Eastman Chemical Company is the world’s largest supplier of polyester plastics, specialty chemicals and basic chemicals.

Many of Eastman's manufacturing processes involve liquid and gas-phase unit operations with mixing, reactions, separation and non-Newtonian flow. The company frequently uses CFX software to improve the reliability of these processes and to add value to the various plant sites.

This case study, describes an example of how CFX helped to solve a recurring problem, and saved Eastman US$2M in the process.

Situation

In one particular production facility, a corrosive additive is injected into a flush-mounted injection tee. The additive is introduced onto the inner wall of the process pipe, and with very little radial flow in the process flow field, the only mechanism for mixing is diffusion. At other plants, where similar additive injection methods have caused corrosion problems and costly shutdowns, engineers have gone to great lengths to design injection systems to minimize corrosion. For this specific plant, the engineer wanted to study low-cost options for improving the mixing of the additive.

Integrated design solution

Because of the ease with which CFD can perform complex parametric studies involving potentially costly process modifications, all in the framework of the virtual world of computer analysis, we decided that it was the ideal tool for optimizing the design of the injector. To achieve optimal results within a limited project time frame, we selected CFX-5.

CFD model

We looked at twelve different basic designs, with further parametric variations on several of these. For each one, the computational model of the mixer was set up using CFX-5's preprocessor, CFX-Build.

For the particular model presented in the illustration below, it took about one hour to build up the geometry of the mixer, and five minutes for the machine to create the 3-D mesh which contains about 272,000 cells. The power of CFX-Build allowed us to build the geometries parametrically in a very short time and thereby study many possible alternatives.

In the calculations, the flow in the domain has been considered as single-phase with the additive being introduced as a species of identical density and viscosity. In the original design, the additive was injected at the inner wall of the process pipe, but we have considered different injection methods, including side pipes and centerline injection.

Simulation

Accurate prediction of mixing behaviour is essential in this type of simulation. To minimise numerical diffusion, which can cause over-prediction of mixing, we used CFX-5's high-resolution bounded discretization scheme. This ensures high accuracy and good convergence while still maintaining physical results that are free of spurious spatial overshoots and undershoots in the solution variables. Minimal numerical diffusion was also achieved by using CFX-5's grid adaption capability. In this case, the grid was adapted 3 times according to the additive concentration gradient leading to a final mesh of about 490,000 elements.

Original mesh.	Mesh after 3 adaption cycles.
CFX-5 dynamically adapts the mesh at the mixing interface between the additive and the process fluid.

Typical CPU time required for the calculation was two hours for a 272,000-element run, and two hours more for the calculation with the three adaption cycles. All simulations were performed on an NT workstation with a single 500MHz processor

Mesh adaption refines the contours of the additive in an axial plane through the pipe. Left, without adaption; right, with adaption. This shows that the adaption produced significantly better results by reducing numerical diffusion of the additive.

Results

Our CFD investigations allowed us to identify a number of options for generating secondary flow and mixing in the process fluid. To make quantitative comparisons of each option, we used a normalization parameter, the quotient of the additive reduction factor and the relative increase in pressure loss.

Conclusion

Now finished, the study has allowed Eastman to develop a new design which will be implemented shortly, with an estimated added value from this work of around US$2 million due to reduced corrosion, and hence frequency of plant shutdowns.