Database for thermal cracking of wall-on-slab structures

A dataset was compiled of approximately 200 cases from laboratory and in-situ studies on externally restrained RC members. The majority of the data were compiled from own research materials and experimental records of the authors, supplemented by selected values from published studies where available. Each entry in the database corresponds to a structural wall element with associated geometric, material, reinforcement, restraint, and cracking-related parameters.

The database was compiled manually from project documentation, measurement records, and laboratory results. It includes 196 records representing 78 bridge structures, 72 tanks and containment structures and 46 laboratory specimens. For in-situ cases, input parameters were taken from construction drawings, design notes, and inspection reports, while crack widths were derived from monitoring records. For laboratory cases, values were extracted directly from test reports. A detailed set of features was included for each record, covering:

• Structural and geometric parameters: wall thickness, length, height, foundation dimensions, L/H ratio, Aw/Af ratio.

• Material parameters: concrete class (expressed with characteristic and, wherever available, measured mean compressive strength), cement type, cement content.

• Reinforcement details: reinforcement size and spacing, reinforcement ratio, concrete cover.

• Restraint conditions: type and number of restrained edges.

• Response measures: measured or calculated thermal, shrinkage and total strain, observed maximum crack width, crack/no-crack classification.

To improve model interpretability and reduce redundancy, raw variables were combined into meaningful ratios. Instead of directly using wall length and height, the L/H ratio was employed; similarly, wall area to foundation area and reinforcement ratio were derived from basic input dimensions. All variables were harmonized to consistent units and definitions to enable like-for-like comparison across sources. 

Several records contained incomplete information. While full geometric characteristics, reinforcement details, and the design concrete class were available in all cases, some material-related parameters were missing, such as concrete mix composition and experimentally confirmed mechanical properties. These gaps were supplemented using typical values corresponding to the given concrete grade and the function of the structural element. Strain measurements were not always available; in such cases, shrinkage strain was calculated according to the provisions of Eurocode 2:2004, and thermal strain was estimated following the guidelines of CIRIA C766. The imputation procedure was deliberately conservative, aiming to maintain dataset consistency without introducing artificial variance.