All Search Results matching “lou heavner”

Neural networks, not the ones in our brains, but the ones that came out of the artificial intelligence research, began back in the 1940s. This advanced process control (APC) technology has been used in the process industries for many years for applications ranging from virtual sensors to end of batch prediction. Neural networks have been discussed in numerous ModelingAndControl.com blog posts. Several software packages, such as DeltaV Neural, exist to build these neural networks applied in process automation-related control strategies.

I received an email yesterday from an engineer trying to predict quality for a product made in his manufacturing process. Their current way to do this was to get quality data from lab samples drawn every 4 hours. Obviously if quality problems are not discovered until the results are back from the lab analysis, it can impact 4+ hours of production. The person asked me what historical data collection frequency should be used to build the neural network model. In his case, he was using DeltaV Neural.

I turned to Lou Heavner, whom you may recall from an earlier neural network-related post. Lou had a great response and I asked if I could highlight some of it here.

Lou opened:

The sample frequency is related to the process dynamics. Initially you can make an educated guess and later revise if necessary. The Neural model is based on observing inputs that will affect the modeled output. There is likely to be some delay between when an input is observed and when the output is affected.

For example, if I had a conveyor with a 10-minute transit period from the point at which some measurement is made (e.g. temperature) and that measurement is important to modeling the output, say quality (e.g. moisture) from a sample port in a downstream hopper. And, the residence time in the hopper from the conveyor to the sample port is 15 minutes--I would have a sample delay of approximately 25 minutes. This is the time between when the material's temperature was measured and when it was extracted as a lab sample.

We refer to this in DeltaV as the time to steady state (TSS), although it is more accurately described as the process residence time. DeltaV Neural divides the TSS into 50 equal intervals and reads all inputs at that frequency. In this example with TSS = 25 minutes, the data frequency would be every 30 seconds. The sampling frequency of the DeltaV Neural block can be faster, but should be no slower. Having the Neural block update once per second may put some additional load on the controller, but that is the only penalty of setting the sampling rate faster.

If you use a Lab Data Entry block to enter the lab results, the Neural block will "poke" the lab data where it belongs in the file. In truth, it isn't poking as much as it is defining a delay. It is important to get the lab data timestamp right to build a successful model. That is why we go to such pains to get the delays and sample times correct. So the lab data needs to be tagged with the time stamp corresponding to when the sample was extracted. This may be and in fact probably is different from the time the sample was scheduled to be extracted, or when the lab analyst performed the analysis, or when the sample was reported, or when the historian was updated.

If you use the Lab Data Entry block and select the correct TSS, there should be no problems building a Neural model due anything related to sampling frequency. If you are unsure of the correct TSS, it is usually better to overestimate than underestimate. The biggest problems are models where some variables have a very long delay and others have a very short delay. In order to see the effect of the variables with the long delay, you may not have the granularity you need on the variables with a shorter time delay. You will need data that is accurate and that exhibits typical process variability, since the model is trying to correlate various model inputs with the model output.

Lou also passed along some guidance when attempting to build the data set for use in the creation of the Neural Network in the offline mode:

It sounds like you are limited to developing an online Neural model using historical data. There are many pitfalls to watch out and avoid. This is called data mining and sometimes success is simply not possible. The same rule for data frequency applies. It doesn't matter if you are using data from 1 hour, 1 shift, 1 day, 1 week, or 1 year. If the data frequency is slower than the real process TSS, the best you can do is create a steady-state model which is probably going to suffer in accuracy unless the process is truly at steady state and has been for 1 TSS prior to each lab sample.

Next, historical data is often conditioned, compressed, averaged, or in some way filtered. This will weaken the correlation and make a poorer model. Ideally, the process inputs should be raw snapshots. When you collect lab data, it is often collected at irregular intervals. The sample schedule may call for samples to be taken at 6:00am every morning, for instance, but the actual sample time may vary (depending on what the sample taker is doing) from somewhere between 5:30 and 6:30.

What I typically do is import my process data (Neural inputs) from the historian into Excel. Neural will find the correct time delays, so you do not need to worry about that. I create a separate column for the lab data. I populate that column with the word MISSING in each row. Then I copy the lab data into the position that corresponds with the time the sample was extracted. Neural ignores words and uses only numeric values. Another thing I do is look for any bad or missing values from the historian and replace those cells with the word BAD or MISSING. Once I have the values correctly saved in the spreadsheet, I can save it as a tab delimited value (.txt) file. Make sure the header information is correct and you are ready to train.

There is a mathematical minimum requirement for the number of lab samples required to build a model and it is related to the number if model inputs. From practical experience, I find that a model with 10 inputs will require at least 100 lab samples and all of those samples must show some variability. With 20 inputs, you would need to at least double and possible triple that amount. With lab data collected every 4 hours, you would probably need about 5 weeks' worth of data and with a TSS of 25 minutes, the process data would need to be 30 second data. You should also verify the resulting Neural model against a completely different set of data to assure yourself you have a reasonable model.

If you're building neural network models for property estimation, virtual sensors, time prediction, or other applications, I hope some of Lou's experience serves as a shortcut to your successful implementation.

MP3 | iTunes

For those attending the Emerson Exchange with distillation processes, you don't want to miss Emerson's Lou Heavner's presentation, Advanced Distillation - 102. Lou is an advanced control consultant, whom you may recall from earlier posts. He'll be presenting Wed, 9/30 at 4:15pm in the Sun 4 room and again Thurs, 10/1 at 9:00am in the Captiva 1 room.

Lou's preference is to have each session be an open, interactive discussion on distillation control challenges and solution approaches. He does have a presentation and will talk if that is what the session attendees would prefer, but he's hopeful an interactive discussion will blossom. I've gleaned some highlights from the presentation.

This is a picture of a traditional binary distillation column. The feed may be vapor, liquid, or a mixture, but is most commonly a liquid. Its composition is usually variable. Liquid in the bottom of the column is boiled up through the column and vapor leaving the top is condensed and returned as reflux. The product flows must equal the feed flow or the process won't operate very long. So, the liquid level in the bottom of the column and the liquid level in the reflux accumulator are controlled by manipulating product flows or sometimes by manipulating heat to the reboiler or reflux to the column.

Usually, composition is inferred key from temperatures in the column and controlled by manipulating heat to the reboiler and reflux flow to the column. Pressure can be controlled by venting non-condensibles (if they are present) or by controlling the amount of condensing in the column overhead.

Feed is usually not available for control, but may be in some cases. When it is available, it can be a good choice for optimization--maximizing throughput. In some cases, online analyzers are available and if they are, they may be used for control or simply monitored.

Lou stresses the fact that you can't control something that you can't measure. Online analyzers or product purity measurements are one of the key requirements for good distillation control. If there is not an appropriate online analyzer, then some kind of inferential measurement will be required. Other measurements such as flooding (a column operating constraint) and reflux ratio can be used to track performance.

Interaction is one of the defining challenges of distillation control. Interestingly, there are many ways to pair controlled and manipulated variables. Some will work well in one column and poorly in another. The whole study of relative-gain array (RGA) analysis has been developed to understand the best way to pair control and manipulated variables. This is largely dependent on factors like product purity specs, feed composition, number of theoretical stages in the column and typical reflux ratio.

Another source of interaction outside of many columns is thermal integration. It is common for the hot product to be cooled against the feed to provide some preheat and efficiency to the column. Variability in the bottom product temperature or flow rate will be recycled back into the column through the feed. Sometimes a heat pump arrangement is used to boil the bottoms against the overhead vapors, which have been compressed. This configuration is rarely seen and only practical when the overhead and bottom products have similar boiling points and non-condensables are not present.

Another factor that makes distillation control difficult is the actual process dynamics such as long time delays associated with the time it takes for liquid reflux to cascade all of the way down to the bottom of the tower. This is more problematic when one considers that vapors will rise much more quickly up the tower. The controls need to respond to both of these kinds of dynamic responses.

When looking at the distillation process, it's a classic multi-variable process with controlled and manipulated variables. You can include the material balance loops (i.e. the level control loops) in the model predictive control (MPC) strategy. The levels would be the control (or constraint) variables and the product flows would be the manipulated variables. There are extra, manipulated variables, so an opportunity exists to include optimization. There may also be additional constraints (e.g. flooding or valve positions) and measured disturbances (e.g. reflux temperature and feed temperature).

Lou goes on to describe interaction and the relative gain array analysis process to identify the best pairings to minimize the effect of loop interaction and simplifying the matrices. He also covers pressure compensated temperature pros and cons, multi-component distillation, level control, azeotropes, and batch distillation.

If you've been battling distillation issues in meeting quality specs, energy usage, yield, and/or capacity, bring these to one of the two sessions and see what thoughts Lou and fellow attendees have to offer. It will hopefully be worth your while!

Update: Here's Lou in action:

I caught up with Emerson's Lou Heavner the other day and we traded a few teenager "war stories." Lou is an advanced automation consultant and I've shared some of his expertise in posts over the years. He mentioned he had done a very basic presentation to show the interaction of operations, control, advanced control, and control strategies using the DeltaV system as his example.

I asked if I could post his presentation in my SlideShare account and discuss it here in this post. He was kind enough to agree.

Lou starts by describing key areas of the operator graphics describing the navigation, toolbars, alarm banner, and buttons to the model predictive controller (MPC) display that he typically will add to the advanced control project. He shows the loop faceplate which comes up when the operator clicks on an alarm. He notes that the operator:

...can change mode, SP, etc. He has one click access to loop tuning, alarm acknowledgement, trending, and with the appropriate privilege, he can access the engineering environment.

He shows the operator faceplate for a PredictPro MPC controller where the operator can view the optimizer, change modes and setpoints, and view the trend prediction horizon. For those who may be unfamiliar with an MPC controller, Lou shows the optimizer, which shows the variables being maximized or minimized and their associated economic value.

Lou next switches to the engineering environment where the modules and their associated parameters are located. Advanced control functions like loop tuning, neural networks, and MPC are available along with the regulatory control options. Lou shows the engineering side of creating MPC controllers from initiating automation step tests to creating and downloading the MPC controllers into the Delta controllers.

He shows an example of a composite block to calculate heater efficiency using the heat loss method. The calculation nests these composite blocks and Lou shows how to drill down and back out. He closes the brief presentation showing an on-line view which aids in troubleshooting the control strategy. This on-line view is available when the strategy is in simulate mode or actually running on-line in the controllers.

I hope Lou's simple descriptions and screen captures helps show the interaction of the advanced controls from their design through to their operation. We'd also like your thoughts on if you'd find this valuable to be seen in another form like a screencast.

My colleague, Deb Franke, pointed me to a great article in her RSS feeds. The ChemicalProcessing.com article, Innovative Fixes for Saving Energy in Plants, describes some ideas to help reduce plant energy costs. Although energy costs have come down in recent weeks, they are still one of the largest controllable costs as I have mentioned in an earlier post.

The article points out innovative solutions including dual drive pumps, variable speed motors, water/glycol systems, automated blowdown systems, low BTU sweep gas and wireless sonic leak detectors. Give the article a read if you think some of these might apply in your plant processes.

I forwarded the article to Emerson's Lou Heavner, whom you may recall from earlier advanced process control application posts. I asked what new and innovative, energy saving ideas he might have to share.

Lou had a couple of ideas. But, being the modest sort, he added a caveat that they may not qualify as new or innovative. To me, if you're looking for ways to reduce your energy costs and you didn't consider one of these, it's definitely new.

Lou's first thought was on distillation processes. He writes:

In distillation, relative volatility and hence difficulty of separation tends to improve at lower pressure. When cooling water and/or air are used to condense the overheads, the pressure is often tightly controlled for stability in the face of changing ambient conditions and the extra cooling capacity available during nights or colder weather is not fully utilized. If pressure is allowed to "float" and as much condensing occurs as is possible, pressure will fall in the column and separation will normally improve. This means less heat is needed in the reboiler and hence energy savings when using steam or some other "costly" utility stream to provide reboil.

His second thought was around combustion processes burning fuel gases with changing compositions. Lou notes:

In heaters or boilers where the gaseous fuel consists of a hydrocarbon mixture of varying composition (like refinery fuel gas), a change in fuel can have an effect on the heat generated by combustion and on the excess air level in the flue gas for a given fuel flow rate. Sometimes, if variability of the flue gas justifies, companies will install fuel quality analyzers that measure composition or heating value. In many cases, the same thing can be achieved and better flow control at the same time, by using a Coriolis mass flow meter. It turns out that the mass flow of a hydrocarbon and the "btu" flow are directly related since both are related directly to MW.

You can't do this with PT compensated flow, because it knows nothing of MW. But Coriolis measures mass directly and can be used to reduce variability of "btu" feed to the burner. This can be dramatic where the fuel gas varies significantly. It is not a good solution if the "btu" content changes due to the presence of inerts (like N₂ or CO₂) or non-hydrocarbons (like H₂ or CO), since they do not exhibit a linear relationship between mass flow and "btu" flow. But if they are present in small quantities and don't vary much, the concept can still work.

On processes that degrade the "quality" of energy, Lou shares:

Saving energy can be as simple as minimizing thermodynamically irreversible operations. Mixing, heat transfer, and throttling of process flows are common examples of irreversible processes. In general, industry should avoid over-purifying/heating/cooling followed by mixing or blending to achieve the target composition/temperature. Process design should attempt to get as much work as possible out of utilities and recover as much heat as possible. Pinch technology is one approach to heat integration design used by process engineers. Of course, there are practical limitations like capital cost considerations, dynamic response and controllability, and availability/reliability of utilities, especially ambient cooling.

Also, control valves should be selected to minimize throttling losses and allocation and valve position should be used to minimize overall pressure drop in systems like utilities where resources are shared by different units or equipment. For example, if multiple reactors are cooled with a shared refrigeration unit, the coolant temperature setpoint can be raised (reducing the refrigeration required) until one of the user's demand exceeds the capability of its corresponding control valve to deliver.

Let's hope that something between the ChemicalProcessing.com article and Lou's thoughts provides you at least one idea that can help reduce your plant's energy bills.

Continuous manufacturing processes have long benefited from the application of advanced process control (APC) in their processes to improve upon their regulatory control. Batch manufacturing processes have recently been able to take advantage of these technologies. I received an email the other from Lou Heavner, part of Emerson's Advanced Applied Technologies team. We've featured Lou's work here a few times in the past.

I'll summarize a few of these applications with the hopes that it might spark some ideas for application in your batch manufacturing process.

A manufacturer of sweeteners was having scheduling problems caused by the unpredictability of batch cycle times. End of batch could vary between six and twelve-plus hours. The operators could determine when end of batch was reached but not predict when this would occur. The APC consultants worked with this manufacturer to apply neural network technology as an inferential estimator to predict the end of batch time. The model can successfully predict the end of batch plus or minus ten minutes up to four hours before the completion of the batch. Scheduling downstream equipment is more manageable given these accurate predictions.

A second example Lou mentioned was again around batch cycle time, but in this case poor distillation control, which resulted in longer batches. Model Predictive Control was used in this pharmaceutical manufacturing process to control the batch distillation, specifically the reflux. Distillation time was reduced with the overall batch cycle time reduced by more than three hours per batch on average. The net effect of this improved control performance was a five-plus percent increase in production capacity. The quality of the product produced was also improved.

A third example is in a specialty chemical manufacturer's semi-continuous fluid bed hydrogenation reactor. In this process, cold solids are added to the top batch-wise based on level in the vertical reactor. Heated feedstock (gases) enters the bottom to provide the fluidizing medium and heat to drive the reaction. The reactor was a bottleneck, limited by temperature control and high temperature constraint. Adding model predictive control around the reactor provided more stable temperature control. The controller reduced temperature variability and allowed target to be moved closer to constraint limit with fewer high-temperature trips.

I thought these were great examples of advanced control technologies combined with people like Lou with process and APC application knowledge that are solving process problems and improving process efficiency. Perhaps these ideas will spark some ideas for improvement in your operations.

A great question came in on the Operating Fired Heaters More Efficiently and Reliably blog post:

Jim I work with natural draft heaters on a daily basis and have initiated several efficiency tests with improved burner internals. I am looking for an opportunity to optimize dual firetube treater by first off improving the combustion efficiency to 80% in each tube and then staggering the temperature controls so that one tube runs 90 to 100% of the time and the other tube only fire during high load requirements.

I sent the comment around our advanced automation consultants for any comments that they might have and I received a great reply from Lou Heavner whom you may recall from earlier posts. Lou describes how to approach optimizing these heaters:

Heater efficiency is calculated using heat loss or input/output method. Input/Output method is difficult because you have to account for lags and delays between fuel firing rate changes and the measurement of process heat absorption changes and in the specific case where there is incomplete phase change on the process side (e.g. partial vaporization) you cannot easily solve with reasonable instrumentation. The heat loss method measures heat loss in the flue gas and assumes any other losses are negligible and constant. If not, they need to be measured and added as well.

Heat loss requires knowledge of the supply air (and fuel) temperatures and the flue gas exhaust temperature as well as the composition of the fuel and flue gas, just like with a boiler. In perfect combustion, there would be no unburned fuel in the flue gas and no sensible heat losses. But due to practical considerations, there are sensible heat losses and to calculate them, you need to know the delta T between the exhaust and ambient and how much excess oxygen remains in the exhaust. Efficiency calculations made using this technique can be pretty accurate in a natural draft heater, but if there is air leakage after the combustion zone, tramp air will show up as lower efficiency due to increased O2. And there is usually an optimum cost operation where the trade-off between sensible heat losses and unburned fuel losses require some level of unburned or incompletely burned fuel leaving in the flue.

When you are ready to control, the goal is to minimize excess O2 while not allowing excessive fuel to go unconsumed. CO analyzers are often used to detect incompletely burned fuel and the goal is usually to keep it below 150 ppm or some lower target. O2 is controlled to stay as low as possible without exceeding the CO limit, which is usually 2% O2 or less for the fluegas.

You can do this with simple feedback control, but feed forward control can help do better. Information on fuel quality, if it varies, and process side temperatures and flows (the heater load demand) can be used to adjust the fuel and air for combustion to meet the heating demand at maximum efficiency. Fuel and air cross limits are often used to maintain fuel and air ratio without getting into a fuel rich condition in the firebox during load changes. But airflow is usually difficult to measure. Therefore, it is often inferred from damper position.

When evaluating an application, we would want to know what instrumentation already exists and what the process variability looks like. What efficiency are they currently obtaining? Then we would look at the control valves and any other contributors to variability to see if they warrant repair or replacement. We would similarly evaluate the instrumentation and analyzers to see if they need anything there.

Then we could evaluate the control strategy and performance and recommend reconfiguration or tuning as appropriate, which may include advanced process control (APC). The person evaluating the controls would have to weigh the cost against the improvement from better loop tuning, valve repair/replacement, CO analyzer, etc. to come up with the best solution. Dampers are often the weak link in fine control of a natural draft heater.

As my colleague Doug Simmers in Emerson's Rosemount Analytical business noted, "The commenter is probably correct with the strategy to fire one heater full out, and bring the second unit on only when needed. Running at full fire develops the best turbulence for fuel/air mixing, and the excess O2 can be kept lower." This is a load allocation problem when two heaters are firing simultaneously. If we can model heater efficiency for each heater as a function of load, then we could optimize the load allocation across both heaters when both must be fired. Actual testing would identify the models, uncover the best strategy, and verify or disprove this assumption."

He may also be interested in the efficiency calculator, developed by Doug's team.

Join the conversation and add a comment if you have experience to share.

Last October, I featured one of Emerson's advanced automation service consultants, Lou Heavner, and how he worked with Lukoil to create virtual sensors based on neural networks.

Their efforts were told in more detail in the March 2007 issue of InTech magazine. The article, entitled, Crude gets smart, described the Russian refiner's challenge to keep their refined products within specification. They had been relying on lab samples that came back from the lab to the operators only once or twice a day.

To get feedback on product quality and composition more frequently, Lou and the team used neural network blocks in their DeltaV system's controllers to create property estimators. As the article states:

The goal of a property estimator is to provide an accurate gauge of product quality, especially after lab results have become stale, which is most of the time. Property estimators are not intended to eliminate lab analyses, although the frequency of analyses may lessen once estimators are proven. Even though estimators may not be as accurate as lab analyses, they can be worthwhile calculated variables to help engineering and operations personnel monitor, troubleshoot, or understand and control the process.

The article describes the steps the team took to collect the data to train the neural network models. It offers guidance for those looking to implement property estimators. Some examples of their recommendations include:

• The time stamp should reflect the time of data extraction from the process--not when it was scheduled for sampling, or when the lab technician performed the analysis, or when they reported the lab results.
• Avoid filtering or manipulating the process data. Raw snapshot data usually makes for the best models.
• If the process does not vary much, the model will not be reliable if the process wanders into a range with no collected data... the model will be changed to "Uncertain" and the operator can be alerted.

The team believes they may have one of the world's largest installations in terms of neural network models. Currently operating models include ones measuring boiling points, flash points and viscosity on the pre-flash, atmospheric, and vacuum towers.

If operators at your plant are waiting on lab information to make quality adjustments to the process, you may have a business case for creating property estimators to augment the lab sampling process.

Emerson's Lou Heavner, a consultant on the Advanced Automation Services team, recently co-authored a paper, Using Neural Network Technology for Virtual Sensing in Crude Refining Units, at the recent ISA EXPO 2006. Lou worked with engineers from Russian oil refiner, Lukoil.

Always a great presenter, Lou not only shared his expertise on this project, but the fact he homebrews beer. I imagine this keeps his control application skills honed!

The paper describes the need to improve quality of refined products to improve upon the current process of taking lab samples once or twice a day. The operators would make changes to the process based upon the lab readings. They tried to control key temperatures and other process variables on the pre-flash, atmospheric, and vacuum columns by manipulating flows including reflux, furnace fuel, pump-arounds and product draws.

To make these quality adjustments more in real-time, the refinery engineers want to use something other than costly and maintenance-prone on-line analyzers. They decided to use neural network technology to build real-time inferential property estimators which could run inside the refinery's existing DeltaV controllers.

Lou work with the Lukoil engineers to build ten artificial neural networks measuring gasoline, kerosene, diesel, VGO, and residue on the pre-flash, atmospheric, and vacuum towers. They believe this application of virtual sensors to be one of the world's largest on a single crude unit.

The real work comes in collecting the data needed to train the neural networks. They needed around 100 lab samples for each model and the continuous historical data for process variables over this sample period. The DeltaV Neural software helped automatically perform the data collection and model training need to build and prove the neural networks. Up to 20 process variables were collected as inputs in training each of the ten neural networks. Any abnormal operating conditions were identifies to exclude the data from this time period from the model. Any of the variables that had little or no effect on the model outputs were eliminated.

The largest challenge in the data collection effort was in the lab data. It had to be accurate in terms of precise time of taken sample and the proper analysis of the sample. The quality of the neural networks is directly impacted by the accuracy of the samples. Another important factor is to make sure the process data is not filtered or manipulated, but instead a raw snapshot.

The resulting inferential sensors predicted what the lab results showed within a few degrees. The ones estimating lighter refined products were more accurate. The engineers have not closed the loop to run the control strategies based on these readings, but they do present the information to the operators to make adjustments more frequently then they could with samples coming only once or twice a day.

I caught up with Lou Heavner who you may remember from an earlier model predictive control post. When it comes to blending operations in refineries, Lou knows this application very well.

He's recently created a Blend Lite application which is a scaled down version of the Blender Control application that has been developed and deployed for numerous refiners around the world.

The full blend control application has in-line blending which offers several advantages over sequential blending, especially in the reduced time required to blend large volumes. Some products like gasoline involve many components and the finished product must meet many specifications. The standard blending package addresses the many component/many spec problem and frequently includes the use of an in-line analyzer and blend optimization.

There are a number of blending applications that involve only a few components and don't require all of the advanced functionality that is usually required for gasoline blending. Refineries have some products that fit this simpler application like jet fuel and asphalt. Other industries have processes that fit this simpler blend application. The key value in these applications is the ability to simultaneously blend multiple components--all in the correct ratio--even if one of them becomes constrained. These simpler blend applications still need functions such as: blend recipes, start/stop sequencing, failure monitor, ratio controllers, ramping and pacing.

Lou said to consider the Blend Lite Package if your application has the following:
- In-line blending for up to 5 components with one flowmeter and control valve per component
- Ratio control of current flows or totalized flows with pacing during the blend and ramping at the beginning and end of blend
- Up to 32 blend recipes
- Blend reports
- Start/stop capability for one valve and pump per component
- Basic operator graphics like those provided for the Blend Package

One of the guys who has been around as long as I have is Lou Heavner, a Consultant in our Advanced Applied Technologies team. As an MIT graduate, he is one really sharp person, and also someone who can simplify and communicate complex ideas.

I asked Lou about what he's been working on recently in the world of applying advanced process controls for process manufacturers. Lou's been recently involved in a project in the European region where he consulted with a refiner to get DeltaV PredictPro working on a crude unit and to take advantage of its optimization capabilities.

During that project, it became clear to Lou that the term optimization conjures up a different vision for most people. He was confronted with matching the capabilities of the system to the expectations of the customer and needed to dive pretty deeply into how optimization works in the DeltaV software, just to explain and train the local refinery staff.

The crude unit is the first unit in a refinery and the fractionator is a great place to start applying APC technology. An optimizer can take advantage of the capabilities of PredictPro and the extra degrees of freedom in a typical crude unit to drive the process to maximum throughput and/or minimize the energy required (as fuel in the fired heaters.) It can also support an objective to maximize yield of the more valuable cuts (products). The system is working and delivering benefits, but Lou is doing further work to quantify the results and turn this skeptical refiner into a good reference site.

Applications like these have been developed for many industries and branded as SmartProcess Optimization application packages. Some of these applications for refineries include the Fractionator Optimizer and the Heater Optimizer packages.

As one who applies the advanced process control software on a regular basis, Lou's feedback to the Emerson technology development teams has enhanced the software over a number of releases.

Lou is a regular presenter at the Emerson Exchange sharing his expertise with the Emerson customers who attend. This year he'll be presenting a short course which explains how to implement optimization with DeltaV PredictPro and surveys several optimization techniques including non-linear searches for minima, load allocation, and LP (linear programming) solutions. These optimization techniques are a key and inseparable part of those who employ use model predictive control techniques.