Prediction of preheat refractory temperature profile of a ladle furnace

Patent metadata

Field	Value
Publication number	US-12032359-B2
Application number	US-202217652125-A
Country	US
Kind code	B2
Filing date	Feb 23, 2022
Priority date	Feb 25, 2021
Publication date	Jul 9, 2024
Grant date	Jul 9, 2024

Abstract

The present disclosure addresses the technical problem of prediction of a preheat refractory temperature profile of a ladle furnace. Operational temperature of the ladle furnace, stability of sensors and placement make sensors not feasible. Computational Fluid Dynamics (CFD) simulations require large computation time and cannot be used for runtime applications in plants. The method and system of the present disclosure uses CFD modeling to carry out parametric study to generate data which is further processed to train an Artificial Neural Network (ANN) model that serves as a prediction model for predicting the preheat refractory temperature profile for at least a portion of the side refractory and at least a portion of the bottom refractory layer separately for which a new set of input data is obtained. The trained prediction model of the present disclosure provides a quick runtime prediction in plants.

First claim

What is claimed is:

1. A processor implemented method comprising the steps of:
obtaining, via one or more hardware processors, a plurality of training parameters comprising a set of geometrical parameters of a ladle furnace for processing a metal and a set of process parameters for associated ladle operations, wherein the ladle furnace is characterized by at least a side refractory and a bottom refractory;
identifying, via the one or more hardware processors, a set of key training parameters using an influence index computed based on the plurality of training parameters;
generating, via the one or more hardware processors, sample data corresponding to the identified set of key training parameters by dividing the identified set of key training parameters into a plurality of sets, wherein each of the plurality of sets represents the set of geometrical parameters and the set of process parameters;
generating, via the one or more hardware processors, simulated data comprising a preheat refractory temperature, an outer layer temperature and a preheat refractory temperature profile, for at least a portion of the side refractory and at least a portion of the bottom refractory, using a Computational Fluid Dynamics (CFD) model which receives the sample data as input;
transforming, via the one or more hardware processors, the preheat refractory temperature profile into a reduced form using a regression equation; and
training, via the one or more hardware processors, an Artificial Neural Network (ANN) model using the preheat refractory temperature, the outer layer temperature, the reduced form of the preheat refractory temperature profile and the sample data to obtain a trained ANN model.

2. The processor implemented method of claim 1 , wherein the set of geometrical parameters includes a range of values pertaining to at least (i) ladle top and bottom diameters, (ii) ladle height, (iii) plug position, (iv) plug diameters, (v) number of refractory layers, (vi) thickness of each of the refractory layers, and (vii) physical properties of each of the refractory layers including density and thermal conductivity; and the set of process parameters for ladle operations includes a range of values pertaining to at least (i) tapping time, (ii) initial tapping temperatures, (iii) slag height or slag weight, (iv) refining duration, (v) holding duration, (vi) casting speed and (vii) turnaround time.

3. The processor implemented method of claim 2 , wherein the step of identifying a set of key training parameters comprises:
selecting iteratively, a training parameter from the plurality of training parameters as input to the CFD model, the training parameter having a maximum value and a minimum value obtained from the range of values, until each of the plurality of training parameters has been selected in an iteration;
simulating output values by the CFD model, in each iteration, using the maximum value and the minimum value of the selected training parameter while values of remaining training parameters from the plurality of training parameters are maintained constant for a corresponding iteration, wherein the output values represent temperature of the metal at a top of the ladle furnace and a bottom of the ladle furnace respectively, corresponding to each of the maximum value and the minimum value of the selected training parameter;
computing the influence index for the top of the ladle furnace and the bottom of the ladle furnace as a percentage variation between output values simulated for the maximum value and the minimum value of the selected training parameter respectively; and
identifying the selected training parameter as a key training parameter in the set of key training parameters, if at least one of the influence index computed for the top of the ladle furnace and the bottom of the ladle furnace is greater than a predefined threshold or within a predefined range of threshold values, wherein the predefined threshold and the predefined range of threshold values are based on required accuracy of prediction of the preheat refractory temperature profile.

4. The processor implemented method of claim 1 , wherein the step of generating sample data is based on a Design of Experiments (DOE) method.

5. The processor implemented method of claim 1 further comprising: predicting, via the one or more hardware processors, a preheat refractory temperature profile for input parameters pertaining to the ladle furnace and the ladle operations of a plant using the trained ANN model, for at least the portion of the side refractory and at least the portion of the bottom refractory of the ladle furnace, wherein the step of predicting comprises:
outputting, the preheat refractory temperature and the outer layer temperature using the trained ANN model; and
predicting the preheat refractory temperature profile, using the outputted preheat refractory temperature, the outputted outer layer temperature and the reduced form of the preheat refractory temperature profile, for at least the portion of the side refractory and at least the portion of the bottom refractory of the ladle furnace.

6. A system comprising:
one or more data storage devices operatively coupled to one or more hardware processors and configured to store instructions for execution via the one or more hardware processors to:
obtain, a plurality of training parameters comprising a set of geometrical parameters of a ladle furnace for processing a metal and a set of process parameters for associated ladle operations, wherein the ladle furnace is characterized by at least a side refractory and a bottom refractory;
identify, a set of key training parameters using an influence index computed based on the plurality of training parameters;
generate, sample data corresponding to the identified set of key training parameters by dividing the identified set of key training parameters into a plurality of sets, wherein each of the plurality of sets represents the set of geometrical parameters and the set of process parameters;
generate, simulated data comprising a preheat refractory temperature, an outer layer temperature and a preheat refractory temperature profile, for at least a portion of the side refractory and at least a portion of the bottom refractory, using a Computational Fluid Dynamics (CFD) model which receives the sample data as input;
transform, the preheat refractory temperature profile into a reduced form using a regression equation; and
train, an Artificial Neural Network (ANN) model using the preheat refractory temperature, the outer layer temperature, the reduced form of the preheat refractory temperature profile and the sample data to obtain a trained ANN model.

7. The system of claim 6 , wherein the set of geometrical parameters includes a range of values pertaining to at least (i) ladle top and bottom diameters, (ii) ladle height, (iii) plug position, (iv) plug diameters, (v) number of refractory layers, (vi) thickness of each of the refractory layers, and (vii) physical properties of each of the refractory layers including density and thermal conductivity; and the set of process parameters for ladle operations includes a range of values pertaining to at least (i) tapping time, (ii) initial tapping temperatures, (iii) slag height or slag weight, (iv) refining duration, (v) holding duration, (vi) casting speed and (vii) turnaround time.

8. The system of claim 7 , wherein the one or more hardware processors are configured by the instructions to identify the set of key training parameters by:
selecting iteratively, a training parameter from the plurality of training parameters as input to the CFD model, the training parameter having a maximum value and a minimum value obtained from the range of values, until each of the pl

Assignees

• Tata Consultancy Services Ltd

Inventors

• Singh Umesh
• Basavarsu Purushottham Gautham

Classifications

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G05B2219/45132

  Forging press, combined with furnace · CPC title

• G06N3/048

  Activation functions · CPC title