Wyniki wyszukiwania - BazTech

1

Optimization of Ferrite Stainless Steel Mechanical Properties Prediction with artificial Intelligence Algorithms

Honysz R.

Archives of Metallurgy and Materials

|

2020

|

Vol. 65, iss. 2

749-753

EN

The article presents a computational model build with the use of artificial neural networks optimized by genetic algorithm. This model was used to research and prediction of the impact of chemical elements and heat treatment conditions on the mechanical properties of ferrite stainless steel. Optimization has allowed the development of artificial neural networks, which showed a better or comparable prediction result in comparison to un-optimized networks has reduced the number of input variables and has accelerated the calculation speed. The introduced computational model can be applied in industry to reduce the manufacturing costs of materials. It can also simplify material selection when an engineer must properly choose the chemical elements and adequate plastic and/or heat treatment of stainless steels with required mechanical properties.

2

Prediction optimization of mechanical properties of ferrite stainless steels after forging treatment with use of genetic algorithms

Honysz R.

Archives of Materials Science and Engineering

|

2019

|

Vol. 100, nr 1/2

13--20

EN

Purpose: The paper describes the use of artificial neural networks to research and predict the effect of chemical components and thermal treatment conditions on stainless steel's mechanical characteristics optimized by genetic algorithm. Design/methodology/approach: The quantity of input variables of artificial neural networks has been optimized using genetic algorithms to enhance the prediction quality of artificial neural network and to enhance their efficiency. Then a computational model was trained and evaluated with optimized artificial neural networks. Findings: Optimization, with the exception of tensile strength, has enabled the creation of artificial neural networks, which either showed a better or similar performance from base networks, as well as a decreased amount of input variables As a consequence, noise data is decreased in the computational model built with the use of these networks. Research limitations/implications: Data analysis was required to confirm the relevance of obtaining information used for modelling to use in training procedures for artificial neural networks. Practical implications: Using artificial intelligence enables the multi-faceted growth of stainless steel engineering, even though there is only a relatively small amount of descriptors. Built and optimized computational model building using optimized artificial neural networks enables prediction of mechanical characteristics after normalization of forged ferritic stainless steels. Originality/value: In order to decrease production expenses of products, an introduced model can be obtained in manufacturing industry. It can also simplify the selection of materials if the engineer has to correctly choose chemical elements and appropriate plastics and/or heat processing of stainless steels, having the necessary mechanical characteristics.

3

Prediction optimization of mechanical properties of ferrite stainless steels after rolling treatment with use of genetic algorithms

Honysz R.

Journal of Achievements in Materials and Manufacturing Engineering

|

2015

|

Vol. 68, nr 1

17--24

EN

Purpose: The article discusses the use of artificial neural networks for research and prediction of the impact of chemical elements and heat treatment parameters on the mechanical properties of stainless steels optimized by genetic algorithm. Design/methodology/approach: To improve the quality of artificial neural network models and improve their performance the number of input variables of artificial neural networks has been optimized with use of genetic algorithms. Then a computational model build with optimised artificial neural networks were trained and verified. Findings: Optimization, except of tensile strength Rm case, has allowed the development of artificial neural networks, which either showed a better or comparable result from base networks, and also have a reduced number of input variables. As a result, in computational model constructed with use of these networks the noise information is reduced. Research limitations/implications: Data analysis was needed to verify if obtained data used for modelling are relevant to use them in artificial neural networks training processes. Practical implications: The use of artificial intelligence allows the multifaceted development of stainless steels engineering, even if only a small number of descriptors is available. Constructed and optimised computational model build with use of optimised artificial neural networks allows prediction of mechanical properties of rolled ferritic stainless steels after normalization. Originality/value: Introduced model can be obtain in industry to reduce manufacturing costs of materials. It can also simplify material selection, when engineer must properly choose the chemical elements and adequate plastic and/or heat treatment of stainless steels with required mechanical properties.