) användes på KL-trä för att analysera intern fuktabsorption och fuktavgivning. Hyperspektral NIR-avbildning i kombination med röntgenavbildning användes för att klassificera materialegenskaperna hos KL-trä. NIR användes också för mögeldetektion och för att jämföra effektiviteten hos kantförseglingsföreningar för KL-trä. Fukt och dess roll i materialbearbetning utforskades, och absorption- och desorptionsexperiment visade att kantförsegling av KL-trä kan minska risken för mögelbildning.

Experimentella procedurer optimerades genom faktoriell design. MATLAB integrerades genom generativ AI-understödd kodgenerering och decentraliserad högpresterande databehandling. För att förbättra materialbedömningen tillämpades multivariat dataanalys (MVDA) och multivariat bildanalys (MIA) för multimodal analys genom huvudkomponentanalys (PCA) och projektion till latenta strukturer – partiell minstakvadraters diskriminantanalys (PLS-DA). Hierarkisk klusteranalys (HCA) användes för att klassificera påverkan av olika kluster på modellering. Djupinlärning med ett konvolutionellt neuralt nätverk (CNN) extraherade ytterligare förfinade egenskaper från trästrukturen.

Mekaniska egenskaper härleddes genom multimodal integration av bildbaserade data, vilket demonstrerade potentialen i att kombinera traditionella principer med modern teknik för att upptäcka tidigare okända materialegenskaper. Genom att integrera mättekniker och inkorporera AI, förbättrar arbetet som beskrivs i denna avhandling materialprestanda och resurseffektivitet i kompositmaterial, med särskilt fokus på KL-trä.

Place, publisher, year, edition, pages
Luleå University of Technology, 2024
Series
Doctoral thesis / Luleå University of Technology, ISSN 1402-1544
Keywords
AI, AI-Driven Material Assessment, Artificial Intelligence, Building, Building Materials, Classification, CLT Assembly, CLT Manufacture, Computer Vision, Construction, Crosslam, Computed Tomography, CT, Data Science, Data Mining, DIC Analysis, DIC, Digital Image Correlation, Digital Image Processing, Digital Speckle Photography, Experimental Mechanics, Finite Element Method, FE, Full-Field Mechanics, Generative AI, Hyperspectral Imaging, Image Analysis, Imaging, Interdisciplinary Research, Laminated Wood Product, Machine Learning, Mass Timber Engineering, Materials Science, Modelling, Multidisciplinary, Multidisciplinary Research, Multivariate Data Analysis, MVDA, Near-Infrared Spectroscopy, NIR, Neural Network, Non-Contact Measurement, Non-Destructive, Optical Measurement, Panel Configuration, Partial Least Squares, PLS, Principal Component Analysis, PCA, Projection to Latent Structures, Strain Localisation, Structural Analysis, Timber, Unsupervised Learning, Wood Anatomy, Wood Science, X-Lam, X-Ray, Alternativ byggmetod, Bildkorrelation, Hållbart byggande, KL-trä, Korslimmat trä, Massivträ, Skjuvtöjning, Träkonstruktion
National Category
Wood Science
Research subject
Wood Science and Engineering
Identifiers
urn:nbn:se:ltu:diva-81437 (URN)978-91-7790-715-2 (ISBN)978-91-7790-716-9 (ISBN)
Public defence
2024-12-17, A193, Luleå University of Technology, Skellefteå, 09:00 (English)
Opponent
Supervisors
Available from: 2020-11-18 Created: 2020-11-18 Last updated: 2025-10-22Bibliographically approved

Open Access in DiVA

fulltext(8890 kB)589 downloads
File information
File name FULLTEXT01.pdfFile size 8890 kBChecksum SHA-512
10a533ca18092213dcbe725c61e500e09c57edfedb957aef547792773ec820b4545559ca233f94b6d40d4e6108d2cb3a3983b7e873c38660304326cc4d87d9b4
Type fulltextMimetype application/pdf

Other links

Publisher's full textScopus

Authority records

Buck, DietrichSehlstedt-Persson, MargotÖhman, Micael

Search in DiVA

By author/editor
Buck, DietrichSehlstedt-Persson, MargotÖhman, Micael
By organisation
Wood Science and Engineering
In the same journal
Materials & design
Wood Science

Search outside of DiVA

GoogleGoogle Scholar
Total: 591 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 1149 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf