Learning control apparatus for a reversing rolling mill

A learning control apparatus for a reversing rolling mill that, when controlling said reversing rolling mill in performing a rolling of N passes, taking N as an integer of 2 or more, computes required setting values for a mill operation of each pass by performing an optimum pass schedule calculation using a mathematical model and, at the same time applying learning control, comprising:

• N in number actual data gathering means that, taking i as any pass out of 1˜N passes, collect actual values obtained in pass i rolling after a current material has been respectively rolled in the i passes, find actual-calculated values by executing model calculations in every pass i using said actual values, and store said actual values and actual-calculated values;
• (N−1) in number pass to pass learning calculation means that include:
• a pass 1 pass to pass learning calculation means that, from among said actual data gathering means, computes a learning term that assimilates a pass 1 pass to pass model error based on said actual value and actual-calculated value stored in said pass 1 actual data gathering means, and stores said learning term, and
• pass i pass to pass learning calculation means that, taking any pass from among 2˜(N−1) passes as pass i, compute learning terms that assimilate model errors of sequential pass to passes i based on said actual values and actual-calculated values that are stored in the respective pass i actual data gathering means together with said learning term for pass 1 that is stored in said pass 1 pass to pass learning calculation means, and store said learning terms;
• (N−1) in number next pass set up calculation means that sequentially compute required setting values for an operation of said mill for 2˜N passes based on said learning terms respectively stored in said (N−1) in number pass to pass learning calculation means;
• N in number direct pass to pass learning calculation means that compute learning terms that assimilate errors for a next material from pass 1 to pass N based on said actual values and actual-calculated values for pass 1 to pass N of a current material that are respectively stored in said N in number actual data gathering means;
• a lot to lot learning calculation means that, based on model actual values and actual-calculated values for each pass i from pass 1 to pass N of current material that are respectively stored in said N in number actual data gathering means, educes model error parts that depend on said material and product dimensions, finds a learning term that assimilates lot to lot model errors, and stores said learning term in a table;
• and
• N in number next material pass set up means that, taking N as an integer of 2 or more, and based on said learning terms that are stored in said direct pass to pass learning calculation means and said learning term stored in said lot to lot learning calculation means, sequentially computes required setting values for said operation of said mill for each pass from pass 1 to a final pass for said next material,
• wherein said learning control apparatus executes a rolling of pass 2 onwards in conformity with said setting values of said next pass set up calculation means and executes said rolling of pass 1 to pass N for said next material in conformity with sad setting values of said next strip pass set up means after completion of the rolling of pass 1 of said current material.