Process Performance Improvement-Part 3

Sept. 2, 2010
McMilland and Weiner Concluding Their Interview with Mike Brown. It Focused on the Bottom-Line Performance of Process Control Tools
By Greg McMillan and Stan Weiner

Greg McMillan and Stan Weiner bring their wits and more than 66 years of process control experience to bear on your questions, comments, and problems. Write to them at [email protected].

Stan: We're concluding our interview with Mike Brown, who's focused on the bottom-line performance of process control tools for basic and advanced control systems as vice president of North American Solutions at Matrikon Inc.

Greg: Do you see standard ways in which we can use metrics to better demonstrate the economics of good process control and the impact on process improvement?

Mike: Try to get the metrics associated with the key sensitivities of the plant management and the economic drivers of the plant. If a paper mill is suffering from a high level of sheet breaks, then report the benefits in "% sheet break reduction." This simple correlation analysis of paper breaks with control performance (loops in AUTO or variability) should allow these metrics to be generated. If the sensitivity is on production rates, then display the control performance benefits as incremental product revenue and/or flow rate (e.g. $/hr and ton/hr).

Process control engineers are often quite shy about generating these economic indicators and metrics. However, they must present them to management and be visible within the organization if they want their work supported.

Stan: What will be some of the biggest challenges on the metrics side?

Mike: One challenge is integrating the metrics into normal plant review meetings and management reports. Sustaining these technologies is always the challenge, so the metrics and visibility need to be integrated into normal plant practice and oversight. This needs to be a constant effort until they are just part of the normal daily plant reporting practice.

The other challenge is not letting people get bogged down in efforts to attain 100% accuracy in economic metrics. Some companies do not like having economic benefit metrics tied to process control applications because it can be difficult to get these calculations 100% accurate, given the complexities of plant operation. The reality is that metrics do not need to be 100% accurate to do their job. They need to bring visibility to ensure the organization focuses time and effort on the problem.

In the ethylene plant we discussed in Part 2 of this series, where poor analyzer performance was a significant cause of "loss of benefits" of the MPC controller, you can easily compute an order of magnitude economic calculation on this loss factor. Whether the number is $10,000 or $20,000 per day is not the issue. Getting the KPI visible, having it point in the right direction, and ensuring that it is big enough to justify the effort now required to fix the analyzers is how we will improve the economic return. Let the visibility of the economics drive the need to support our process control applications and the behavior we want in our plant resources.

Greg: Where can this bring us?

Mike: Better technology and tools to help drive process control activities towards performance improvement. The tools need to enable flexibility in resourcing in order to improve resource utilization. They need to be much simpler and drive simple workflows to attract good, young people into the practice. They must establish an economic reporting capacity that people can integrate into key plant operating metrics. People want to be able to deliver and see the results of the improvement efforts.

Stan: Operators can have a significant impact on plant economics. How do we get them to use performance metrics to drive their behavior? How can they use the control systems for process improvement?

Mike: Operators are a critical part of the equation, and often the one most ignored. I have spent 20 years watching operator behavior in control rooms. In most cases, operators want to be challenged and be better enabled to make the process move. They want to use the control system to drive plant performance, but doing that needs to start with accountability and equipping them with the information they need to drive performance.

The operators need to know that we are trusting them with the responsibility to drive unit performance and that they are expected to deliver. We need to start getting the key performance metrics, such as energy efficiency, capacity utilization and off-spec product, clearly published and trended on a shift basis in the control room. We need to let the operators see how they are doing and what is achievable. Don't be afraid to associate this information with operator shifts. Performance as a whole will rise.

Then we need to take our simple control system performance metrics, associate them with the plant economic metrics and trend them with the same level of operator visibility. Control room visibility of performance metrics, such as number of loops in automatic, percent loop saturation, plant oscillation index, analyzer online factor and MPC effective utilization rates, will ensure that the operators make the important connection between control system and plant performance. 

Some people still believe we should remove the operator from the process by automating everything in the control system. This isn't the path to better plant performance. We need to automate the plant's ability to move from one state to another as optimally as possible, and we need to provide the online models to compute the optimal states, given the economic conditions. However, we also need the operator to be very interactive with that process to understand if the control system is being best utilized to deliver on those objectives. We need to move to the concept of the operator as the pilot of the plane, and give him the same level of accountability for performance.

One company has started this process by simply changing the way operating orders are provided to the control room. Instead of simply giving the operator a set of targets and setpoints for key variables, it now provides only recommended setpoints with a clear explanation of the unit driving force for profitability. A second step in this effort is to ensure that the electronic logbook computes the unit's key performance indicators (KPI) with a link to trend them. Consolidating this information is a great justification for companies to move towards electronic operator logbooks as a way of driving performance.

Greg: What can we do to foster an understanding by process and mechanical engineers and operators of how dynamics set by process, equipment, final control elements (e.g. valves), measurement and strategy design and implementation affect control system performance?

Mike: I have always said that dynamics are what separate the thinking of the control engineer from the process engineer. The key there is to provide an understanding of how process and control system dynamics need to be understood and considered in the design of the plant and the setting of the daily operating states. One of the challenges for process engineers is to recognize the importance of final control element and measurement performance. The speed of response is critical in fast applications, such as compressor, furnace and polymer pressure control. Analyzer sample and cycle times are a concern for all loops. Analyzer and valve resolution and sensitivity limits create limit cycles.

Top 10 Control Room KPIs That Will NOT Lead to Process Improvement

10) Number of coffee cups consumed on night shift
9) Number of chairs broken by our favorite operator, "Big Bob" Gibson
8) Average time an operator can keep his eyes closed before falling asleep
7) Shortest recorded time for operator handover at shift change
6) Longest recorded time to bring up a DCS graphic following a control system "upgrade"
5) Number of times the operator gets the DCS alarm "MPC error—please re-invert the matrix"
4) Number of process engineers that cannot spell PID
3) Number of instrument tuning technicians named Ziegler or Nichols
2) Number of times an operator shouts "Bingo!"
1) Any KPI that needs to be charted with a control range of -1000% to 1000%.