ANNUAL TRANSACTIONS OF THE NORDIC RHEOLOGY SOCIETY, VOL. 22, 2014 A New Numerical Method for Correction of Wide Gap Rheometry Data by CFD Naser Hamedi1, Johan Revstedt1, Eva Tornberg1, and Fredrik Innings2 1 Lund University, Lund, Sweden. 2 Tetra Pak Processing System ABSTRACT must be used. The flow of a Newtonian In the present study an approach based fluid will then deviate more from the on computational fluid dynamics (CFD) is classical Taylor-Couette solution and, used to find appropriate correction factors hence, the shear rate distribution over the for rheometer data. gap is not known a priori. This longstanding Two fluids, a mineral oil and Carboxyl problem in rheometry is often referred to as Methyl Cellulose (CMC) 1.5% were used. the “Couette Inverse Problem”1. To solve Hence, both Newtonian and non-Newtonian this problem, the flow curve is derived from effects are considered. CFD together with the measurement data obtained by the the integration approach was used to correct experiment. Note that the terms narrow and the measurement data and the calculated wide should not be interpreted in an viscosity. The results of the wide and absolute sense, but rather as relative narrow gap rheometry were compared and quantities, i.e. the distance between the the torque contribution for each part of the rotating bob and the stationary cup in bob was specified. By the new approach, a relation to the height of the bob. correction factor for the end parts was found Several methods have been proposed to for narrow and wide gap measurement data. solve this problem. A widely used method is The results show that the correction factor to based on the analytical solution of the be larger for the wide gap. In addition, governing equations and is the “integration higher values of the correction factor were approach”2. Other approaches are mainly found for the non-Newtonian liquid. The based on numerical solution of the ill-posed CFD approach has a good potential of integral in Eq. 13: further improving the accuracy of rheological measurements. (1) INTRODUCTION The Couette viscometer is widely used In Eq 1, Ro and Ri are the outer and to measure the viscosity of the fluid. In inner radius of the cup and bob, order to keep the level of measurement respectively. is the rotational speed and uncertainty low, one normally aim at using is the shear rate. as narrow gap as possible between the bob There are studies addressing the solution and the cup. However, for of the Couette inverse problem based on the fibrous/particulate fluids where the size of solution of Eq. 1. To extract from Eq. 1, the particles is relatively large, a wider gap the integral should be inverted. This integral inversion is the source of the problem in the 175 wide gap rheometry. To solve it, numerical In Eq. 4, the effects of the end parts of the differentiation should be carried out on the bob are neglected. noisy experimental data which requires The aim of the present work is to selecting suitable algorithm4. Among the develop a method, based on computational numerical methods used to achieve this, one fluid dynamics (CFD), to find a correction may mention the method by Yeo et al.5 factor that will in combination with the based on Tikhonov regularization and the integration approach improve the accuracy wavelet-veguelette decomposition method of wide gap rheometer data. by Ancey6. One may also employ Furthermore, we aim to quantify the experimental methods in order to find torque contribution from the end surfaces proper coefficients for the measurement and how this is influenced by the width of correction. Barnes7, for example, proposed rheometer gap, and the power law index of an experimental method to correct the wide- the fluid. gap viscometry data based on the narrow- The dimensions of the wide gap gap approximation. rheometer are Ro=13.75mm, Ri=7mm, Neglecting the end effects one may hbob=21mm and hcup=62.5mm. For the directly relate the shear stress on the bob- narrow gap the bob dimensions are surface to the torque (T): Ri=12.5mm, hbob=37.5mm and the same cup. (2) MATHEMATICAL FORMULATIONS The description below is based on the where h is the bob height. assumptions that the flow is steady, laminar However, the shear stress on the end and without secondary motion, which is surfaces will also contribute to the measured 8 indeed the case considering the very low torque. Based on Eq. 2, Nguyen and Boger Reynolds and Taylor numbers. Considering proposed a method for calculating the shear the modelling of the non-Newtonian fluid stress on the cylindrical surface accounting used, the power law behaviour is only valid for the shear stress on the end surfaces. As a in a certain range of shear rates. Hence, to as first approximation they assumed that the accurately as possible model the flow of end surface shear stress is evenly distributed CMC and to avoid numerical difficulties at and equal to the shear stress on the very low shear rates a constant viscosity was cylindrical surface: applied below a certain value of . Finally, the estimation of the flow properties namely (3) the “power law index” (n) and the viscosity/consistency index (m) is crucial and sensitive to the existing noise in the By solving Eq. 1 and using Eq. 2 and the experimental data. Therefore, a local linear relation between shear rate and shear averaging method was applied to reduce the stress for a Newtonian fluid one may relate noise before calculating m and n. the shear rate to the angular velocity of the Eq. 5 and Eq. 6 are the result of solving bob, as has been shown by for example 9 Eq. 1 using Eq. 2 to relate the torque to the Steffe : shear stress, i.e. what is here denoted as the integration approach, for Newtonian and (4) power-law fluids, respectively. Power-law index (n) is calculated as and the 176 consistency index (m) could be calculated Fig. 1b shows the result of the grid afterwards. is the rotational speed in rad/s. sensitivity study using the computational domain for wide gap model. X-axis represents the number of grids across the (5) gap and Y-axis denotes the measured CFD torque. As can be seen, there are only minor (6) differences in the torque for the three finest grid resolutions. The same series of simulations were performed for narrow gap (Pa.s) is the viscosity in Newtonian numerical model for =1 rad/s. fluids in Eq. 5. From Eq. 5 and Eq. 6, one can define and as the coefficients, which are used to extract the fluid properties in the wide gap rheometry in Newtonian and non-Newtonian fluids, respectively. For further details on this, see Steffe9. In the present method, the integration (a) (b) approach is used to get a first estimate of the Figure 1. 1a) Discretized wide gap model, viscosity. This value is then introduced into 1b) Calculated torque for several grid the CFD model. resolutions. The coefficient C using Newtonian fluids has a unique value for every experimental EXPERIMENTAL SETUP set-up, i.e. it is only dependent on the The experiments were performed in two geometry of the measuring unit. For non- series measuring the rheological properties Newtonian fluids it also depends on the of Newtonian and non-Newtonian fluids. A power-law index which means that for every mineral oil is selected as the Newtonian rheometer and for every non-Newtonian fluid and Carbon Methyl Cellulose (CMC) fluid, a correction factor must be calculated, 1.5% is selected as the non-Newtonian fluid. which is usually neglected in the The experiments were examined in measurements activities. Kinexus® rotational rheometer (Malvern Instrument Ltd., Malvern, UK) with both NUMERICAL MODEL narrow and wide gap bobs. A 30º slice of the three dimensional (3-D) model was considered as the computational RESULTS AND DISUSSION domain, Fig. 1a. The simulations were performed on a Shear rate distribution structured hexahedral mesh using ANSYS Fig. 2 shows the shear rate distribution CFX 14 (ANSYS Inc., Canonsburg, PA). A in the wide and narrow gap rheometry for a periodic boundary condition was used for Newtonian fluid. As shown in Fig. 2, the the side surfaces and for all solid surfaces shear rate clearly changes along the wide no-slip conditions are used. The pressure gap while in the narrow gap less variation reference point was placed at the top cup can be observed. We expected to have a side. variation in the shear rate distribution in the 177 wide gap, but in the narrow gap the effect of Table 1. Relative contribution to the torque, the top and bottom parts on the shear rate in percentage, of the different parts of the was observed, too. This is shown by the bob. small shear rate zones on the bob ends in Gap bottomtop face rod both gaps and a slight difference between Oil narrow 3.4 2.7 93.5 0.4 the shear rate values on the bob-face and the wide 6.7 5.9 85.5 2.0 cup. CMC narrow 6.2 5.1 87.4 1.4 As is evident from Fig. 2b, in the wide gap, wide 8.0 6.8 82.2 3.0 the shear rate changes along the wide gap Comparing narrow and wide gap height while in the narrow gap, it is rheometry results for Newtonian fluids in approximately constant over the height. Table 1, one finds that the relative contribution from the end surfaces is about twice as large in the wide gap case.