## Description

This work reports on the vibrations of an argon-oxygen decarburization (AOD) converter and its causative dynamics excitations, a topic of great interest for metallurgists. Hence, more physical insight is available, especially in the processes of decarburization and reduction. Existing converters are equipped with accelerometers and strain gauges for the in situ screening of process and heat parameters. In order to improve the physical understanding of such processes numerical simulations of the system are performed in Abaqus. In order to reproduce the physical behavior of the plant the vessel with its trunnion ring, drivetrain, bearings, gearbox and foundation are included in the model. The model is verified by comparing eigenfrequencies and mode shapes of the physical system with the finite element model (FEM). The results show that the numerical model adequately represents the mechanical behavior of the converter. The first five eigenfrequencies, which are in the relevant frequency range, correspond well. Furthermore, the measuring setup is also included in the numerical model. Specific transfer functions are calculated by means of a 'steady-state-dynamics' analysis, where the input variables are the forces and torques in the center of the crucible of the converter. As a consequence of this modal reduction the efficiency of the computation is increased drastically. Modal damping is assumed for the whole model. The transfer functions are exported to Matlab and the transfer matrix is assembled. Based on the measured vibration signals and the computed transfer matrix an inverse calculation of the excitations is performed in the frequency domain. Therefore, the transfer matrix has to be inverted at each sampling frequency. In typical industrial applications over-determined systems occur, where the number of outputs exceeds the number of inputs. In such a case the transfer matrix is not quadratic any more and a Moore-Penrose Pseudoinverse can be calculated. The computed input variables are transformed back into the time domain by an inverse fast-fourier-transform. Then these signals are used in a forward dynamics simulation ('modal dynamics') in Abaqus. By using this chain of inverse and forward simulations the measured vibrations can be reproduced exactly. If nonlinearities are included in the model, several iterations are necessary in order to attain convergence. As convergence criterion the root mean square error of simulation outputs and target signals is used. The speed of convergence depends on the type of nonlinearity and the operating point where the model is linearized. The dynamic causative forces and torques are available in the time domain as well as in the frequency domain. Based on these signals the forces which are introduced into the foundation are calculated. Furthermore, the movement of the converter plant is known. The whole procedure is carried out for several process steps and statistical data can be obtained for the metallurgists. Now a couple of parameters like process step, blow rate, carbon content and silicon or manganese content with significant influence on the level of vibration can be identified.Period | 9 Nov 2010 |
---|---|

Event title | Simulia Austria Regional User's Conference: null |

Event type | Conference |

Location | Vienna, Austria |