Komputerowe wspomaganie doboru materiałów na części maszyn

W pracy przedstawiono oryginalną metodykę modelowania zależności między składem chemicznym i hartownością stali konstrukcyjnych stopowych z wykorzystaniem sieci neuronowych. Bazując na wynikach doświadczalnych badań hartowności metodą Jominy'ego, opracowano i w pełni zweryfikowano doświadczalnie model sieci neuronowych, umożliwiający obliczanie krzywych hartowności Jominy'ego na podstawie składu chemicznego stali. Opracowano również i w pełni zweryfikowano numerycznie model sieci neuronowych, umożliwiający projektowanie składu chemicznego stali na podstawie znanego przebiegu krzywej hartowności Jominy'ego. Ponadto przedstawiono przydatność własnej metody modelowania krzywych hartowności do oceny wpływu pierwiastków stopowych na hartowność stali. Przeprowadzono symulację komputerową wpływu poszczególnych pierwiastków stopowych oraz ich kombinacji na hartowność stali oraz zamieszczono uzyskane wyniki.

---

The adequate models of relations between the chemical composition and the hardenability of the alloy constructional steels were developed, basing on the experimental results of the Jominy hardenability tests. The models employed the neural networks. The hardenability models developed are useful for forecasting Jominy hardenability curves shapes, basing on the knowledge of the steel chemical composition. The preliminary classification of the steels within the particular groups into six classes altogether proved to be sufficient for developing the neural network models obtained. It was showed, that the classification can be made basing on the alloy factor value AF. It was found, as the result of the verification procedure made in the paper, that the value of the coefficient of the evaluation of the calculation method adequacy s, which was employed in the paper, is always lower (1.5-2.4 HRC) for the new method than for the others (reaching even 10 HRC). The value of the coefficient is lower than the minimum value of the coefficient for the measurement errors committed during experiments (2.5 HRC). It was proved, basing on the comparative results, that the developed method of calculating of the hardenability curves employing the neural networks, guarantees the best conformance of the calculation results and the experimental data among the existing methods. The method of designing of the steel chemical composition basing on the required hardenability curve shape was also presented in the paper. It makes it possible to design the steel chemical composition basing on the assumed Jominy hardenability curve shape. The preliminary steel classification employed was the same as in case of the modelling of the hardenability. All generalizations are based on the vast set of the experimental data. The results of the tests carried out on about 400 heats were taken for neural networks' training in the case of each model. About 550 heats of the carburizing and heat-treatable steels with various chemical compositions were used for testing. The developed method of forecasting of the steel chemical composition basing on the knowledge of Jominy hardenability curve shape is useful in practice, e.g. for the real-time control of the chemical composition of the steel with the strictly demanded hardenability curve shape during the heating process. Moreover application of the presented method, using neural network models developed, enables a scientist to make free analyses of the effect of the alloying elements occurring in carburising and heat-treatable alloy constructional steels using only computer simulation, without having to carry out additional and expensive experimental investigations.