TY - THES

T1 - Sensorless Temperature Control in a Metal Forming Process

AU - Löbl, Patrick

N1 - no embargo

PY - 2021

Y1 - 2021

N2 - In metal forming, the specimen temperature has a decisive influence on the forming process and the resulting material properties. However, the specimen temperature within the tool can scarcely be measured directly by physical sensors, which also makes temperature control difficult. Sensorless temperature control by means of virtual sensors is a solution to this problem and also very attractive for forming processes due to its low cost and possibilities of easy adaptation. The input-output data required for modelling a virtual sensor are also often already available from simulations in Finite Element Analysis software. These provide great cost and time savings compared to measurements conducted on real experiments, especially if variable time increments are used in the simulation. The goal of this thesis is to develop a sensorless temperature control system for a special forming process, which is to be based on a virtual sensor to estimate the specimen temperature and thus eliminate the need for a direct, physical measurement. The virtual sensor is to be implemented using a dynamic model. Input-output temperature data with uneven sampling intervals from finite element simulations of the specimen are available for modelling, and uniformly sampled measurement data for validation. The central research questions include the modelling of systems on the basis of input-output data with variable sampling intervals and the development of a suitable sensorless control system of the specimen temperature for the test bed. In order to fulfil these, a methodology for model construction was derived from literature and the global least squares system identification method was numerically implemented and applied. PID and model predictive control were evaluated for the control system and the implementation was carried out using a PID controller. The results were simulated and validated experimentally. A sensorless temperature open-loop control system was successfully implemented at the Equal Channel Angular Pressing (ECAP) test bed at the Chair of Metal Forming at the Montanuniversität Leoben. The use of the global least squares system identification method for parameter estimation has proven suitable for models in state-space representation with a defined structure by utilising input-output data with uniform as well as non uniform sampling intervals. As a result of the developed control system, the heating-up time on the test bed could be reduced to 30% of the time required in operation without the control system. A final, constant set point deviation of less than 1% was also made possible by the control system, whereas in operation without the control system the specimen temperature could only be kept constant for 100 - 150 s and much larger set point deviations were to be expected.

AB - In metal forming, the specimen temperature has a decisive influence on the forming process and the resulting material properties. However, the specimen temperature within the tool can scarcely be measured directly by physical sensors, which also makes temperature control difficult. Sensorless temperature control by means of virtual sensors is a solution to this problem and also very attractive for forming processes due to its low cost and possibilities of easy adaptation. The input-output data required for modelling a virtual sensor are also often already available from simulations in Finite Element Analysis software. These provide great cost and time savings compared to measurements conducted on real experiments, especially if variable time increments are used in the simulation. The goal of this thesis is to develop a sensorless temperature control system for a special forming process, which is to be based on a virtual sensor to estimate the specimen temperature and thus eliminate the need for a direct, physical measurement. The virtual sensor is to be implemented using a dynamic model. Input-output temperature data with uneven sampling intervals from finite element simulations of the specimen are available for modelling, and uniformly sampled measurement data for validation. The central research questions include the modelling of systems on the basis of input-output data with variable sampling intervals and the development of a suitable sensorless control system of the specimen temperature for the test bed. In order to fulfil these, a methodology for model construction was derived from literature and the global least squares system identification method was numerically implemented and applied. PID and model predictive control were evaluated for the control system and the implementation was carried out using a PID controller. The results were simulated and validated experimentally. A sensorless temperature open-loop control system was successfully implemented at the Equal Channel Angular Pressing (ECAP) test bed at the Chair of Metal Forming at the Montanuniversität Leoben. The use of the global least squares system identification method for parameter estimation has proven suitable for models in state-space representation with a defined structure by utilising input-output data with uniform as well as non uniform sampling intervals. As a result of the developed control system, the heating-up time on the test bed could be reduced to 30% of the time required in operation without the control system. A final, constant set point deviation of less than 1% was also made possible by the control system, whereas in operation without the control system the specimen temperature could only be kept constant for 100 - 150 s and much larger set point deviations were to be expected.

KW - System Identification

KW - Control Engineering

KW - Virtual Sensors

KW - Sensorless Control

KW - Non-uniformly sampled data

KW - PID Control

KW - Model Predictive Control

KW - Systemidentifikation

KW - Regelungstechnik

KW - Virtuelle Sensoren

KW - Sensorlose Regelung

KW - Daten mit ungleichmäßigen Sampling-Intervallen

KW - PID Regelung

KW - Model Predictive Control

M3 - Master's Thesis

ER -