Blast furnace hot metal temperature prediction, by mean of mathematical models, plays an interesting role in blast furnace control, helping plant operators to give a faster and more accurate answer to changes in blast furnace state. In this work, the development of parametric models based on neural networks is shown. Time has been included as an implicit variable to improve consistency. The model has been developed departing from actual plant data supplied by Aceralia from its steel works located in Gijón.KEY WORDS: ironmaking; blast furnace; neural networks; forecasting; simulation; hot metal temperature.coke rate, was the variables selected as input. The control effort exerted over some of these input variables by means of step changes is clearly observable. For example, the correction applied to blast moisture during the period comprised between the 16 and 32 h it is significant, probably with the objective of restraining the steady decrease of hot metal temperature. Clearly visible as well is the strong correlation between pulverised coal injection (PCI) and oxygen enrichment in the blast. Furthermore, it is possible to check the increase of ore/coke rate in the burden at the same time that PCI increases until it reaches its set point. Figure 2 shows a more detailed view for hot metal temperature during the same period. In it, the vertical lines and circles represent the beginning of tapping and the time when hot metal temperature was measured, respectively. It is important to emphasise some features of this variable, which are will determine the way in which the problem of its prediction will be addressed.First of all, it is obvious that neither the tapping rate nor the time when the tapping hole is open is regular. In addition, the number of temperature measurements taken during tapping times and when they are taken are irregular too. Both depend on the plant operator's decision according to their estimation of the state of the blast furnace.In general, between two and three hot metal temperature measurements are taken by cast. The first is taken shortly after piercing the tap hole, the second when slag starts to flow out and the third near the end of the cast. The process computer records the data when a new measurement is carried out, and repeats this value until the next measurement. This is the reason why the signal obtained for hot metal temperature evolves stepwise.These specific features must be taken into account for dynamic modelling purposes. The fact that the value of the variable that is intended to forecast can not be measured with a sampling rate equal to the inputs variables sampling rate, makes it necessary to pre-treat it before being introduced in the model, in order to obtain regularly distributed values. This can be done by means of interpolation among the data available.7) Another possibility is to include the time explicitly in the model. This latter was the approach employed in this work. Model DevelopmentThe structure chosen for the model can be considered as belonging to a class of...
The present work presents a model based on fuzzy logic tools to predict and simulate the hot metal temperature in a blast furnace (BF). As input variables this model uses the control variables of a current BF such as moisture, pulverised coal injection, oxygen addition, mineral/coke ratio and blast volume, and it yields as a result of the hot metal temperature. The variables employed to develop the model have been obtained from data supplied by current sensors of a Spanish BF. In the model training stage the adaptive neurofuzzy inference system and the subtractive clustering algorithms have been used.
The paper will present the work done to assess the usefulness of the thermographic analysis in the design of a detection system for longitudinal defects in coils manufactured by a tandem cold mill. The approach started with a first phase devoted to the development of a computer system for the acquisition of high resolution thermal maps of the whole strip. In the second phase the thermal maps are classified and related to the process variables. s INTRODUCTIONThe pressing demand to improve the quality in the steel rolling industry requires the use of the latest technology or to explore new ones, in order to detect and avoid product defects (1, 2), and in this way, to make a better adjustment of the manufacturing parameters, producing better products.The work presented in this paper uses the latest hardware technology in infrared sensors to acquire the temperature of a tin plate strip during its rolling at a high frequency. The thermographic sensor uses a scanning system that allows the temperature measurement in several points of the strip, that is, scan acquisition, instead of spot acquisition. This sensor capability joined to the fast acquisition frequency, allows the temperature measurement of the entire strip while it is being rolled in the cold mill. Finally, a thermographic image of the strip is obtained by combining all the scans captured using the information about the speed of the strip while it was rolled.The goals of the work are to show the relationship between temperature and flatness, and the detection of longitudinal defects using the thermographic information obtained through the sensor (3).Longitudinal defects are mainly related to flatness problems produced by miscellaneous factors in the processing of the cold mill, like refrigeration, or rolling strength. If the defects are detected while the strip is being processed, the rolling parameters can be modified in order to solve the problem and improve the quality of the final product.s THERMOGRAPHY ACQUISTION Description of the systemThe system is composed of four elements, shown in figure 1, whose description and functionality is briefly presented below.• An infrared sensor, specially developed by LAND Company for this application (4, 5), scans the temperature across the strip 80 times per second. During each scan, the sensor generates an analog signal proportional to the temperature of the strip.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.