| In modern large steel plants, the steelmaking-continuous casting (SCC) process consists of multiple converters, refining furnaces, casters, and multi-refining stages. The molten steel is firstly poured into a transfer ladle after being processed on the converter. We call the molten steel in the same converter a’charge’or job. Then the charge of molten steel is transported by a crane to a refining furnace for refining. At the refining stage, the molten steel undergoes multiple refining. Finally, the molten steel is transported by a crane to a caster where multiple charges are processed consecutively in the form of a cast. At the continuous casting stage, the molten steel is poured from the ladle into a tundish and solidifies in a crystallizer.The casts (including the sequence of charges within the cast), the process routes of all the charges, the processing times of all the charges on all the facilities and the transportation time between different facilities are known in practice. The performance indexes are that each cast should start on time, and the charges within a cast should be casted consecutively on the same caster, and the processing time of the charges cannot overlap on the same facility. The SCC scheduling is for each charge an allocation of one time intervals to one machine at each stage with performance indexes under given constraints, which is in the form of production timetable.Usually, the scheduling plan needs to be adjusted in practical production due to the short supply of the molten iron or scrap steel, the unqualified temperature of molten steel after converter processing, the unqualified temperature or composition of molten steel after refining and the delay of the start time of transfer facilities. The rescheduling of SCC is similar to the scheduling of SCC except that it has more constraints such as the limits of the casting time of each charge on a caster and the processed or processing charge.For multiple refining cases, the confliction of the processing time of two adjacent charges on the same facility can not be formulated by a mathematical model. The existing methods for the SCC rescheduling do not consider multiple refining, only a simulation study, and can not be applied effectively to multi-refining steelmaking scheduling in modern plant. Therefore, manual scheduling is taken to tackle the problem, which often results in the unacceptable waiting time between different facilities and breaking casting.This dissertation takes a large steel plant as the research background, whose production unit consists of three converters, six refining furnaces, three casters and tri-refining steps. The rescheduling methods for the SCC are studied under the environment of disturbance.The major contributions are listed as follows:1. For the small-delay problem caused by the delay of the start time of a charge, the performance index is that the charges within the cast should be casted consecutively on the same caster. The constraint equation is that the processing time of the charges cannot overlap on the same facility. The decision variables are that the allocation machines to the unperformed operation of each charge and the start time of each operation. Since the problem cannot be solved by existing optimization methods, this dissertation presented a rescheduling strategy as follows. Firstly, by fixing the processing time of casting and keep the allocation machines for each charge of the primal plan, the model becames a linear programming model with the objective to minimize the breaking cast time of the adjacent charges within a cast, and it can be determined the start time of unperformed charge on caster. Secondly, based on the precious result and the LP model, it can be determined the start time of unperformed charge on converter and refining furnace with the objective to minimize the waiting time. Finally, the start time of unperformed charge and performing charge on caster can be adjusted within a limited range by heuristic methods. 2. For the problem of a large delay caused by the delay of the start time of a charge, theperformance index is that the charges within the cast should be casted consecutively on the same caster. The constraint equation is that the start time of the charges cannot overlap on the same facility. The decision variables are that the allocation machines to the unperformed operation of each charge and the start time of each operation. Since the problem can not be solved by the existing optimization methods, this dissertation presents a rescheduling strategy including five steps. Step one is to distinguish the unperformed charges and performing charges. Step two is to adjust the start times by fixing the casting time and keeping the processing facilities of performing charges unchanged in the original scheduling plan. Part three is to determine the allocation machines (the converters and refining furnaces) for the unperformed charge by fixing the processing time of casting. Part four is to determine the start time of unperformed charge on converter, refining furnace and caster with the objectives to minimize the waiting time, breaking cast time, and the punctualities of tardiness and earliness of unperformed charge. It is realized through fixing the start time of the casting. Part five is to guarantee the continuous casting by adjusting the casting time of unperformed charge and performing charge on a caster within a limited range. Rules are employed in Step one, and heuristic methods are employed in Step two, Step three and Step five.3. For the problem of the breakdown of a converter or a refining furnace, the performance index is that the charges within the cast should be casted consecutively on the same caster. The constraint equation is that the start time of the charges cannot overlap on the same facility. The decision variables are that the allocation machines to the unperformed operation of each charge and the start time of each operation. Since the problem can not be solved by the existing optimization methods, this dissertation presents a rescheduling strategy including three steps. Step one determines the allocation machines (converters and refining furnaces) for the unperformed charges using the heuristic method by fixing the processing time of the casting. Step two determines the start time of unperformed charge on the converter, refining furnace and caster by solving a LP model with the objective to minimize the waiting time and breaking cast time. Step three adjusts the start time of unperformed charge and performing charge on caster within a limited range by a heuristic method.4. The rescheduling methods of this dissertation have been coded to a scheduling software. And we have applied to the aforementioned steel plant successfully. The rescheduling plan including 20 charges can be obtained by the presented methods within 10 seconds, while manual methods for the scheduling plan including 20 charges need 10 minutes. The average waiting time for all charges is 3 minutes, which is much shorter than that with manual methods (9 minutes). The practical running results show that waiting time of the average daily 66 charges dropped from the 814.76 minutes of the manual methods to 196.98 minutes of the presented methods, which is close to one fourth of the manual methods. |
