APPLIED INDUSTRIAL CONTROL SOLUTIONS LLC

HOME PAGE Case Studies Published Articles Control Tips/Tricks Training/Mediation

FIELD SERVICE REPORT No. 32

To:

(Proprietary)

From:

Brian Boulter ApICS LLC
Robert Lockhart ApICS LLC

Date:

November 24th 1999

Subject:

Field Trip Report, ALUMAX Inc.

Introduction

At the request of Tom Kennedy of Mannesman-Demag and Chris Depiero of Rockwell Automation, ApICS LLC was commissioned to make a field trip to the Alumax Inc. No. 3, 5 Stand Aluminum Hot Rolling Mill at Lancaster PA to address instability problems in the interstand tension control loops. The author and Chris Depiero made a field trip to the site to identify and correct the stated problem. Two trips were made to the site; 11/11/99 – 11/13/99 and 11/18/99 – 11/20/99. The following observations and corrective actions were taken.

Observations, Problem Identification:

The mill interstand tension exhibited random parasitic instabilities with harder wide materials, and consistent instabilities with narrower hard materials. The instability that was observed was of the "classic" resonant energy kind. This parasitic instability results when the rotational speed and/or the torsional natural frequency of one or more of the rotating elements in the system co-incides with an interstand span natural frequency. When this occurs, energy at the resonant frequency is injected into the system by the drives, and feeds upon itself through the exchange of energy between the two driven inertias and the strip (or spring) connecting them. This problem can be overcome using tension regulator cross-coupling techniques that provide damping of the torsional and/or system natural frequency. These schemes must be frequency scheduled as a function of strip cross sectional, or by implication span natural freqeuncy.

Software revisions and architecture reconfiguration recommendations: Random trip outs were observed due to thermal overload faults and random SCR failures. The current loop rate limitation was set to 25 [pu] on stand #2, and 35 [pu] on stands #3 and #4, GE specifies 15 – 20 [pu] for their mill grade solid frame motors. The higher than rated current rate resulted in the motor on stand #2 flashing early in the morning on 2/13/98.

The gauge control vendor requested that we provide a "knockdown" feature in the mill software. The purpose of the "knockdown" feature is to reduce the mill speed to a rolling speed that coincides with the base speed of the #2 stand motor whenever a current overload, a thermal overload, or a maximum speed condition occurs.

Corrective actions:

The "hot" tuning for the speed and tension loops on stand #2 and #3 were contributing to the excitation of parasitic resonances in the machine. The source of the resonances were identified, as stated above, and were verified to coincide with mass-spring resonance frequencies that were calculated using an ApICS LLC proprietary tension system analysis program. The nuisance trips that occurred in the machine were addressed by adjusting appropriate thresholds. Following is a summary of the systems engineering corrective actions:

  1. Adjusted the speed and tension loop tuning on stands #2 and #3.
  2. OLD

    Bandwidth

    NEW

    Bandwidth

    #2 Kps = 18

    (15 – 18 [rad/sec])

    7

    (6 – 7 [rad/sec])

    Wlds = 3

     

    0.1

     

    #3 Kps = 20

    (15 – 18 [rad/sec])

    15

    (12 – 14 [rad/sec])

    Wlds = 3

     

    0.1

     

  3. Lowered the anti-aliasing frequency on the Reliance A/D card. There was a significant amount of machine and electrical noise on the signals from the load cells. To alleviate the potential for aliasing the second order butterworth low-pass anti-alias corner frequency was lowered on both input channels.
  4. OLD

    NEW

    300 [rad/sec]

    75 [rad/sec]

  5. Lowered the CML rate on stands #2, #3, and #4. The current rates for all three motors were above the rated values specified by GE.
  6. STAND

    OLD

    NEW

    #2

    25 [pu]

    15 [pu]

    #3

    35 [pu]

    20 [pu]

    #4

    35 [pu]

    20 [pu]

  7. Opened up the % control on the tension major loops. The temperature compensation scheme that is employed by the gauge control vendor in the automatic mode may result in vernier trims that exceed 10% of speed set-points to maintain tension.
  8. STAND

    OLD

    NEW

    #2

    10%

    20%

    #3

    10%

    20%

  9. Opened up the thresholds on the SCR fault diagnostics. The nuisance SCR faults trips at low loads were due to conservative SCR diagnostic settings. See the AMX DCS manuals for details.
  10. STAND

     

    OLD

    NEW

    #2

    Deadband

    10

    20

     

    Trip-point

    1500

    3000

    #3

    Deadband

    10

    20

     

    Trip-point

    1500

    3000

    #4

    Deadband

    10

    20

     

    Trip-point

    1500

    3000

  11. Thermal overload recalibrated. The motor thermal overload time constant was adjusted to reflect the value recommended by GE.
  12. Variables for trending the following were created and compiled into the Sigma database.
  13. Overspeeds on #2. #3, and #4

    Overvoltages on #2. #3, and #4

    These fault conditions were also logically linked into the drive fault ladders.

     

  14. The signal that is fed back from the vernier on the tension regulator from stand #3 to the speed summer on stand #2 was scaled as a function of the interstand speed ratio. A gain block was also included and set to unity (basically not used).
  15. A CEMF "spillover" regulator was designed and coded into the controller. The field regulator scheme that was included in the shipped software was a feed-forward regulator that set field strength with speed reference. The actual armature voltage could drift depending on the IA drop for the given operating currents and temperatures. To guarantee that the motors would not run in an overvoltage condition A CEMF spillover regulator was installed and set to spill at 680 [VDC]. It was tuned to ensure that there were no adverse interactions with the field and speed loop tunings.
  16. A cross coupled frequency and gain scheduled tension regulator was implemented (Figure 1). The existing tension regulator gain scheduling (as a function of cross-sectional area) was rescheduled as follows:

OLD

NEW

The equations for scheduling the tension regulator feedforward lead/lag are:

Figure 1.

Conclusions

After making the above software revisions and architecture reconfiguration recommendations, adjustments, design changes and calibrations, the machine was observed to operate with considerable robustnes, and no nuisance trips were observed. Tom Kennedy was given permission by Alumax Management to put Brian Boulter on call should any further problems occur.

HOME PAGE Case Studies Published Articles Control Tips/Tricks Training/Mediation