Grain size measurement and various characterization methodologies can pose challenges in determining trends or making important decisions in an industrial setting. Therefore, it is important to have a consistent method of interpreting and collecting data. The primary method for analysis in industrial or university environments typically involves manual labour, which can be tedious and time-consuming, leading to errors in measurement or sample preparation [1,2].

As a result, various solutions have been developed to overcome these setbacks. For instance, Bordas et al. utilised a neural network to analyse different methodologies. It has been shown that even for algorithms, identifying data can be challenging [2].

Additionally, the quality of the gathered data is an important factor [3]. Therefore, it is crucial to place great importance on other human factors, such as sample preparation.

The main goal of this work is to generate data for training and producing results using Python-programmed AI. The first step of the work will demonstrate various methods for sample preparation and compare the results obtained by different operators and AI. Subsequently, an evaluation will be conducted on the results obtained from different datasets previously measured within our databank. Finally, the various steps will be presented and weighed to understand how to create and train an AI to measure different microstructural features.

Referenzen

[1] H., Peregrina-Barreto., et al. Measurement, 2013, 1, 249-258.

[2] A. Bordas., et al. Molecules, 2022, 27, 15.

[3] K, S., Tan, et al. Energy and AI, 2024, 16