• Results for steam rate simulations for A, B and C strikes indicate the different times taken to complete the strike due to the different boil-on rates.

  • Based on these relationships simple correlations can be developed between the average massecuite production rate and steam rate for each pan.

  • The steady-state model of the pan stage has been developed to calculate average flow rates of process streams using mass balances at each vacuum pan, stock tank and fugal.

  • The model determines the average production rates of vacuum pan massecuite, C sugar remelt, molasses and sugar streams given the syrup purity and flow rate to the pan stage.

  • Crystal sizing determinations assist in calculating the necessary quantities of C sugar needed for the A and B pans to ensure final product sugar is of the required size.

  • Once the empirical model for each pan has been established then the boil-on rates for each feed stream at the different stages of the pan stage schedule can be determined by summing the liquor, A molasses and B molasses feed rates for all the pans at that point in the schedule.

  • Given the expected liquor production rate, C sugar remelt production rate and raw wash return from the refinery to the liquor tank during this interval, the predicted tank levels can be determined for the liquor tank.

  • Similarly the predicted tank levels for the A and B molasses streams can be calculated from the production rates of the molasses at the centrifugal station and the sum of the consumption rates on the individual pans at a specific point in the pan stage schedule.

  • Stock tank level prediction requires the utilization of the previously presented models.
    A forward prediction of stock tank levels was made drawing from cane receival, stock tank and vacuum pan control system data for Racecourse mill on 03/09/2004.

  • Using forward prediction of vacuum pan states and the key process models discussed, a 90 minute future forecast of stock tank levels was made utilizing a rolling error average over the three previous 15 minute intervals. 

  • High correlation between predictions and actual tank levels.

  • The overall strategic management of the pan stage is quite difficult.

  • Often the pan stage is managed in a sub-optimal manner because an overview of operations encompassing various sections is not available.

  • The system seeks to provide a unifying structure to assist pan stage operators in making early decisions with respect to sugar quality, sugar recovery and minimization of steam consumption on the total pan stage.

  • The smart supervisory control system is a hybrid fuzzy expert system incorporating fuzzy logic, explanatory capabilities and industrial process models of the crystallisation stage.

  • Presently the most difficult part of process modelling has been completed and the construction of the advisory system is under way.

  • The system hopes to contribute both to the field of fuzzy logic based expert system design and to the development of industrial process models for the crystallisation stage within a sugar factory.

  • The project contributors would like to thank the staff at Racecourse Sugar Mill (Mackay, Australia) for their cooperation and support.

  • Support for this research is provided by means of an ARC Industry Linkage grant and PhD scholarship.

  • The funding assistance provided by SRI is also acknowledged and appreciated.