By developing high-fidelity models that can accurately predict the blending time required for fluids of a wide range of viscosities, ABB is helping process industries, eg, oil and gas industry, to minimize homogenization time and ensure quality mixing for better performance.

Subscribe to ABB innovation and technology news

_{Mahesh Vaze}, _{Subhashish Dasgupta, ABB Corporate Research, Process Automation Bangalore, India, mahesh.vaze@ in.abb.com, subhashish.dasgupta@in.abb.com}
_{Keila Echart, Nicolas Vairyo, Alexandre Caillot}, ABB Process Automation Aix-les-Bains, France, keila.echart@fr.abb.com, nicolas.vairyo@ fr.abb.com, alexandre.caillot@ fr.abb.com

To produce lubricants, paints, resins, pharmaceuticals, etc., process industries require innovative solutions for technologies that mix raw materials according to specific blending formulations to yield high-end products. The mixing technology needs to ensure a satisfactory level of homogeneity, and stability, within the produced fluid, while expending minimal energy during the mixing process. By optimizing blending time, and saving energy, the final product can reach the intended market faster, thereby reducing costs. The development of new blending technologies that take advantage of advances in modeling methods could help these process industries achieve customer-specific targets. Enter ABB to leverage advancements in physics-based computational modeling to simulate and predict the blending process for better performance.

Where it all happens

With more than 70 years of experience in process system implementation associated with formulation, blending and batch production, etc., ABB’s researchers have joined forces with business experts to explore how they could leverage their knowledge and experience to help customers improve blending time for a range of viscous fluids to better achieve accuracy and cost targets for customers.

To understand the entire blending process, ABB considered a generalized lubricant oil plant segregated into three different zones →01:

Raw material storage farm: Typically, this zone is made of several storage tanks for base oils and additives.
Blending units: These units are the core of the plant. They are used to mix base oils and additives in the required proportions to produce a portfolio of finished products with various specifications.
Finished product area: Finished product storage tanks and filling lines are used to fill final products into small packs, drums, bulk trucks, or railcars.

The physical blending process

The performance of such process plants is significantly dependent on the efficacy of mixing to achieve the proper combination of petroleum derivatives and other components through blending technologies. The goal is to reach the desired level of homogenization swiftly and accurately.

Jet mixers are commonly used for blending in tanks because they are relatively easy to install, operate and maintain. Jet nozzle mixers →02 are based on the Jet-Venturi effect in which the transfer of the kinetic energy from a high velocity liquid is accompanied by a balancing drop in pressure as it passes as a jet into the surrounding tank.

Homogenization of the various precursor liquids within the tank is attained as some of the liquid is entrained through a pump that then returns the liquid to the tank as a high velocity jet through a nozzle of a specific cross-sectional area. This action generates a circulation pattern within the tank that results in agitation and mixing of the contents.

Why change the process?

Given the jet mixing process described above, ABB’s experts had questions: Is the recirculation process through the jet nozzle mixers, which is utilized in the raw material storage tanks and final product storage tanks, adequate to assure the level of homogenization and mixing desired in terms of performance prediction? And, if not, how can it be improved? The designs are considered conservative.

Moreover, ABB pondered whether they could adequately predict mixing time by adjusting the various design parameters that are known to impact the efficiency of jet nozzle mixing technology.

By addressing these questions, ABB’s aim was to help plant operators establish an optimized and shorter mixing time; this would enable the homogenized product to be ready for the next operational step more quickly. The more rapid process would not only reduce energy consumption, it would also accelerate the freeing of storage tanks to accommodate another batch of base oils and additives sooner.

Analyzing the process

First, to address the fundamental question – whether the expected homogenization level matches the actual achieved level – ABB analyzed the mixing process challenges. Despite the seeming simplicity of the process, experience with suppliers on past projects shows that the high and variable viscosities of the fluids in the tanks complicate the blending process immensely. It becomes all but impossible for jet nozzle suppliers to estimate the impact of various fluid viscosities on the mixing process. Consequently, suppliers are unable to guarantee with certainty the expected performance associated with blending time – a conundrum for the process industries.

With this in mind, ABB’s experts endeavored to create a numerical model that could accurately predict the blending time for miscible fluids with a large range of viscosities, eg, non-viscose fluids, and those with low, medium, high and very high degrees of viscosity. In addition, variable influencing parameters, eg, tank geometry, physical properties of the fluid,number and arrangement of jet nozzles inside the tank and the cross-sectional area of nozzle orifices, have been taken into account to develop a preliminary computational fluid dynamics (CFD) model.

CFD modeling – the basics

Once the physical process of mixing was understood, ABB set out to specify flow and determine a suitable mathematical model to employ. Because the efficacy of CFD models to predict complex fluid dynamics is well known, ABB chose to investigate CFD modeling of a jet mixer, an eductor-based mixer, that utilizes an external source of energy or pump to mix fluids.

By combining the literature survey results [1-5] with modeling expertise, ABB established best modeling practices in turbulent flow during jet mixing. The resultant CFD blender model, also known as the blender’s digital twin, should enable mixing time to be predicted, ensure a satisfactory level of homogeneity, and potentially minimize the cost of testing. Ultimately, the model results would allow the minimum time required to reach the desired homogenization level in oil mixtures of various viscosities to be estimated, thereby improving blending performance.

Because complex flow patterns are encountered whenever multiple oil strata of differing properties, eg, viscosity, are mixed, ABB relied on best modeling practices in their study: incorporating assumptions that have minimal impact on prediction quality, developing the best discretization schemes, meshing; and selecting the best turbulent flow model to employ.

Assumptions and conditions

For this investigation, assumptions were made about the modeling system →03 and the computational geometry →04. Assuming a half-symmetric system, allowed a half-symmetric model →03 to be developed. The fluid space was assumed to comprise six horizontal layers; each layer was prescribed a unique viscosity and density. The absolute height of oil strata, oil densities and their viscosities ranged from 0.015 m to 4.651 m, from 828 to 1,000 kg/m³ and from 0.1059 to 7.788 kg/m-s, respectively. An exterior piping that recirculates the fluid mixture was added to the tank model →03.

Setting boundaries

To ensure an accurate solution to the Navier-Stokes equations (or equations for mass, momentum and conservation) used, boundary conditions were applied to the domain boundaries for CFD modeling. These included inlet, outlet and wall boundary conditions, among others. As mentioned previously, a recirculation piping system →03-04 was provided to the exterior of the mixing tank, connected to the base of the tank to simulate the recirculation effect that is driven by the real-world pump in the factory. This way, modeling of an actual pump and the associated complications could be avoided.

Exploring modeling methods

To determine the longest possible mixing time needed to reach homogenization, ABB considered an extreme case of oil stratification in this study: In this worst-case scenario, six completely segregated layers of oils with distinct viscosities (the lowest viscosity was 0.1059 kg/m-s and the highest was 7.788 kg/m-s) were simulated along the height of the tank.

To select the best modeling method, ABB explored species modeling and multiphase modeling. The first method models oil layers as different chemical species while the latter models the layers as physical phases. During this comparative study, the mixing and transport of species was simulated by solving fluid flow equations and species transport equations comprising convection and diffusion terms. Despite the suitability of the method for investigating the blending process, since oils are not diffusive, running the species transport model was found to consume a significant amount of computational time. The multiphase alternate approach was found to converge to a solution using less computational time than did the species transport model. Thus, the main simulations were performed using the mixture multiphase flow model. Computations were performed by solving the equations of fluid dynamics, high-end turbulent flow models simulated the mixing of the oil layers as a function of time.

Multiphase model predictions

For the simulation experiment, the eductor was inclined with respect to the vertical axis of the tank by 20 degrees. All necessary data related to geometry, flow and fluid were obtained from the business data sheet supplied →03-04.

To investigate mixing homogeneity using a qualitative approach, viscosity contour plots, across the symmetric plane, were observed at various time instances. ABB experts assumed that an adequate homogenization level would be reached when the contour plots’ layers were indistinguishable →05. Prior to the experiment, the arbitrarily determined mixing time was set as t(arb). The viscosity contour plots →05 were monitored continuously until t(arb). The results show that only a negligible difference in the viscosity plots was observed at: t(arb) – Δt. Therefore, a mixing time of t(arb) – Δt, is considered an acceptable mixing time to achieve a satisfactory level of homogeneity within the tank. The positive results not only demonstrate that it is possible to achieve a reduction in the required mixing time, but imply a reduction in the mixing energy too.

Advanced modeling with ROMs

While CFD models are globally accepted as high-fidelity tools to predict complex processes, using the models to perform multiple parametric studies is perceived to be challenging, due to the vast time and memory requirements. For these reasons, ABB evaluated the feasibility of using reduced order versions of CFD simulations also known as reduced-order-models (ROMs). While ROMs keep the indispensable features of CFD models, they utilize substantially less time and memory, thereby reducing the associated costs. Not surprisingly, this promising yet simpler modeling approach is gaining attention [6].

The process of ROM development involves running multiple simulations by changing a few important input parameters and computing the output. Advanced fitting techniques, eg, vector fitting, are then used to obtain a response surface; this surface provides the desired output that corresponds to any given set of input parameters.

To develop the model, ABB designed a simplified mixing tank with two oil layers of varying heights and properties →06. A ROM was constructed and the heights of the oil strata were varied in multiple simulations. The height of oil 1 was varied in proportion to the height of oil 2, while the total oil height was assumed to be fixed. The result is a relationshipbetween homogeneous mixture viscosity, mixing time and oil height →07. By using this relationship, the mixing time needed to reach homogeneity for a given oil strata height can be deduced. These results support the the use of the ROM model to investigate parameter impact and develop a physics-based digital twin.

Providing the best results for customers

Aiming to offer world-class blending solutions to customers →08-10, ABB is pursuing the development of advanced analytics aided by high-fidelity models to minimize homogenization time and ensure quality mixing. The collaboration between business experts and research scientists has led to the development of multiphase CFD and ROM models capable of predicting the blending process. Blender digital twins are not only useful tools for predicting performance and estimating the minimum mixing time required, but can be used to study important mixing parameters to suit the needs of specific customers.

_{References

[1] F. Al-Qaessi and L. Abu-Farah, “Prediction of mixing time for miscible liquids by CFD simulation in semi-batch and batch reactors”, Engineering applications of computational fluid mechanics, vol. 3, no. 1, 2009, pp. 135 – 146.

[2] J. J. Derksen, “Blending of miscible liquids with different densities starting from a stratified state”, Computer & Fluids, vol. 50, 2011, pp. 35 – 45.

[3] A. W. Patwardhan, “CFD modeling of jet mixed tanks”, Chemical Engineering Science, vol. 57, 2002, pp. 1307 – 1318.

[4] J. Thomas, et al., “A CFD digital twin to understand miscible fluid blending”, AAPS PharmSciTech, vol. 22:91, 2021.

[5] K. L. Wasewar and V. Sarathi, “CFD modeling and simulation of jet mixed tanks” Engineering applications of computational fluid mechanics, vol. 2, no. 2, 2008, pp. 155 – 171.

[6] M. Vaze and S. Dasgupta “Modeling flow”, ABB Review 2/2023, pp. 146 – 151.}

Select region & language

In the mix: Computational fluid dynamic modeling of blender jets