Do not use the Bad Actor Method for Operating and Alarm Limits
You’ve rationalized your Alarm and Operating Limits with CVE and are getting the best alarm performance you have ever had. The CVE method is a multivariable method and its success comes from considering all the limits as the Set of the vertices of the enclosing hypercube of the Operating Envelope that delivered your Operating Objectives. This is what is meant by ‘Consistent Alarms’ or ‘Consistent Operating Limits’. You will destroy the Consistency very quickly if you go back to univariate methods, such as the Bad Actor method, to adjust individual limits and you will soon undo all the good effects of your rationalization.
Why would you want to adjust individual limits and how should you do it? The Capability of many processes changes over time as heat exchangers foul; fired heaters create scale on tubes; compressor blades (esp. gas turbines) and fin-fan coolers become dirty; catalyst activity decreases and many other slow degradation effects occur. Think, for example, of a heat-recovery heat exchanger cooling hot product and heating the incoming feed stream. As the exchanger fouls the heat transfer coefficient (which you probably monitor today as part of process performance monitoring and maybe use when deciding when it is time to clean the exchanger) so the product exit stream is hotter than it used to be and the feed exit stream is cooler. If the next unit the feed stream encounters is a fired heater then more fuel has to be burnt in the heater to raise the feed stream to the desired temperature. The product output stream is hotter than it used to be and may be nudging acceptable limits for storage tanks. So at least three variables (fuel gas flow, tank inlet temperature and exchanger outlet temperature are operating closer to limits and may have rising numbers of alarms. How should you correct the situation caused by the slowly fouling heat exchanger which you consider insufficiently fouled to justify the process downtime and/or maintenance cost of cleaning it now?
The correct method, assuming you are reviewing alarm performance monthly, is to collect the last month of process history from your historian and append it to the data you used during the rationalization using the same CVE ‘Saved State’ so that you can see the Operating Envelope that was correct at Rationalization time. Now move the limit on one of the affected variables to a value that accommodates the current capability and lock it and other significant variables such as the Operating Objectives and CVE will show you the Consistent Operating envelope that results. Run Alarm Performance Prediction and observe that during the last month you would not have had the alarms that promoted the Reviews’ concern.
Best of all would be to include the heat transfer coefficients from your periodic heat balances and use them directly to indicate the consequences of fouling on other variables. You might even be able to perform and store the heat transfer coefficient calculations in CVE and perhaps simplify the job of the process engineer who calculates them today and the making of the decision of when to clean the fouling exchanger.
But don’t use univariate methods such as the Bad Actor method for alarm performance monitoring and make sure that those involved in Alarm Monitoring have had training in CVE as they may not have been involved in the initial rationalization. And write a Business Procedure for Alarm Monitoring because it probably wasn’t covered in depth in the Alarm Philosophy document.
First presented on: 14 December 2022
Presenter: Alan W. Mahoney, PhD