Bridging the gap
Careful planning will connect your Islands of Automation
By Darren L. Goodlin and Robert Rice
When developing the control strategy for a process, it is important to consider more than just the individual characteristics. An effective control strategy needs to also factor the limitations of the process’ instrumentation and other associated process equipment.
Sensitive instrumentation, in general, is any inline instrument or analyzer affected physically or through error in its ability to correctly measure. Oftentimes the operation of the process does not include how sudden pressure or temperature changes will impact inline analyzers. Here is some food for thought about how these instruments and analyzers feel the impact when a team comes up with a process control strategy that does not include them.
Membrane analyzers: These analyzers have a process interface that allows them to measure the items under analysis. The basic concept of membrane analyzers is a gas dissolved into solution like carbon dioxide permeates through the membrane in a dissolved state and then undergoes measurement on the other side of the membrane. One of the most common methods of measuring carbon dioxide is through a temperature/pressure method known as the Zahm principle. When the process line pressure suddenly changes or it creates pressure spikes, this type of analyzer feels the affect and may not perform. If the pressure surge is large enough, it could even permanently damage the analyzer’s process membrane. Pressure surges can also be in the form of negative swings to the extent of a vacuum within the process line. This will “suck or pull” the membrane from its supporting structure and stretch to the point of damage or tear. When these sudden pressure swings contain a dissolved gas like carbon dioxide, it could cause the dissolved gas to release from a solution just like opening a bottle of soda. This is stringing carbon dioxide out of solution. Once the gas leaves the solution, it has an impact on the ability of the analyzer.
Analyzers with a glass process interface like a pH sensor can feel the impact as well with hydraulics but in a slightly different way. pH sensors utilizing a glass process interface also contain a section known as the reference junction. This junction provides a reference to the sensor about the properties of the medium. This type of flowing junction has an internal sensor electrolyte that very slowly flows out into the process as needed by the sensor. When high pressure surges occur, the process fluid drives the electrolyte in reverse, and the process fluid contaminates the reference junction and causes the sensor to not be able to compensate for the chemical properties of the measured fluid. This can cause measurement drift or shift and can be permanent or cause a delay in the ability of the sensor to correctly measure the pH of the process fluid or product.
Magnetic and Coriolis flowmeters: These instruments feel the impact when a solution with dissolved gas in it comes out because of poor hydraulic process control. Both these instruments will not correctly measure the flow and density of liquids when the pipe is not full or has gas in the measuring element.
People often do not take hydraulic impact into account when they design process systems or when they develop and then implement control strategies. Flow control strategy is a very important part of any process system. Abrupt flow changes can cause hydraulic hammer to the extent where it damages equipment, inline instrumentation, and analyzers. You can see signs of hydraulic impact through missing or damaged pipe insulation and signs of piping systems moving in the hangers or supports.
It is highly recommended that anytime there is an installation of an inline process analyzer, you also bring in a pressure transmitter. This will provide a means to diagnose analyzer failure or measurement error caused by pressure surges. This measurement will need to log into the control system if available or have built in maximum and minimum pressure logging capability. This will also assist in holding analyzer vendors accountable of their analyzers integrity if this line hydraulic information is available.
When connecting two modular systems together, there are considerations that everyone must understand. Incompatible flow patterns between one piece of equipment and another can wreak havoc on the piping and instrumentation that connect these two islands. Two of the most common process conditions that result in instrumentation failure are pressure spikes and hydraulic hammering. These conditions frequently occur from large fluctuations in flow rate.
Oftentimes, hydraulic hammering presents itself as a loud banging sound while pressure spikes may be silent. Both can cause significant damage to a plant’s piping and instrumentation.
Pressure spikes occur for a number of reasons, but they most commonly occur when valves open and/or close too quickly. A pressure gauge with its needle permanently stuck at one extreme is a clear sign of spiking. In most cases, pressure spikes last for less than a second. Even though of short duration, a spike can cause damage that will take days to recover. In many cases, the pressure spike may not even be evident to staff when they review pressure data collected by a plant’s process data historian. This is most likely due to process data aliasing. If the historian collects data at the rate of once every 30 seconds, then there is a 1 in 30 chance of actually witnessing the 1 second pressure spike. This can make detecting pressure spikes from historically trended data extremely difficult.
Hydraulic hammering occurs from a pressure surge imparted by the kinetic energy of a fluid in motion when it stops or changes its flow rate suddenly. Consider the initial charging of an empty pipe. A 6” ID pipe containing water that flows at a rate of 2,000 gallons per minute (GPM) down a 100 foot length of straight pipe to a 90° elbow is the equivalent of a small car driving toward a wall at 15 miles per hour. In such a case, the phenomenon is audible, and it can rip pipe hangers right off the wall and strip pipes of their insulation.
Other sources of pressure spikes and hydraulic hammering include Clean In Place pumps, discrete control of tank filling or line charging, as well as poorly tuned flow and jacket temperature controllers. A stream supply or return system is also susceptible to hydraulic hammering. In this type of system, the steam has a tendency to condense into water in horizontal sections of pipe. The water then flows at high velocity through the system and stresses pipes upon impact at each pipe elbow. You can correct this condition by improving condensation draining practices.
Surge tanks counteract fluctuations in flow characteristics that would otherwise damage upstream or downstream systems. Surge tanks are common across the process industry and work in a variety of applications to improve control. Oftentimes, surge tanks go between two process systems that have incompatible flow patterns. The uncontrolled flow associated with the surge tank is the “wild stream” and can either be the flow in or the flow out of the surge tank. The primary control objective of the surge tank is to absorb the fluctuations of the wild stream without significantly impacting the controlled stream. To achieve this result, the level in a surge tank should swing between an upper and a lower level limit. The more the tank can swing, the greater the surge capacity of the tank.
Here is a surge tank between two processes that have incompatible flow patterns. You can control the level in the beer surge tank by adjusting the flow of draft beer as it pumps from large storage tanks to the beer surge tank. The “wild stream” is the flow of beer out of the surge tank to the SVK Kegging System. The flow rate within this system can vary from 0 to 180 GPM. As stated earlier, much of the instrumentation shown can feel the affect of the large flow changes required by the kegging system.
Keep it S.I.M.P.L.E.
Good control is in the eye of the designer. One engineer’s concept of good control can be the epitome of poor control to another. In some facilities, the ability to maintain operation of any loop in automatic mode for a period of 20 minutes or more is good control. One assumption of good control is the control loop’s ability to achieve and maintain the desired control objective. This leads to an important question: What is the control objective?
Understanding the control objective suggests the engineering team has a firm grasp of what the process should accomplish. This must be the case whether the goal is to fill bottles to a precise level, maintain the design temperature of a highly exothermic reaction without blowing up, or some other objective. Truly the control objective involves this and more.
When developing a control objective, it is important to take a macro view of the process.
There are over 70 different and documented techniques for quantifying controller performance. Here are a few different ways to benchmark performance and to analyze interacting processes. The key to useful performance benchmarking is choosing a performance measure compatible with the desired control objective.
As mentioned earlier, surge tanks buffer the fluctuation in liquid flow between upstream and downstream processes. They absorb the force associated with large fluctuations in flow rates that would otherwise have a negative impact on instrumentation and/or product quality. As such, the desired controller performance does not relate to set point tracking or disturbance rejection. Rather, it all boils down to how well the surge tank suppresses excessive flow dynamics in the “wild stream.” The performance metric used to describe the successful operation of a surge tank should therefore describe how well the surge tank can contain the wild stream.
One visual technique useful in evaluating surge tank performance involves plotting the “wild-stream” and the controlled stream on the same trend. Ideally, the controlled stream will not perfectly match fluctuations in the wild-stream.
Another useful performance benchmark associated with surge tanks correlates the movement of tank level with the controller output. Since the goal of surge tanks is to minimize controller output travel by maximizing level swing, an equation can help evaluate the relative performance of any given surge tank. The surge impact index is a value you can calculate according this equation. One note is if the surge impact index is smaller, the better the surge tank is able to suppress surges.
Interacting processes can be troublesome in any manufacturing process. By identifying which systems interact, you can counteract the disturbances rather than perpetuate them throughout the system. Even if you can not eliminate an upstream disturbance by identifying the source, a feed-forward controller could improve downstream loop performance.
Cross-correlation analyzes the relationship between two data series. By calculating a set of correlation values at increasing time delays, a picture develops that shows how the data series relate through time.
Cross-correlation values are always between negative one and one. Positive values indicate that process A directly affects process B, so an increased deviation from average in process A causes an increased deviation in B. Negative values indicate an inverse relationship such that an increased deviation in process A causes a decreased deviation in process B. If there is no relationship between the data sets, then the cross-correlation values will be close to zero.
Cross-correlation can also determine exactly how much time elapses before reaching the downstream process. At the point when there is greatest impact on the downstream loop, there will be a peak in the cross-correlation trend.
Additionally, cross-correlation can identify when disturbances occur from a recycle stream. If a recycle stream occurs within a single control loop, an autocorrelation can identify how the recycle influences the system.
Power spectrum can also identify and analyze interacting loops. The same events also affect interacting loops and therefore have power spectrum peaks at the same frequencies. Power spectrum cannot identify how long it takes for a change in one system to reach another like cross-correlation can, but it can be more useful when there are many processes separating the suspected interacting loops. Cross-correlation can blur when there are many processes in between with varying relationships.
Modular system suppliers design their specific systems to perform a set task under controlled conditions. Oftentimes, the same system can work in different applications and under varying operating conditions. Suppliers designed these systems to be generic and flexible enough to handle these different conditions. When discussing the specifics of an operating environment, it is important to ask questions about the supplier’s system without dictating how it should perform. It is easier to learn their system and requirements first, before introducing them to your operating environment and control objectives.
When determining the control objective or devising the associated control strategies for your production processes, careful consideration of their impact on the surrounding equipment should occur. Process events such as hydraulic hammering and pressure spikes are common sources of instrumentation failure. You can avoid these same events by applying the correct control strategy.
Performance measures are an integral part of optimizing and maintaining system performance. In some cases, performance assessment methods may only identify the start of a problem as opposed to its source. By understanding these basic principles and the disturbances that impact a system, engineers can determine what to expect during normal operation.
ABOUT THE AUTHORS
Robert Rice, Ph.D. is director of solutions engineering at Control Station, Inc. Darren L. Goodlin is manager instrumentation technologies brewing engineering and technology at Anheuser-Busch Inc.