Most steelmaking ladle modelling strategies are based either on pure physical modelling or statistical relations. In this paper, a relatively simple model for steel temperature and ladle temperatures based on ordinary differential equations is presented and compared with a number of more complex ODE models. The model is semi-physical since the model structure is obtained mainly from physical relations, but the model parameters are calibrated using collected process data and a systematic calibration methodology. A main objective has been to obtain a model suitable for on-line purposes. Different model expansions are compared to the reference model. It is found that a simple model structure gives a good trade off between accuracy and complexity.
This brief considers modelling of steel and wall temperatures in steelmaking ladles used as liquid steel transportation vessels. Two measurement campaigns in order to collect data for model calibration and validation were conducted at the facilities of a Swedish stainless steel manufacturer. The process data were utilized in order to calibrate and validate an ordinary differential equation (ODE) model using a grey-box approach. The grey-box approach is justified since model structure and parameters are only partially known and few data are available for calibration. It is concluded that even though the model structure is non-complicated, the model shows promising accuracy for the steel temperature estimation. The model will be further evaluated in online simulations at the steel mill.
This paper describes the on-line implementation and evaluation of a model describing heat transfer in steelmaking ladles. The implemented model is based on ordinary differential equations which ensures a short computation time. Model calibration was performed utilizing a grey-box approach using real process data in order to describe the specific ladles used by a Swedish steel manufacturer. The main purpose has been to design and implement a simulation prototype for on-line simulation in the steel mill. To evaluate the model and the implementation, simulation of 14 steel charges was performed in real-time in the steel mill. Further evaluation of different models and different conditions is needed, although the performed experiment seems promising.
This paper deals with model predictive control using a grey box model. The process is modelled using a priori information and unknown parts are described as general continuous nonlinear functions, which are approximated by means of Taylor series including higher order terms. In the Taylor series, the partial derivatives are estimated from measured data by minimising the maximum likelihood function. The grey box model is used to design and tune a model predictive controller to control the pickling process. The process is a bottle neck in the production line and the simulation shows that the production can be increased approximately 15 % by using the controller.
This paper deals with a case study of grey box modelling where known parts are modelled using a priori information and unknown parts are described as general continuous nonlinear functions, which are approximated by means of Taylor series including higher order terms. In the Taylor series, the partial derivatives are estimated from measured data by minimising the maximum likelihood function. This approach is used to keep the number of estimated parameters low. The modelling procedure follows a structured approach including; basic modelling, data acquisition, model calibration, expanded modelling, stochastic modelling and model appraisal.
This paper deals with grey box modelling of an industrial process, in which known parts are modelled using a priori information and significant unknown parts are described as general continuous nonlinear functions. The modelling procedure follows a structured approach, which includes basic modelling, data acquisition, model calibration, hybrid expanded modelling, stochastic modelling and model appraisal. The general functions are approximated by means of the Taylor series including higher order terms, where the partial derivatives are estimated from measured data by minimising the maximum likelihood function. The Taylor series approach is used to keep the number of estimated parameters low in comparison with other nonlinear black box identification methods. The model is suitable for formulating algorithms to control the process, for example, a model predictive controller. The model can also be used to simulate various different production situations in order to improve the capacity of the total production line. Further, the relevant parts of the Taylor series can be used to explain in what way unknown process parts influence the behaviour of the process and give ideas for further investigation concerning the studied process
This paper deals with condition monitoring and failure diagnosis of a steel strip rinsing process. Modelling and identification of the process is based on a priori knowledge about the process and data from the process. Some parts of the process wear out slowly. It is not possible to measure the wear with any transducers. In the model, the worn parts are modelled explicitly and estimated on-line by an Extended Kalman Filter. The parameter estimation is used for supervision and as an advisory system for the process operators to decide which worn parts should be changed at the next planned stop. In addition to the normal wear, other types of abrupt failures may suddenly occur. It is not possible to detect these failures directly and the failures will give a biased parameter estimate and mislead the process operators into thinking, that a part subject to wear should be changed although it is performing well. Therefore, the condition monitoring system is complemented with a fault detection and diagnosis system, which distinguishes normal wear from sudden abrupt failures.
Within the process industry, many economic benefits can be achieved by controlling industrial processes at a supervisory level. From an engineering viewpoint, this means using increasingly sophisticated techniques and state-of-the-art theory to optimise process through put and deliver ever more exacting property specifications. Supervision and Control for Industrial Processes considers ways of improving and developing processes using these new methods. In particular, the book demonstrates the use of the grey box modelling method to find an appropriate process model which can be used for control, fault detection and isolation at a supervisory level. This publication will be of particular interest to engineers and graduate students involved in the areas of process control, supervision of industrial processes and failure diagnosis.
This paper deals with model based fault detection and isolation of a pickling process within the steel industry. The model is based on the grey box methodology and reflects the physical behaviour of the process. Possible faults are included in the model as parameters, which are estimated on line. The estimation is based on minimizing a loss function using past data from a defined moving time window. The procedure of finding the faults starts by estimate all defined fault parameters. One fault parameter is removed from the set of prospective list of faults by removing the parameter with the smallest saliency. The saliency is defined as the quote between the parameter estimate and the corresponding element of the inverse of the hessian matrix. The parameter with the smallest saliency gives a measure of the relevance of the estimated parameters relative all estimated parameters. The procedure is repeated until all fault parameters are eliminated from the list. To isolate the faults, the Akaike's Information Criterion (AIC) is used to detect faults. This gives the threshold when a fault relevant parameter is removed from the list of prospective faults. © 2010 IFAC.
This paper deals with methods and experiences of incorporating a priori knowledge into mathematical models of industrial processes and systems. Grey box modelling has been developed in several directions and can be grouped into branches depending on the way a priori knowledge is handled. In this paper we divide the groups into the following branches; constrained black box identification, semi-physical modelling, mechanistic modelling and hybrid modelling. Experiences from case studies demonstrate the different branches of grey box modelling procedures. In the applications, the grey box models have been used for model based control, soft sensors, process supervision and failure detection
This paper deals with modelling and identification of a river system using physical insights about the process, experience of operating the system and information about the system dynamics shown by measured data. These components, together, form a linear model structure in the state space form. The inputs of the prospective model are physical variables, which are not directly measured. However, the model inputs can be found by a nonlinear transformation of measured variables. Unknown parameters of the model are estimated from measured data. The modelling work focuses on the principle of parsimony, which means the best model approach is the simplest one that fit the purpose of the application. The goal of the model is to control the water level of the river where the water flow is mainly determined by the demand for energy generation produced by the hydropower stations along the river. The energy requirement increases in the morning and decreases in the evening. These flow variations, caused by the energy demand, have to be compensated by controlling the power plants downstream, in such a way that the water level between the power stations is guaranteed. Simulation of the control system by using an adaptive model predictive controller shows that the water levels vary less and can be maintained at a higher level than during manual control. This means that more electric power can be produced with the same amount of water flow.
In the steel industry, there are many processes that include measuring and control of temperatures. With higher demand on quality, increased production, and effective energy consumption, the use of noncontact temperature measuring techniques has increased. After the cooling section in a continuous annealing-pickling line, the strip temperature is estimated by using the grey box technique. Temperatures are measured in the cavity between the strip and the roller using radiation thermometers. A model is made for estimating strip temperature using the measured temperatures and knowledge of the physics of the process.
In the steel industry, there are many processes that include measuring and control of temperatures. With higher demand on quality, increased production and effective energy consumption the use of non-contact temperature measuring techniques has increased. A model is developed for estimating strip temperature using measured temperatures and knowledge of physics of the process.