Water and its associated electrolytes influence a vast number of chemical reactions and processes, including oxidation and reduction, neutralization, pH control, flue-gas treatment, waste stream clean-up, mineral scaling and corrosion. These, in turn, affect the design of equipment and processes as well as lifecycle optimization — and, therefore, productivity and profitability.
So, properly accounting for water chemistry and its evolution is crucial.
“However, while water-chemistry-based process modeling optimizes the performance of assets and processes, it’s very hard to do right,” cautions A. J. Gerbino, vice president, application consulting and client success for OLI Systems, Parsippany, N.J. “It’s not something that should be considered just as a side application.”
Doing it right means having comprehensive capabilities for electrolyte modeling and simulation — so the company’s software products cover electrolyte chemistry analysis, aqueous corrosion modeling, process flowsheet calculations and process modeling integrations.
“What’s different about our approach is that we use electrolyte thermodynamics to understand the core mechanisms that are at work in, for example, redox reactions, ion exchange processes and surface complexation. It is essential that we model all of these,” he adds.
Gerbino cites the growing use of reverse osmosis (RO) membranes in water treatment as a classic example of why this approach is important.
In one project, the operating company had multiple species of dissolved boron present in its wastewater.
“Standard RO modeling packages treat these species — and different species of many other contaminants — as the same. They aren’t. We have developed RO object calculations that use the properties of the different cations, anions and neutrals in the water to predict their permeability across membranes,” he explains.
Once the composition of the permeates and concentrates was calculated accurately, OLI Systems’ software tools estimated the scaling hazards they pose and advised the operating company on how to optimize the use of scale inhibitors.
“Note that we don’t solve the problem; what we do is provide a better understanding of the water being used and how much precipitate there could be — so dosing can be much more accurate,” says Gerbino.
This same knowledge also is key to optimizing emerging membrane technologies, for example, nanofiltration, osmotically assisted RO, ion-selective membranes and membrane distillation.
Figure 1. Alloy corrosion is one of many issues that modeling is tackling. Source: OLI Systems.
It can provide insights on issues as diverse as amines and their associated hydrochlorides, mercaptan chemistry, alloy corrosion and working out the optimum salinity in cooling water applications (Figure 1). A common use is to identify causes of mineral scaling, e.g., with acid gas scrubbing; there, the risk of such scaling rises if the concentration of the metal carbonate gets too high. Condensation in the overheads of hydrofluoric (HF) acid alkylation units, used by refineries to boost octane levels and reduce knocking, can cause corrosion problems, too.
The June 21, 2019, blaze at the Philadelphia Energy Solutions refinery, which released more than 5,000 lb of HF acid into the atmosphere, underscores the risk. The U.S. Chemical Safety and Hazard Investigation Board subsequently found that the likely culprit for the initiating event, a faulty pipe elbow, was just 0.012-in. thick (learn more here). The company has since gone out of business.
(Editor’s note: HF alkylation also poses significant challenges for sealing of flanged connections; for details on how one refinery has achieved unprecedented sealing performance, see: “Refinery Tames Tough Sealing Service.”)
“The bottom line is that modeling should be used wherever water is being used in a process,” stresses Gerbino.
The company now is focused on unlocking the benefits of advances in software automation, the Industrial Internet of Things, analytics, cloud platforms, artificial intelligence and predictive maintenance.
At the heart of this effort is its electrolyte chemistry digital-twin framework that carries out data reconciliation and analysis of plant-level chemistry models using operations data. This will deliver improved accuracy in three areas: prediction of plant electrolyte chemistry behavior; sensitivity analysis and planning for asset/process design; and soft sensors to non-invasively predict many key parameters including pH and composition.
Early 2021 should see the launch of a new series of products, including OLI Cloud Apps, OLI App Builder Platform, OLI Cloud application programming interfaces (APIs) and OLI Optimizer Tool, that take advantage of the development.
“Our Cloud APIs will give users rapid access to the information they need in a way that is much simpler for them to scale and integrate with other process simulation and optimization tools which a plant is using,” claims Vineeth Ram, the company’s chief revenue officer.
Evolving Constraints
The challenge for many water-treatment systems is that their original design constraints don’t remain fixed over their lifecycle, notes Jason Nichols, advanced analytics leader, Suez – Water Technologies & Solutions, Niskayuna, N.Y.
Operating an existing design with new constraints while maintaining safety margins requires an operational knowledge of the water chemistry and how that chemistry evolves over time, he stresses.
“If the evolution of water chemistry is considered at design time, rather than designing to static maximum load/minimum quality requirements, operators could better manage changes in constraints during the lifecycle of a system. It often isn’t. However, you do need to know a heck of a lot about the water chemistry involved when designing, or reimagining the design, for water treatment systems to be optimal over their entire lifecycle,” he explains.
Usually economic pressures are the short-term drivers for Suez customers needing to reduce operating costs, most commonly by producing more or cleaner water with the same size equipment, or by cutting energy or maintenance costs.
Heat exchangers exemplify this, Nichols notes. Typically only investigated during planned maintenance outages, they are coming more under the spotlight as companies look to use renewable fuels and decrease greenhouse gas emissions.
“People want to see what is happening in terms of underlying trends within heat exchangers which affect, for example, fuel and anti-scalant consumption. Even if heat transfer efficiency is improved just by an extra 0.25%, it’s worth a lot of money over time. So we use models that can be executed in real time from plant data to find ways of improving their availability and reliability,” he says.
Corrosion chemistry is another hot topic for Suez, reflecting operating companies’ concerns about expensive retrofits as well as harder access to capital. So, the firm has developed a number of proprietary models for various water chemistries to help equipment last longer. “Understanding how corrosion is actually evolving in a system in real time is key for condition-based maintenance plans to make that happen,” Nichols notes.
Driving all this modeling work is an effort to understand the true variance of process data (Figure 2).
Figure 2. Understanding the true variance of process data requires a data-driven approach. Source: Suez – Water Technologies & Solutions.
Nichols points out that many chemical processes typically run at steady state — so even with large amounts of data, the signal often doesn’t show much variance. However, when external pressures, such as economics, weather leading to shutdowns or differing water qualities, disrupt steady-state assumptions, process monitoring and analytics can’t account for much of the variability. Hence, data-driven approaches are necessary, he says.
“The solution is to not try to infer that which you already know. In my experience, the amount of data needed to statistically infer a process model is far less than what’s needed to calibrate one.”
As an example, he cites an ultrafiltration (UF) fouling model the company is working on. It has studied the reasons for underlying changes in system performance and generated equations to describe the non-fouling-related variations in the system.
“So, we can model a wastewater UF system and account for process variances as the process goes up and down in production. This is simply real-time process normalization but it allows us to then use data-driven models to infer the long-term fouling and short-term external factors that aren’t accounted for. We take the same approach with heat exchangers, falling-film evaporators, compressors, UF/RO systems, etc.,” he explains.
The next step for Suez is getting process models to correctly simulate dynamic controls. However, uncharacterized hysteresis in system responses and long time constants for degradation in its models make predicting where a system is going into the future challenging.
“This is also tough to get using a data-driven solution. Most plant owners won’t let us drive their plant to the edge of the performance envelope just to collect some data for our models! Being able to model non-steady-state process dynamics means we can do simulations and drive the simulated plants into known fault zones and create algorithms to spot these situations in the field,” Nichols concludes.
Assessing Scenarios
The value of modeling lies in the ability to run a lot of ‘what if?’ scenarios without the need for laboratory or pilot testing, says Matt Gerhardt, vice president of industrial water at Brown and Caldwell, Walnut Creek, Calif.
“It’s both faster and less expensive. So, for example, we can demonstrate if an existing water-treatment system is adequate or whether it needs to be expanded/upgraded — and the implications of such decisions on overall plant performance, optimization, opex, capex, etc.,” he points out.
Operating companies typically call upon the firm to carry out modeling for material compatibility, permit compliance, evaluation of wastewater treatment chemicals when a plant is expanding or a new product is being brought online, and assessment of alternative treatment chemicals.
Brown and Caldwell models physical and chemical treatment processes with OLI Systems Flowsheet and Studio, and biological treatment processes with BioWin from EnviroSim, Hamilton, Ontario.
One recent project involves a specialty organic chemicals manufacturer that produces a waste requiring anaerobic treatment. The waste also contains a lot of nitrogen. “We used BioWin to model and create a series of anaerobic and anoxic treatment tanks linked in series without inter-stage sludge removal,” notes Gerhardt. Subsequent pilot tests with the new process convinced the chemical maker to build a full-scale $100-million facility.
Another involves a starch manufacturing plant in the U.S. that modifies starch with phosphorus to use as a thickener in food processing. “However, the company’s wastewater contains an enormous amount of phosphorus — a sizable fraction of the state’s total phosphorus discharge,” he explains. Following requests by the local regulator to reduce this, Brown and Caldwell modeled the process and now is doing pilot tests to recover the phosphorus using calcium and then turn it into fertilizer for resale.
Another project posed an interesting situation: the treatment plant was deemed to have too much total ammonia — the product of un-ionized ammonia (NH3) plus ionized ammonia (NH4+) — in its wastewater. Here, modeling the wastewater flow from the outfall pipe provided an understanding of how pH and total ammonia concentration change with distance from the outfall.
“We showed that the concentration of un-ionized ammonia and pH dropped within a very short distance, due to both dilution and conversion to ammonium ion. It turned out that this was happening in such a small volume that the existing effluent discharge wasn’t an issue: it was an argument accepted by the local regulator.”
New processes demand new wastewater-treatment strategies, stresses Gerhardt, citing as an example the situation faced by a company developing a COVID-19 vaccine.
“I can’t say much about this except that it was a brand-new process with new compounds for which treatability and toxicity data were not available. We developed a plan for evaluating treatment in parallel with the vaccine development and other testing to prevent wastewater treatment from becoming the rate-limiting step. I’d say what was unique about this was that a lot of things that ordinarily would have been done sequentially were happening in parallel to meet schedule.”
A couple of modeling issues still pose challenges for his engineers, Gerhardt admits. One relates to the adsorption strategy the company uses to remove arsenic and selenium from water and wastewater. A common treatment system involves adding an iron salt, adjusting pH to precipitate iron oxyhydroxide, and then adsorbing the arsenic or selenium onto the iron oxyhydroxide surface. “I do not know of a good way to model this, because the amount adsorbed per unit mass of iron is not fixed and is not easily predicted.”
Also, he would like to be able to simulate chemical and biological treatment technologies at the same time because they increasingly are being used together. “At the moment, [we] have to export data from one into the other. I think that’s because they are really two very different fields of expertise,” he notes.
Seán Ottewell is Chemical Processing's editor at large. You can email him at [email protected].
