Microsphere Image Classification for FCC Catalyst Porosity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Characterizing the porosity and elemental composition of Fluid Catalytic Cracking (FCC) catalysts is difficult and time-consuming, making it challenging to analyze and optimize their performance accurately.
Innovation Solution
A microsphere analysis tool (MAT) using machine learning techniques to classify spatially-contiguous image regions of backscatter electron scanning electron microscopy (BSE-SEM) images, segmenting them into particles, intra-particle pores, and inter-particle voids, and characterizing porosity and elemental composition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional methods are used to characterize porosity and elemental composition of FCC catalysts, then measurement accuracy can be maintained, but the process becomes extremely time-consuming and difficult
Solution Approach 1:
The patent replaces traditional mechanical and chemical characterization methods with an automated image analysis system using machine learning. The system captures images of catalyst particles and uses trained classifiers to automatically determine porosity and elemental composition, eliminating manual analysis and significantly reducing characterization time while maintaining accuracy
Solution Approach 2:
The patent creates digital copies (images) of the catalyst particles and analyzes these copies to extract compositional information. By working with image data rather than physical samples, the system enables rapid, non-destructive characterization that can be performed on many particles simultaneously
2Measurement precision
If detailed characterization of FCC catalysts is performed to improve performance analysis, then accuracy increases, but time consumption increases significantly
Solution Approach 1:
The patent performs preliminary action by training machine learning classifiers on labeled datasets before actual characterization. Once trained, these classifiers can rapidly and accurately characterize new catalyst particles without requiring time-consuming manual analysis for each particle, achieving both high precision and speed in production mode
Solution Approach 2:
The patent segments the characterization task into distinct classification categories (e.g., different elemental compositions and porosity levels). The machine learning system is trained to recognize and classify different particle types based on image features, enabling parallel processing of multiple characterization attributes simultaneously
Data Source
AI summary
Methods, systems, and computer readable media are disclosed for characterizing the porosity of a microspheric material, training a classifier model for classifying spatially-contiguous image regions of backscatter electron scanning electron microscopy (BSE-SEM) images of a microspheric material according to the particle composition of the image regions, and characterizing the elemental composition of a microspheric material. The methods include: receiving image data representative of a microscopic images of a sample microspheric material; segmenting the image data into a plurality of spatially-contiguous image regions; classifying, by a trained machine-learning model, each of the plurality of spatially-contiguous image regions; and characterizing the microspheric material. The disclosure also relates to compositions for use as an FCC catalyst comprising microspheres which comprise alumina and/or clay and/or a zeolite.