J. Kusiak

Modelling of microstructure and mechanical properties of steel using the artificial neural network

Abstract The paper presents some results of the research connected with the development of new approach based on the artificial intelligence of predicting the volume fraction and mean size of the phase constituents occurring in a steel after thermomechanical processing and cooling. The independent variables in the model are austenite grain size and cooling rate over the temperature range of the occurrence of phase transformations. The dependent parameters are proeutectoid ferrite, Widmanstatten ferrite and pearlite fractions as well as ferrite grain size. Furthermore, a preliminary model for the prediction of mechanical properties was elaborated based upon the same idea.

Conventional and Multiscale Modeling of Microstructure Evolution During Laminar Cooling of DP Steel Strips

Physical and numerical simulations of the hot rolling and laminar cooling of DP steel strips are presented in the paper. The objectives of the paper were twofold. Physical simulations of hot plastic deformation were used to identify and validate numerical models. Validated models were applied to simulate the manufacturing of DP steel strips. Conventional flow stress model and microstructure evolution model were used in the hot deformation part. The approach to the complex systems analysis based on global thermodynamic characterization and detailed microstructure characterization was applied to determine equilibrium state at various temperatures. Finally, two numerical models were used to simulate kinetics of austenite decomposition at varying temperatures: the first, conventional model based on the Avrami equation, and the second, the discrete Cellular Automata approach. Plastometric tests and stress relaxation tests were used for identification of the hot rolling model for the DP steel. Dilatometric tests were performed to identify the phase transformation models. Verification confirmed good accuracy of all models. Validated models were applied to simulate the manufacturing of DP steel strips. Influence of technological parameters (e.g., strip thickness and velocity, active sections in the laminar cooling, and water flux in the sections) on the DP microstructure was analyzed. The cooling schedules, which give required microstructures were proposed. The numerical tool, which simulates manufacturing chain for DP steel strips is the main output of the paper.

Data Exploration Approach Versus Sensitivity Analysis for Optimization of Metal Forming Processes

Product properties for innovative materials, e.g. dual phase steels, require precise control of production processes. Difficulties in optimization of process parameters correspond with large number of control variables, which should be considered in the technology design. Sensitivity analysis allows evaluating the importance of all process inputs on the final properties of material. Information on the most important inputs is crucial for further design of the process. Application of sensitivity analysis requires detailed knowledge of the process phenomena as well as the definition of the mathematical model of the thermomechanical process. Furthermore, some sensitivity analysis algorithms are of the high computational cost. Presented work concerns possibility of the application of data exploration approach in evaluation of the importance of process inputs as the alternative for sensitivity analysis. Use of data mining algorithms eliminates necessity of mathematical model development, it also does not require any apriori knowledge about the process. Authors presents the comparison of sensitivity analysis and data exploration approach in evaluating relationships between inputs and outputs of the hot rolling for dual phase steel strips. The presented approach and the perspectives of the practical application could lead to significant decrease of time necessary for the computations of process design. The theoretical considerations are supplemented with the results of both types of analysis.

Filtering of Thermomagnetic Data Curve Using Artificial Neural Network and Wavelet Analysis

New methods of filtering of experimental data curves, based on the artificial neural networks and the wavelet analysis are presented in the paper. The thermomagnetic data curves were filtered using these methods. The obtained results were validated using the modified algorithm of the cubic spline approximation.

Industrial control system based on data processing

The goal of the work is presentation and discussion of the idea of innovative approach to industrial control system based on data processing. The key issue of proposed control system is the analysis of a history of considered industrial process, it means the analysis of registered data (process parameters and signals) during the past production. The system searches similarities among the current production period and registered past production episodes (episodes are atomic periods of production). Each of episodes is characterized by controlled and measured signals. An episode which is similar to the present period and which is characterized by the best possible value of quality criterion is being selected and becomes a pattern for control of the present production. The searching procedure was based on the multi-agent methodology, while the control function of the chosen episode was modeled using the artificial neural network. The developed idea of the control system was implemented and tested using the data obtained by simulation of the virtual industrial experiment.

Image Filtering Using the Dynamic Particles Method

The holistic approaches used for image processing are considered in various types of applications in the domain of applied computer science and pattern recognition. A new image filtering method based on the dynamic particles (DP) approach is presented. It employs physics principles for the 3D signal smoothing. The obtained results were compared with commonly used denoising techniques including weighted average, Gaussian smoothing and wavelet analysis. The calculations were performed on two types of noise superimposed on the image data i.e. Gaussian noise and salt-pepper noise. The algorithm of the DP method and the results of calculations are presented.

Computer Aided Design of Manufacturing of Fasteners - Selection of the Best Production Chain

Computer aided design of the manufacturing technology for the fasteners is presented the paper. The particular objectives of the work were twofold. The first objective is evaluation of applicability of various materials for fasteners. Analysis of different technological variants is the second objective of the research. In the material part, bainitic steels are considered as an alternative for the commonly used carbon-manganese steels. This is a continuation of [1,2]. Possibility of elimination of the heat treatment was evaluated. Main features of the new generation of bainitic steels are discussed briefly in the paper. Rheological models for all steels investigated in the project were developed. The models were implemented into the finite element code for simulations of drawing and multi stage forging. Simulations of various variants of manufacturing chain were performed and the best alternative was selected. Criteria for the selection composed dimensional accuracy and tool life. Industrial trials were performed for the selected cycle and the efficiency of this cycle was evaluated. Finally, the optimization task was formulated. However, solution of the optimization problem is costly at this stage and improvement of the efficiency of the formulation will be the objective of further research. References 1. Kuziak R., Skóra M., Węglarczyk S., Paćko M., Pietrzyk M., Computer aided design of the manufacturing chain for fasteners, Computer Methods in Materials Science, 11, 2011, 243-250. 2. Kuziak R., Pidvysots’kyy V., Węglarczyk S., Pietrzyk M., Bainitic steels as alternative for conventional carbon-manganese steels in manufacturing of fasteners - simulation of production chain, Computer Methods in Materials Science, 11, 2011, 443 – 462.

On Aggregation of Stages in Multi-criteria Optimization of Chain Structured Processes

This work is concerned with complex optimization problems which can be divided into multiple, multi-dimensional problems arranged linearly (as can be observed in the multi-stage industrial processes). The relations between complexity of the problem, level of aggregation of stages into larger groups, and efficiency of search for optimal solution were investigated.

Identification of Material Models of Nanocoatings System Using the Metamodeling Approach

Hard systems of nanocoatings deposited using PVD (physical vapor deposition) are used in the artificial heart prosthesis. Correct determination of nanomaterial parameters is crucial for accuracy of simulation. The objective of this work is identification of material parameters of nanocoatings in hard system using the inverse analysis based on the artificial neural network metamodeling. The inverse analysis was preceded by the development of the Finite Element Method (FEM) model dedicated to the nanoindentation test of the hard nanocoatings system. The performed sensitivity analysis is focused on determination of parameters, having the highest influence on FEM model response. The obtained, reliable FEM model was used next in the inverse analysis. The objective of that analysis was evaluation of the parameters of the individual layers of the nanocoating system. In order to decrease the computation time connected with the inverse analysis, the metamodeling approach was proposed. The used metamodel was based on the artificial neural network technique. The obtained results confirm the usefulness of the presented method in the identification of the material properties of the complex, nanocoating systems.

Artificial Neural Network Predictive System for Oxygen Steelmaking Converter

The main objective of the paper is the presentation of the static control model of steelmaking converter process based on the artificial neural network approach. The results of classical mass and energy balance as well as regression models are also presented. The developed artificial neural network predicts the temperature of the liquid metal and the volume of necessary oxygen blow. The ANN was trained and tested with the real industrial data measured in one of the Polish steel plants. The comparison of the ANN results with the classical calculations is presented.