W. Sitek

The modelling of hardenability using neural networks

Abstract This paper presents a novel method of the modelling of hardenability, its experimental-verification results, and comparison with the methods described in the literature. In addition, the potential of the new method of modelling the hardenability curves for the assessment of the effect of the chemical composition on steel properties is presented. Computer simulation was made of the influence of the particular alloying elements on hardenability and verification of the calculation results obtained was carried out using the experimental data.

The prototype of an expert system for the selection of high-speed steels for cutting tools

Abstract Experience gained by a team of researchers developing a database system designed as an aid for engineers and students who work out the tools made from the high-speed steels shows that this sort of system needs a flexible yet powerful user interface. To solve this task a front-end expert system module is being designed. Present design parameters assume that the intelligent add-on shall consist of some 500÷1000 rules enabling the user to compare the characteristics properties of various high-speed steels in respect to given applications, more, the technology generating module shall include the what-if option. To make all these new features as user friendly as possible it has been decided that the work should start with development of the linguistic task-oriented processor to enable the user to contact with the system in a subset of the natural language. To this end a prototype of a user interface was written in PROLOG and is presently being tested. In order to set the limits to the above mentioned task the dictionary was developed including all the necessary nouns, verbs, adjectives, numerals and sentence structures that are allowed to appear in the interface domain specific language.

The computer simulation of critical compressive stresses on the PVD coatings

The aim of this research is the computer simulation of critical compressive stresses on coatings TiN and Ti+TiN, obtained in the magnetron PVD process on the sintered high-speed steel of the ASP 30. Computer simulation of stresses was carried out in ANSYS environment, using the FEM method and the experimental values of stresses were determined based on the X-ray diffraction patterns. The computer simulation results correlate with the experimental results. The presented model meets the initial criteria, which gives ground to the assumption about its usability for determining the stresses in coatings, employing the finite element method using the ANSYS programme. In order to evaluate with more details the possibility of applying these coatings in tools, further computer simulation should be concentrated on the determination of other properties of the coatings for example microhardness. [Received 10 November 2006, Accepted 13 January 2007]

Designing of the chemical composition of constructional alloy steels

Abstract The paper presents a method of designing the chemical composition of constructional alloy steels based on the required Jominy hardenability curve shape and using neural networks. It presents a new method for the modelling of hardenability curves according to the chemical composition of the steel. Using this approach, a further method was developed—the method of designing of the chemical composition of the steel showing the required properties, including the required hardenability. Both methods are presented in the paper, as well as exemplary calculations using the computer tools developed.

Comparison of hardenability calculation methods of the heat-treatable constructional steels

Abstract Evaluation has been made of the consistency of calculation of the hardenability curves of the selected heat-treatable alloyed constructional steels with the experimental data. The study has been conducted basing on the analysis of present state of knowledge on hardenability calculation employing the neural network methods. Several calculation examples and comparison of the consistency of calculations for various calculation methods employed are included.

Modified Tartagli Method for Calculation of Jominy Hardenability Curve

Basing on the experimental results of the hardenability investigations, which employed Jominy method, the model of the neural networks was developed and fully verified experimentally. The model makes it possible to obtain Jominy hardenability curves basing on the steel chemical composition. The modified hardenability curves calculation method is presented in the paper, initially developed by Tartaglia, Eldis, and Geissler, later extended by T. Inoue. The method makes use of the similarity of the Jominy curve to the hyperbolic secant function. The empirical formulae proposed by the authors make calculation of the hardenability curve possible basing on the chemical composition of the steel. However, regression coefficients characteristic for the particular steel grade must be known. Replacing some formulae by the neural network models is proposed in the paper.

Application of neural networks for selection of steel grade with required hardenability

This paper presents the application of artificial neural network for the selection of steels grades with a required hardenability. The purpose has been achieved in two stages. In the first stage, a neural network model for calculating the Jominy curve on the basis of the chemical composition has been worked out. This model made it possible to prepare, in the second stage, a representative set of data and to work out the neural classifier for the selection of steel grade with the required hardenability. The methodology presented in this paper makes it possible to add new grades of steel to the system. The worked out model may be used in computer systems of steel selection for the heat-treated machine parts. [Received 15 April 2006; Accepted 15 March 2007]