Page images


Session Co-Chairmen

Alan W. Cramb, Carnegie-Mellon University
Jan Kor, Timken Company


J. Kor, Timken Company
P. Koros, LTV Steel Company
J. Fay, ASARCO, Inc.
A. McLean, University of Toronto
D. Hardesty, Sandia National Laboratory
C. Alcock, Notre Dame University
Y. Kim, Lehigh University
R. Guthrie, McGill University
M. Shah, IBM
S. Ray, NIST


A new generation of processes for direct liquid metal production is evolving. These processes will be continuous rather than batch and involve fewer processing steps, resulting in substantially lower capital and operating costs.

The objective for this session is to define the role that modern techniques for intelligent processing, i.e., sensors, modeling, and computer process control, should play in these major process developments. Based on this assessment, specific recommendations should be made for further cause of action.

The session was organized into two parts: the first part was an information exchange between experts on sensor technology, process control, modeling, and the application of advanced computer decision making techniques', while the second part was to define and prioritize specific needs in the area of intelligent processing of liquid metals.

This phrase used instead of terms "artificial intelligence" and "expert systems" which were

overused in this context.

Three major thrust areas were determined: Sensors, Process Modeling and Intelligent Processing. Within each area specific needs were identified. The sensor area was considered to be the area of highest importance and it was the consensus of the group that sensor development and implementation should be the major focus of any endeavor.


Three separate groups of sensors were distinguished: continuous temperature sensors; continuous chemical sensors for liquid metal and hot, dirty gases; and, physical sensors to measure reaction intensity.

Continuous Temperature Measurement

This is an area of immediate need. Discussions indicated that this is not necessarily a problem of temperature measurement itself but more a problem of sensor life in the steel-making environment. It was noted that reactions at the metal/slag interface can be particularly severe when continuous liquid steel measurements are made, and that abrasion in hot, dirty gases can also be a problem. The capability of continuous liquid metal temperature measurement at temperatures between 1450 °C and 1700 °C to an accuracy of plus or minus 1 °C is necessary, while hot gas temperature measurement must be available up to temperatures of 1800 °C.

Continuous Chemical Sensors for Gases and Liquids

This is another area of immediate need and a list of identified sensor needs is given in table 1. Highest priority was given to continuous carbon determination; however, carbon monoxide, carbon dioxide, hydrogen and water contents of the off-gas were also given a high priority. Both of these sensor needs are immediate. A carbon sensor capable of measuring carbon from 0.005 to 4.0 weight percent would be optimum and cover all needs; however, a sensor capable of 0.1 to 4.0 percent would be a good starting point for process control. In gas chemical analysis the major anticipated problems were related to continuous sampling of the hot, dirty gas.

Physical Sensors of Conditions in the System

This area of need is again immediate from the point of view of reaction control; however, since an appropriate sensor could not be identified?, it was felt that this may be a longer term development. The goal of this sensor development is to develop a real time, on-line measurement of reaction intensity within the bath. A vessel sensor, using some physical measurement which can be made either on or within the vessel (such as a sonic or vibrational measurement) would then be interpreted to give an index of reaction intensity. It was agreed that any such sensor development would need tandem development of appropriate computerized pattern recognition techniques to be successful in a reasonable time frame. "Neural Networks" were mentioned in this context.


The area of process modeling was identified as the second most important area for research at this time. Two separate groups of needs were outlined:

(1) Process Control Models
(2) In-Detail Process Models.

Process Control Models

Due to the nature of a continuous reaction vessel, the need was identified for simple on-line process control models, based on sensor and operator input, which will control the operation and act as an operator guide. This is an area of immediate need and the system would be upgraded and developed during pilot plant operation.

In-Detail Process Models

A long range need was determined in the area of process modeling. Although a complete, fundamental heat, mass and fluid flow model of the process was not thought to be realistic, certain portions of the problem were worthy of independent development. Such portions were: heat, fluid flow, droplet size distribution in the foaming emulsion, bubble surface area, jet penetration and solids behavior.

2A number of ideas, documented in table 2, were discussed.


In the area of Intelligent Processing there was a concern that current "Artificial Intelligence" techniques might not be applicable to on-line control in a broad sense; however, in certain well-defined circumstances it may be useful. In addition, it was felt that the computer science involved in process control and decision making was sufficiently rapid that a mechanism should be set up so that appropriate advances can be implemented within the industry in a reasonable time frame. Three separate groups of needs were identified.

Operator Feedback and Instruction

This group is tied to the first group under MODELING. It was felt that "Artificial Intelligence" techniques might be appropriate in operator guidance systems.

System Integration

A problem identified generally with the implementation of computer techniques was standardization and portability. It should not be necessary for each company or plant to custom design their own systems. A group needs to be set up to develop and recommend standards in system integration. Appropriate areas of endeavor are communications, estimation techniques and numerical data bases.

Technology Transfer

To combat the problem of imminent development of appropriate computer technologies, it was felt that a "watch-dog" group should be set up to monitor emerging technologies and to act as a center of technology transfer when appropriate. It was felt that this group should be dynamic, i.e., a group that would be aggressive in attempts to find collaborators interested in industrial development of advances and also be involved in such implementation. The group would also become a center of knowledge in this area and a resource to the industry.

« PreviousContinue »