Furnace Integrity Monitoring

 

SUMMARY


Case Study:
Metals


Company:
Glencore Xstrata


Challenge:
Improve reliability and robustness of furnace sensors


Results:
Increased furnace life, automatic detection of sensor faults, earlier warning of catastrophic failure, reduced opex

BUSINESS CHALLENGE

Continuous operation of modern furnaces poses serious operating challenges. It is of critical importance that the physical integrity of the furnace is known and understood at all times. Furnace refractory linings have failed catastrophically in the past, resulting in unplanned shutdowns and extreme risk to health.
These unplanned shutdowns represent a considerable expense in an industry where continuous operation is of paramount importance to maintaining cost effectiveness.

An additional challenge is in managing and prioritizing the process information displayed to the process operator. It is impossible to manually keep track of the many hundreds of thermocouple signals that may be presented, and traditional univariate alarming schemes can create nuisance false alarms without delivering the sensitivity required to spot faults before they cause irreversible damage.

PERCEPTIVE SOLUTION

Existing signal-by-signal alarm methodologies are ‘reactive’; the variation in the signal's value must be significantly greater than 'normal' signal variation before an alarm is triggered in that section of the furnace.  However, by the time the alarm triggers, the damage is often too large to avoid an unplanned shutdown.  But setting a smaller alarm range can cause spurious and nuisance alarms that lead to unnecessary shutdown.

The premise for the system developed using MonitorMV is that a model-based condition monitor need not focus solely on individual signal values, but also on the correlations and relationships between variables. Typically, thermocouple measurements around a furnace are highly correlated. The temperature measurements are dependent upon factors such as the reactions occurring within the furnace, the heat transfer through the furnace walls, and the spatial arrangement of measurement points. The correlated nature of the sensor data allows a MonitorMV MSPM model to reconstruct highly accurate predictions of signal behaviour. In the event of a potentially dangerous change in the integrity of the furnace, the prediction error becomes very large for signals in the region of the fault. In this way, the changes to the integrity of the furnace are detected as soon as they occur, and can be tracked as they develop. By contrast, by the time a univariate sensor alarm is triggered, usually the only course of action is a reactive emergency shutdown, with a corresponding loss of production.


RETURN ON INVESTMENT

Early warning of failure permits a safe, controlled shutdown of the furnace, with minimal loss of production and improved maintenance planning.  Perceptive's furnace integrity monitor is a fully integrated system installed at a number of industrial sites around the world. It is utilised on a day-to-day basis within smelter control rooms. After a number of years of operation, the following benefits have been delivered:

  • Priority warning of imminent failure enable safer shutdown and better planned maintenance
  • Improved operator information permits extended furnace campaign life
  • Automatic detection of faulty sensors reduces costs of reactive maintenance
  • Reduced insurnace premiums, due to higher furnace safety and reduced production losses

 

The solution has now been patented by Glencore Xtrata.

 

      
Perceptive Engineering Clients & Partners