Analysis of the Possibility of Using Selected Artificial Intelligence Algorithms for the Assessment of the Microstructure of Vermicular Cast Iron
DOI:
https://doi.org/10.24425/afe.2025.153807Abstract
This paper presents an analysis of artificial intelligence algorithms in the context of their applicability to the automatic analysis of microstructure images. In the example presented, reference is made to exemplary images of the microstructure of vermicular cast iron. A characteristic feature of this alloy is the shape of the graphite separations. The microstructure consists of elements that humans can learn to recognise quite simply. Developing an application that recognises ‘dark colour’ and ‘worm shape’ is no longer so straightforward. The determination of the ‘dark’ colour in the algorithm becomes problematic, because depending on the conditions under which the photo was taken (e.g. time: day/night), the actual intensity values are altered. A similar situation occurs in the determination of shape, which varies from case to case. Such a classification is very general and results in large differences between instances of the class. Even a term like ‘relatively large’ can change depending on the size of the graphite separation itself. A dark colour can be represented as a sudden change in image intensity, i.e. large values of the gradient modulus. The question arises: what happens if ‘dark’ can be more than one microcomponent, for example graphite and perlite. A good solution would be to define an associated set of features that would more precisely define just this component of the microstructure - that is, its shape, colour and surroundings. The paper uses the local feature paradigm to do this. Referring to the literature, it can be pointed out that [1] local features are referred to as non-small and specific parts of an image. Distinctive image features need to be distinguished in order to detect these places of interest. In this case, they are: edges, spots and ridges.References
[1] Kosa, E. (2012). Research on the structure of nitrided layers produced on EN-GJL 250 cast iron under glow discharge conditions. Engineering diploma thesis, Warsaw University of Technology, Faculty of Materials Engineering. (in Polish).
[2] Szpak, Ł. (2014). Image classification based on the author's hypsometric representation. Master's thesis, Warsaw University of Technology, Faculty of Electronics and Information Technology, Institute of Computer Science. (in Polish).
[3] de Albuquerque, V.H.C., Cortez, P.C., de Alexandria, A.R. & Manuel, J.R.S. (2008). A new solution for automatic microstructures analysis from images based on a backpropagation artificial neural network. Nondestructive Testing and Evaluation. 23(94), 273-283. https://doi.org/10.1080/10589750802258986.
[4] Han, Y., Lai, C., Wang, B. & Gu, H. (2019). Segmenting images with complex textures by using hybrid algorithm. Journal of Electronic Imaging. 28(1), 013030. DOI: 10.1117/1.JEI.28.1.013030.
[5] Fotos, G., Campbell, A., Murray, P. & Yakushina, E. (2023). Deep learning enhanced Watershed for microstructural analysis using a boundary class semantic segmentation. Journal of Materials Science. 58(36), 14390-14410. https://doi.org/10.1007/s10853-023-08901-w.
[6] Azlan Suhaimi, M., Park, K.H., Sharif, S., Kim, D.W. Saladin Mohruni, A. (2017). Evaluation of cutting force and surface roughness in high-speed milling of compacted graphite iron. In MATEC Web of Conferences, 7-9 June 2017 (pp. 1-7). DOI: 10.1051/MATECCONF/201710103016.
[7] Zhu, C. & Zhang, A. (2007). Experimental study on fracture toughness of vermicular cast iron. Journal of Mechanical Strength. 29(2), 310-314.
[8] Guzik, P. (2014). Methods of searching for characteristic points and their features. Engineering diploma thesis, Warsaw University of Technology Faculty of Electronics and Information Technology Institute of Computer Science. (in Polish).
[9] Nowik, A. (2014). Local descriptors using reduced binary histogram in similar image retrieval. Warsaw University of Technology, Faculty of Electronics and Information Technology Institute of Computer Science. (in Polish).
[10] Function Documentation. (2023). Color Space Conversions. Retrieved November 25, 2023, from https://docs.opencv.org/3.4/d8/d01/group__imgproc__colo r__conversions.html#ga397ae87e1288a81d2363b61574eb 8cab
[11] Color conversions. (2023) Retrieved November 23, 2023, from https://docs.opencv.org/3.4/de/d25/imgproc_color_conversions.html
[12] Smotking Images: Gaussian Blurring. (2023). Retrieved November 23, 2023, from, https://docs.opencv.org/4.x/d4/d13/tutorial_py_filtering.html
[13] Morphological Transformations: Closing. (2023). Retrieved November 23, 2023, from https://docs.opencv.org/4.x/d9/d61/tutorial_py_morphological_ops.html
[14] Structural Analysis and ShapeDescriptors: findContours. (2023). Retrieved November 23, 2023, from, https://docs.opencv.org/3.4/d3/dc0/group__imgproc__shape.html#ga17ed9f5d79ae97bd4c7cf18403e1689a
[15] Arithmetic Operations on Images: Bitwise Operations. (2023). Retrieved November 23, 2023, from, https://docs.opencv.org/3.4/d0/d86/tutorial_py_image_arithmetics.html
[16] Image Thresholding: Otsu's Binarization. (2023). Retrieved November 23, 2023, from, https://docs.opencv.org/3.4/d7/d4d/tutorial_py_thresholding.html
[17] About Keras: Keras & TensorFlow 2. (2023). Retrieved November 26, 2023, from https://keras.io/about/tf.keras.preprocessing.image.ImageDataGenerator, Retrieved 2023-11-26, https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator#used-in-the-notebooks
[18] fchollet, The Sequential model. (2023). Retrieved November 26, 2023, from https://keras.io/guides/sequential_model/
[19] Conv2D layer. (2023). Retrieved November 26, 2023, from https://keras.io/api/layers/convolution_layers/convolution2d/
[20] MaxPooling2D layer. (2023). Retrieved November 23, 2023, from https://keras.io/api/layers/pooling_layers/max_pooling2d/
[21] Flatten layer. (2023). Retrieved November 27, 2023, from https://keras.io/api/layers/reshaping_layers/flatten/
[22] Dense layer. (2023). Retrieved November 27, 2023, from https://keras.io/api/layers/core_layers/dense/
[23] Layer activation functions. (2023). Retrieved November 26, 2023, from, https://keras.io/api/layers/activations/#sigmoid-function
[24] Adam. (2023). Retrieved November 26, 2023, from https://keras.io/api/optimizers/adam/
