In times of major plant upsets, to prevent the operator from being swamped with a multitude of alarms, there is a need for a means of only activating the alarms that are the highest priority in the given situation.
To minimise alarm floods on major process upsets, one has to filter or suppress alarms in accordance with plant operating mode and alarm priority. This is achieved via a suppression logic block located between the initiating alarm signal and the alarm processing circuit, refer to Figure 1. More than one form of suppression can be included within the same alarm input as indicated with the priority four alarm which in this case is a level measurement. The suppression and priority switch logic blocks are shown relative to the HMI graphics priority colour driver, which is described in more detail later. However, one should always use alarm suppression with caution, being aware of the consequences under all operating modes and requires 'buy-in' by all relevant parties, this is discussed in more detail later. It should be mentioned however, that most current central control systems (in the author's opinion) have far too many alarms. This is because they are virtually free, so we tend to give little thought to their real value to the operator and include them on any measurement.
These include several types of alarm:
* Absolute.
* Deviation.
* Rate of change.
* Discrepancy.
* Permissive.
* Status.
* Calculated, etc.
We need to get back to the systematic application approach of the old conventional window type annunciator, where each and every alarm cost the project real money to implement and had to be fully justified. Some types of alarm may require self re-initiation after a certain time; these are typically status alarms such as trip overrides to remind the operator of an unsafe operating mode. As a general guide, it is recommended that the 'total number' of alarms should approximately equal the number of analog inputs. Care needs to be taken in the assignment of alarm priority, there should be no more than three or four levels:
1. Critical (5%) - Personnel at risk, fire/gas, costly equipment damage, serious environmental impact, typically these are final warnings eg trip failed to operate.
2. High (10%) - Possible equipment damage, possible environmental pollution, lost production, typically these are trip alarms.
3. Medium (15%) - Large plant upset, off-spec product, equipment fault, typically these are pre-trip alarms.
4. Low (70%) - Minor plant upset, inefficient operation, potential problem, typically these are informative process alarms which if left could lead to more serious problems.
In addition to the above, it is normal to log certain 'events' such as pump start or auto/man change on controllers, but these are not classified as alarms as no 'audible alarm' is given nor is any 'accept' required. In order to have effective priority discrimination, to assist in the responsiveness of operators to alarms, it is recommended to use the priority percentage guidelines as shown in the brackets above (these vary slightly to those recommended in EEMUA 191). It is interesting to note that at the Texaco Refinery incident of 1994, 87% of the alarms were classified as high priority, with such a high percentage, how could the operators determine which were more important?
On some alarm applications, it is good practice to switch the alarm priority depending on unit operating mode, for example, some measurements such as temperature are more sensitive to rate of change during start-up than when at normal operating load. Alarm priority changes can also be initiated by maintenance functions such as the testing of trip systems. When a trip system is under on-line maintenance or periodic functional testing, the associated trip pre-alarm which would normally be a priority 3, should be switched to priority 2, as the actual trip protection function is temporarily inoperative thereby elevating the importance of the associated pre-trip alarm. The use of different audible tones for alarm priority has been used with limited success; it is however recommended that no more than two tones be used. For critical alarms, short electronic voice messages have been used in recent years. On many alarm systems, it is not possible to detect which alarm occurred first to initiate an equipment trip such as with a compressor, this is due to a relative slow scan rate within the DCS. Where this facility is required for troubleshooting, a specific sequence of event recorder (SER) software package with fast scan times (less than 1 ms), must be implemented for true sequential alarm time stamping.
During plant disturbances, a useful operator aid is to indicate on a process overview graphic which major sections of plant or equipment are in the highest alarm priority state. This has been termed a process equipment relational interface (PERI) display, which is driven by an HMI graphic priority colour driver block, as shown in Figure 1. This is very helpful by quickly directing operators to the process area for most urgent attention during upsets, by using a four or five colour change of respective process equipment. The operator will 'click' on that equipment to automatically display a more detailed graphic indicating the problem area with the specific alarm measurements. Although one should be careful when using colours for such interaction, it is suggested that for healthy operation green is used, then progress from low priority yellow through amber and red to blue, for critical alarm or operating state. The background colour should be nondistracting, and at the same time provide a clear contrast with the graphics. Light grey seems to be a good choice. It has been suggested that 'white' be used for healthy operation which has a tendency to blend in with a light grey background, ie nonconspicuous which is probably ideal for a healthy display operation, as operators attention should only be directed to problem areas. Although the PERI display as discussed here has only been applied to assist in alarm management, it can also be used for equipment and process efficiency indication which would require a separate plant overview display. When the operator selects an item of equipment showing poor efficiency, a more detailed analysis of the equipment operation and specific problem area will be advised with suggested corrective action. This could be a tool for linking to an operator's productivity bonus!
To be continued in the next issue: Part IV
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version