1 March 2002
Away from pedal-to-the-metal production
By Leoncio EstÚvez-Reyes
Triangulation solves the process control performance puzzle.
Pulp and paper manufacturers must improve economic performance to remain viable. Process control is one technology that can help achieve that goal, but not many enterprises manage that technology well.
Why? The main reason is they don't properly measure whether the technology is working. There are three types of measurements to make that determination: reliability, utilization, and variability.
FINDING USEFUL INFORMATION
Pulp and paper manufacturers face the challenge of improving economic performance to remain viable. Consolidation is not the sole answer, as size does not guarantee that earnings will meet the cost of capital.
Even mammoth companies need to operate within customer demand and away from the prevalent pedal-to-the-metal production rates if they want revenues to be sustainable in the long term. Companies need to find the operational point where production brings the most profits.
Process control is one technology that can help achieve that goal. Empirical data collected during the past 50 years strongly supports this notion. Yet few enterprises manage the technology well, and most fail to measure the technology's performance.
Computers and data acquisition systems have helped manufacturing companies automate most process operations. Automation systems are prevalent in all branches of manufacturing. These systems have inundated manufacturers with data about the process and process control.
Sorting through this mass of records and finding useful information is overwhelming, but one must properly determine whether the technology is getting the process to the desired sweet spot of profitability.
KNOWING WHAT'S WRONG
Companies traditionally measure financial performance as it affects corporate health and ultimately dictates survival. However, focusing solely on whether one is meeting shareholders' return, cost accounting targets, or productivity goals does not help businesses become consistently profitable.
This is because those measurements come after any opportunity to change the output is past. Another shortcoming is that management and engineering overlook critical factors that determine long-term business success.
Finally, knowing you are doing things wrong does not tell you what to fix to make things right.
Manufacturers need to measure performance in a way that drives employees to obtain the outputs that satisfy market expectations: constant quality of product, speed of delivery, reliability of supply, and overall service, all provided at the proper price.
They need to monitor performance indicators relevant to total customer satisfaction. This action needs to take place at a point in the productive progression where some output optimization can still occur. It must take place in a way that can easily diagnose and fix any existing problems.
In a nutshell, manufacturers need to measure process performance. This means all employees must be cognizant of the inputs and outputs of the processes they are involved with within the enterprise organizational context. They must track performance and respond to any deviations swiftly and with flexibility.
The above ideas spin around the notion of the process, but what does that mean? A process is a set of tasks and activities that takes place within a manufacturing facility, uses a certain infrastructure (facilities and equipment), and follows a certain strategy (methods and procedures) to manipulate inputs in order to obtain desired outputs.
A process always performs. Therefore, its performance is quantifiable, measurable, and controllable.
The definition of process has an associated concept: the process customer. The process customer is not an anonymous being that buys the finished products. The customer embodies all of the entities that receive the outcome of the process.
Thus, the customer can be the manufacturing personnel in charge of a second process downstream of the originating one, the people affected by the process outputs (environmental, legal, and the like), or the distributing firms that market the finished product. The customers are important because they are the ones that determine whether the outputs are within an acceptable range and meet their expectations.
This measurement approach is analogous to the trigonometric principle of triangulation. If one knows at least one side and two angles of a triangle, one can compute the other two sides and the third angle.
Choose metrics that are a natural extension of the triangle model and are analogous to its sides. The metrics group in three categories: reliability, utilization, and variability.
Variability performance metrics
METRICS GIVE COMPLETE PICTURE
Reliability metrics measure the process incidents that cause downtime or periods of reduced production rates and their main statistics: mean time to fail, mean time of incidents, production losses, root cause, and the like. They are associated with the customer expectations of predictable and speedy delivery of products.
Variability indices measure the performance of process variables such as temperature, consistency, pressures, and others and benchmark them against an applicable standard such as minimum variance control (MVC). They link to the customer expectation of constant and optimal product quality.
Utilization indicators measure who or what is controlling the process, as indicated by the need for human intervention, presence of abnormal conditions, etc. They tie to the customer's expectations for cost and service.
All of these metrics give a complete picture of the influence the elements in the process have on results, such as the following:
- The equipment and facilities (i.e., distributed control systems (DCSs), programmable logic controllers, sensors, and control valves)
- The production strategies and methods (i.e., setting rates, staffing, standard operating procedures, standard maintenance practices)
- The nature and behavior of the inputs and outputs
If the objective is to gain a good perspective of the influence the strategy and people have on the infrastructure, examine the reliability metrics. Use them to launch improvements in control assets management: review of maintenance practices, replacement of obsolete equipment, selection of instruments to withstand hazardous environments, and others.
If the objective is to account for the influence that the people and the production strategies have on the inputs and outputs, observe variability. These figures can help guide improvements in the algorithms, strategies, and elements used for process control. These may include changes in controller tuning, relocation of instruments, and use of advanced controls.
If the objective is to gain insight into the relationship between the people and the influence of the equipment on the inputs and outputs, look at utilization metrics. These can lead to improvements in the operating procedures and methods the operators use to interact with the process through the use of controls. Examples are implementation of best operating practices, alarm management, and the like.
The combination of these three metrics provides a full perspective on the health of the process, which in turn facilitates making adjustments in overall process control.
ALARMS SHOW OPPORTUNITY
A management team at Weyerhaeuser's Grande Prairie, Alberta, pulp mill chartered a team whose responsibility was to evaluate the performance of the process control system on site. Since then the team has put together several computer applications and work initiatives using the concepts outlined above to identify and measure the different metrics mentioned: reliability, variability, and utilization.
For reliability, the team tracked downtime and reduced production rate incidents caused by the controls and instruments for the different mill units, areas, and shift teams. Several trends, including moving averages, give a picture of uptime and throughput.
If the reliability numbers show an opportunity, the area personnel can look to see if that opportunity is in process equipment outages or changes (obsolescence?) or in methods (improper care or operation?).
For variability, the metrics used to monitor every control loop in the DCS are as follows:
- Coefficient of variation (COV), which is the signal standard deviation divided by its average
- Harris Index
- Process capabilities (Sprod, Scap,and Smvc)
- Historical trends of the Harris Index and the COV Index
- Trend of the controller output
- Controller output vs. process variable
- Distribution of the error (process variable minus set point)
Monitoring and subsequently analyzing these figures provide a measure of how loops compare with a theoretical benchmark: the MVC. The analysis points out which loop represents the best opportunity. If variability numbers show an opportunity, it is possible to then look and see whether that opportunity is in process equipment changes (poor tuning, valve wear?) or strategy (inadequate control scheme?).
For utilization, choose metrics that indicate the number of operator manipulations such as tuning or other parameter changes. Also, measure alarms per shift team. The numbers are also expressed on a per-operating-console basis.
Quantification of alarms and interventions will indicate the best opportunities. These measures focus maintenance resources on problematic loops based on the number of alarms they generate and the number of manipulations the operator requires. The utilization indices show whether the DCS or the operators are controlling the process.
Documenting changes and e-mailing adequate reports to management and process and area owners supports change management for all these metrics.
BETTER SAFETY NUMBERS
This pulp mill is in better process health as a result of these metrics methodologies. From 1997 to the present day, the mill has reported production and quality improvements that translate into an overall return on net assets (RONA) well above the known numbers for the pulp and paper industry's best players.
Although it is hard to credit exactly which programs are directly responsible for every percent of increase in RONA, Weyerhaeuser is convinced enough that it intends to deploy analogous tools for process measurements on a companywide basis. The goal is to have the tools deployed in all the manufacturing sites within the pulp, paper, and packaging business by the year 2005.
Reductions in variability and human interventions and increases in reliability correlate with more production, better safety numbers, lower expenses, and streamlined personnel needed to oversee and care for the process. IT
Behind the byline
Leoncio EstÚvez-Reyes is an engineering specialist with Weyerhaeuser Research and Development in Federal Way, Wash. He is a professional engineer with 14 years of experience as an instrumentation and controls technology practitioner in petrochemicals and pulp and paper. He has a Ph.D. from the University of British Columbia and a master's of engineering degree from McGill University.
|COV||coefficient of variation|
|DCS||distributed control system|
|Harris Index||ratio between the actual variability of a process variable and the variability that is achievable with ideal variance control|
|MVC||minimum variance control|
|RONA||return on net assets|
|Scap||capability standard deviation represents the minimum variability that is physically possible in an industrial process|
|Smvc||variability that is possible to achieve in an industrial process when applying minimum variance automatic control|
|Sprod||standard deviation present in a process variable|