knowledge transfer


Intelligent Control of a Rotary Kiln

Submitted by K. Leiviska, 26.06.2001, IBA A


The lime kiln process is inherently difficult to operate efficiently because of complex dynamics and multi-variable process with non-linear reaction kinetics, and long time delays. During its operation many interconnected variables must be considered and control actions must be designed to meet multiple and sometimes conflicting objects, and changing operating conditions. Some measurements are unreliable, and the kiln characteristics may change during a long run. The operation may also be upset by disturbances such as changes in the composition and/or properties of the lime mud. In addition, certain process variables must be maintained within predefined constraints in order to assure the safe operation.


The kiln process has been extensively studied at the Wisaforest mill. A kiln control system based on fuzzy logic was developed already during 1993. Encouraged by good results, the research was continued with the main emphasis on applying a novel linguistic-equation approach for fuzzy modelling and simulation of the kiln process. The original knowledge-based fuzzy system was modified in order to take into account supervisory control and adaptation to changing operating conditions.

The LECont concept was implemented in G2. The compact system corresponds to a three-level cascaded controller:

  • The basic PI type LE controller handles the normal operation with symmetrical membership definitions.
  • The operation condition controller changes the control surface of the basic LE controller by modifying the membership definitions for the change of control variable D u.
  • The predictive LE controller changes the membership definitions for the derivative of the error D e. This level contains both the braking and asymmetrical actions.

Neural networks are used in defining the set points for the control loops.

Status and results

The system has been in continuous operation since the beginning of 1999. According to the statistics, the mean value of excess oxygen has been reduced by about 15% and the quartile range has been reduced by more than 20%. The mean value of the hot-end temperature has been reduced by nearly 40C and the quartile range and the standard deviation have declined by nearly 50% and more than 30%, respectively. The mean of the specific heat energy consumption (5.5 GJ/tCaO) was nearly 7% lower than the respective value during manual operation and a decrease of over 10% in total reduced sulphur (TRS) emissions and a reduction of about 50% in the proportion of high emission periods were recorded.

Adaptivity and portability

A basic linguistic equation controller corresponds to a normal FLC. The strength of actions is controlled by working point models, which could be considered as gain scheduling algorithms using a specific LE structure. The braking and asymmetrical actions correspond to predictive switching control. The use of the system in other kilns requires the tuning of models and controllers.

More information

E. Juuso, M. Jarvensivu, O. Ahava, 2001,Intelligent Supervisory Control of an Industrial Rotary Kiln. In: Industrial Applications of Soft Computing (K. Leiviska, Ed.). Studies in Fuzziness and Soft Computing, Vol. 71, Springer Verlag,.

technical reports & case studies

executive summaries

scientific papers




search eunite with