Plant-wide automation at Elgiloy Specialty Metals. Background of project. Process utilization and efficiency. Production scheduling and monitoring, страница 8

The rolling mill level 2 automation layer is comprised of several major control tasks plus a process model. The interaction between these modules is depicted in Fig. 8.

As shown, movement of material in the mill is detected within the level 1 process controllers and communicated to the PCS system. Prom this point, the material is tracked and, if appropriate, a rolling schedule is computed. The rolling model uses information stored in the database along with adapted quantities to calculate the intended schedule of passes in the mill. The rolling schedule for a coil is comprised of the following setpoint variables: pass exit gauges, pass predicted forces, pass run speeds, entry tensions and exit tensions.

After calculating the rolling schedule, the set-point processing task dispatches this information to level 1 for distribution to the individual regulators. During subsequent rolling, measured values are captured and sent to the PCS for use in logs, pass-to-pass adaptive learning and transmission to the MSC system. A more detailed diagram of the data gathering and transmission to level 2.5 is shown in Fig. 9. It can be seen in this diagram that the data are comprised of a header plus cyclic data.

The header contains all the singular information on the coil taken principally from the PDI. Header information includes items such as coil ID, customer ID, start time, stop time, coil weight, work order ID, session number and alloy.

The header data are placed in a shared memory segment by the set-point processing task once the last pass has been commenced. At the end of the last pass, the final header data are stored.

Cyclic data are gathered during the rolling of the last pass into the shared memory segment. Cyclic data are gathered every two seconds during the last pass and are comprised of the following: thickness, mill speed, entry tension, exit tension, gap position and forces on hydraulic cylinders.

After the end of coil, this information is transferred to the MSC system via a SQLnet connection. Upon successful transmission, the shared memory segment is cleared in preparation for the next coil.

Finally, a neural network system was added to the mill to provide the fastest adaptation of numerous rolling variables to achieve the most repeatable process control. Adaptive learning is segmented into long and short-term components. Pass-to-pass adaptation is also performed during coil rolling. After each pass a new schedule is computed for all remaining passes.

Summary

A plant-wide data acquisition and product scheduling system can be achieved with minimal capital investment utilizing an architecture similar to the one described. Benefits derived from the system such as effective product scheduling, coil process history and customer order tracking have shown the solution to be very cost effective. Complex data concentration layers with the associated hardware and resultant software maintenance have been avoided.

Additional improvements and utilization of the plant data are contemplated for the future. Future enhancements being considered include:

·  Automatic coil PDI transfer from the enterprise system (PPS).

·  Plant product tracking.

·  Work order status communicated to PPS.