(594a) Using CFD to Simulate Tank Clean-in-Place Process

In the chemical, food, process and pharmaceutical industries, large tanks are filled, emptied, and refilled with various products many times. Prior to refilling the tank with new product, manufacturers must ensure the previous batch residue is completely removed from the tank’s internal surface, as not to affect subsequent product quality. A clean-in-place (or, CIP) system is the typical procedure used to clean a tank after use and before product changeover.

Depending on the product being removed and the equipment size, manual cleaning is usually not attractive economically and a clean-in-place system is desired. A clean-in-place process can be automated and standardized. The process involves installing a dynamic nozzle system, in which water or cleaning solution jets from a nozzle assembly, figures 1 and 2. The nozzles may be forced to move due to the liquid flow (passive) or be motor driven (active). The motion is typically bi-axial, figure 2. This bi-axial nozzle motion and the jetting water/solution form a lattice of track marks on the tank walls which eventually overlap to clean the entire tank’s interior. Typical objectives are obtaining a fully cleaned surface while using the optimal amount of water or solution and within the shortest cycle time. Typical input parameters are water pressure, the bi-axial motion of the nozzle assembly, nozzle geometry, and nozzle assembly placement within the tank (e.g. offset from the tank centerline as well as height within the tank). For example, by optimizing the nozzle rotation speeds, flow rate, and nozzle placement it can be ensured that the coverage by the cleaning agent to the mixing tank surface will be maximized. In addition, the impact strength of the impinging jet on the mixing tank surface can be monitored which is directly related to the “mechanical” cleaning action.

Geometrically, this problem spans multiple length scales. The smallest length scale that needs to be captured is that of the jet diameter (mm range). The industrial sized tanks are typically in the multi-meter length scale order of magnitude. Another challenge stems from the variation in time scales that need to be resolved as well. Local jet speeds of multi-hundred meters per second need to resolved stemming in a residence time of around 0.1 seconds form the time of jet departure till impingement. On the other hand, the nozzle assembly is bi-axially rotating at a very low rpm and hence to cover the entire tank with cleaning liquid, multiple “cycles” of rotations are required resulting in physical time in the order of seconds, minutes, or hours.

Due to the large variation in length scales (mm for nozzle vs. several meters for the tank) and time scales (jet flow vs. tank surface coverage) two modeling approaches will be presented. The first approach, tank level, will highlight how the complex bi-axial nozzle motion (combination of spinning and orbiting motion), figure 2, can be controlled to affect the coverage of the mixing tank surface with cleaning agent. A Lagrangian particle tracking approach (Discrete Phase Model or DPM) will be used to model this process combined with a the lagrangian wall film (LWF) model to track and represent the water film spreading on the wall. This model will be very efficient to represent the surface coverage as the nozzles spin. Initial results are shown in figure 3 where the instantaneous wall film thickness is shown on the tank surface after 0.15 seconds.

The second modeling approach, jet level, will focus on the details of the individual jets to accurately predict the impinging force on the mixing tank surface. A detailed Eulerian (Volume of Fluid or VOF) approach will be used to model the details of the jet. To maintain the fidelity of the narrow jet, coming from an orifice of 0.5 mm, over such a long distance, automatic mesh adaption will be implemented to dynamically refine the water/air interface at least up to impinging with the water surface. The main outcome of this simulation will be the wall shear stress that the jet exerts on the tank walls. Initial results are shown in figure 4 where shear rate on tank wall is displayed.

By combining the two modeling approaches the accurate prediction of the impingement force on the mixing tank can be predicted by the Eulerian (VOF) model while the Langrangian (DPM) model will be used to efficiently predict the entire tank coverage. Finally, the effects of operating parameters like injector rotation speed, flow rate and nozzle placement will be highlighted.