Cements and concretes materials characterisation using machine-learning-based reconstruction and 3D quantitative...

Area of Science:

• Materials Science and Engineering
• Computational Science
• Geology and Geophysics

Background:

• 3D X-ray microscopy (XRM) is crucial for non-destructive materials characterization, enabling analysis of composition, structure, and failure mechanisms.
• Challenges in XRM include imaging large, dense, or high-resolution samples, impacting data acquisition and processing.
• Advancements in artificial intelligence (AI) and machine learning (ML) offer solutions for artifact detection, denoising, quantification, and upscaling in XRM data.

Purpose of the Study:

• To apply AI and ML-based reconstruction methods to improve XRM data acquisition and processing for cement and concrete samples.
• To demonstrate enhanced image quality, faster throughput, data upscaling, and quantitative phase identification in 3D.
• To resolve previously inaccessible features and streamline the characterization workflow for challenging materials.

Main Methods:

• Application of three AI/ML-based reconstruction approaches: DeepRecon Pro for image enhancement and denoising, DeepScout for data upscaling, and Mineralogic 3D for quantitative automated mineralogy.
• Utilizing XRM (tomography) for non-destructive 3D imaging of cement and concrete samples.
• Comparative analysis of AI/ML-enhanced data against traditional methods to assess improvements in quality, speed, and resolution.

Main Results:

• DeepRecon Pro significantly improved scan quality and throughput for thick cement/concrete cores via enhanced contrast and denoising.
• DeepScout successfully upscaled XRM data, enabling visualization of larger fields of view.
• Mineralogic 3D provided accurate 3D spatial characterization and quantification of mineralogical/phase components.

Conclusions:

• AI and ML-based reconstruction significantly enhance XRM data quality and streamline processing for challenging materials like cement and concrete.
• These advanced methods enable the resolution of finer details and improve the quantitative analysis of material phases in 3D.
• The integrated workflow accelerates sample throughput and expands the capabilities of XRM in materials characterization.