A non-linear model predictive controller (MPC) was designed to automate and optimise the control of a continuous annealing process. The process is known as a continuously transitional and time delay system. By employing a first principle furnace model, the controller computes the required furnace load and line speed for every control step. A non-linear constrained dynamic optimisation problem is solved in real time by applying piecewise linear models over a receding horizon. The new controller has been tested and implemented on a 400 000 t continuous annealing line. Production using this controller has proved its reliability and yields a broad improvement in temperature accuracy, throughput increase and energy conservation. The details of the furnace model and the validation results are described, followed by the MPC problem formulation, system execution and implementation results.

The European Steel industry is spending considerable efforts in order to improve the socio-economic and environmental sustainability of its processes by promoting any development, which can increase efficiency and lower the environmental impact of the steel production processes. In particular, the European iron and steel sector is strongly committed toward the reduction of energy consumptions and CO2 emissions. Process gases are a very valuable resource: possibilities exist to consider these gases as an intermediate by-product for the production of other valuable energy carriers or products with an associated environmental benefit. Therefore, the process gas networks, especially inside the integrated steelworks, have a fundamental function, as they allow meeting the demand of many processes and producing energy through dedicated facilities. They can also support the production processes by internal electric energy generation and often by supplying energy outside the plant boundaries. On the other hand, such networks are very complex systems interacting with many different production steps and the management of such complex systems is a very difficult task, where many often-counteracting factors need to be jointly taken into account. This paper presents the first outcomes of the research project entitled “Optimization of the management of the process gas network within the integrated steelworks (GASNET)”, which aims at developing a Decision Support System helping the energy managers and other concerned technical personnel to implement an optimized off-gases management and exploitation considering environmental and economic objectives. A series of Key Performance Indicators has been elaborated, in order to monitor the efficiency of the gas management and the objectives of the optimization have been defined. The overall structure of the project and the ongoing work will also be outlined in the paper.

This article presents the application of a recent neural network topology known as the deep echo state network to the prediction and modeling of strongly nonlinear systems typical of the process industry. The article analyzes the results by introducing a comparison with one of the most common and efficient topologies, the long short-term memories, in order to highlight the strengths and weaknesses of a reservoir computing approach compared to one currently considered as a standard of recurrent neural network. As benchmark application, two specific processes common in the integrated steelworks are selected, with the purpose of forecasting the future energy exchanges and transformations. The procedures of training, validation and test are based on data analysis, outlier detection and reconciliation and variable selection starting from real field industrial data. The analysis of results shows the effectiveness of deep echo state networks and their strong forecasting capabilities with respect to standard recurrent methodologies both in terms of training procedures and accuracy.

No abstract

A numerical transient model was designed to describe the dynamics of the annealing process prior to coating in a hot dip galvanising (HDG) line. The model was integrated into a model predictive controller developed in a previous work, including an adaptation mechanism to retain the model accuracy in operation. Complex transient production was studied to formulate proper constraints as input for the optimisation problem to be solved in real time. Simulations and online tests were carried out to verify the transient model and the controller design principles. The new controller has been implemented on a HDG line which annually produces 400 000 t of coated strip products mainly for the automotive and construction markets. Production using this controller proves its ability to correctly predict the future operation and optimise the control automation of the heating, the cooling and the line speed. Detailed data analysis shows significant improvements in terms of strip transition control precision, temperature control accuracy (50% increase inside tolerances), throughput maximisation (3-8% increase) and production consistency through the full automation.