Viscosity Models for Nano Fluids A number of models had been derived to predict the viscosity of nano fluid.

Some of these models are derived from certain fundamental equations and others are made from the experimental data. A few of those models are used here and are as follows: Einstein (1906) Model A. Einstein was the first person who developed a model to calculate the viscosity of solid liquid suspension. He assumed the linearly viscous fluid that contains dilute suspended spherical particles. He calculated the energy dissipated by fluid flow around a single particle and then by associating that energy with the work done for moving this particle relatively to surrounding fluid, he obtained

r = nf / bf = (1+2.5)

(1)

Where and are respectively, fluid dynamic viscosity and particle volume fraction; the subscript bf, nf and r refer respectively to the base-fluid, nano-fluid and to the nano fluid to the base fluid ratio of viscosity. His model is found to be valid for relatively low particle volume fraction say 0.02 approximately. Brinkman (1952) Model H.C. Brinkman model is the extension of Einstein model to concentrated suspensions. The expression is derived for viscosity of solutions and suspensions of finite concentration by considering the effect of the addition of one solute molecule to an existing solution, which is considered as a continuous medium. The model is given as:

r = nf / bf = 1/ (1-)2.5

(2)

Where and are respectively, fluid dynamic viscosity and particle volume fraction; the subscript bf, nf and r refer respectively to the base-fluid, nano-fluid and to the nano fluid to the base fluid ratio of viscosity. This model is found to be suitable for a volume fraction of 4%.

Krieger and Dougherty (1952) Model I.M. Krieger and T.J. Dougherty developed a semi empirical equation for calculating the relative viscosity which can be expressed as
] nf = bf [1 - (/ m)]-[ m

(3)

Where m is the maximum packing fraction which varies from 0.495 to 0.54 under quiescent conditions and is approximately 0.605 at high shear rate, and for randomly mono dispersed spheres, the maximum close packing fraction is approximately 0.64; and [ is the intrinsic ] viscosity whose typical value for mono dispersed suspensions of hard spheres is 2.5. This model gives the value of relative viscosity slightly more than the Einstein model. This model reduces to Einstein model if a monomial expansion is performed, whereas a binomial expansion will lead to another famous model named Batchelor Model. Lundgren (1972) Model In this model, stokes flow through a random, moderately dense bed of spheres is treated by a generalization of Brinkmans (1952) method, which is applicable to both stationary beds and suspensions. T.S. Lundgren proposed the following equation under the form of the Taylor series in

nf/bf= [1+2.5 +6.252+ O(3)]

(4)

It is obvious that when the terms of second or higher order of are neglected, this formula simply reduced to Einstein Formula. Batchelor (1977) Model For > 0.01, hydrodynamic interaction between particles became dominant as the disturbance of fluid around individual particles interacts with that around other particles. The viscosity in such a case was derived by G.K. Batchelor, which is

r = nf/bf = [1+2.5+6.52]

(5)

Brownian motion of nano particles and the interaction between them was taken into account in this model and it is an extension of Einstein Model. This model is valid for a particle volume fraction 0.1, where motion of single particle and pair particle interaction become dominated. Nielson (1970) Model A generalized equation is proposed for the relative elastic moduli of elastic material. By introducing a function which considers maximum volume packing fraction of filler phase, the moduli of many types of composite systems can be calculated. Thus L.E. Nielson model is given as:

nf = bf (1+1.5p) ep(1-m)
Wang et al. (1999) Model Wang et al. proposed a model to calculate the viscosity of nano fluid as follows:

(6)

nf = bf (1+7.3+1232)
Nguyen et al. (2007) Model

(7)

Nguyen et al. has proposed a correlation for computing the CuO-water viscosity, which is as follows:

r= nf/bf= (1.475-0.319+0.0512+0.0093)
Saito (1950) Model

(8)

N. Saito developed a model based on a theory for spherical solute molecules in which a single solute molecule is placed in the field of flow, obtained by averaging over all the possible positions of a second solute molecule. Saito model is given as:

nf = [1+ (2.5/ (1-))] bf = [1+2.5+2.52+ -------] bf

(9)

Evaluation of Models Models listed in section I are evaluated by using them to predict absolute and relative viscosity for a range of experimental conditions and compared the predicted results with experimental data in given range of temperature and volume fraction. Error in predicting the viscosity by various models is presented in table. The following parameters are considered here for the purpose of calculation: 1. Viscosity of (EG + Water) base fluid given in Namburu et al. (2007) from 92cP at 35C to 1cP at 50C. 2. Viscosity of distilled water base fluid given in Nguyen et al. (2007) is varying from 1cP at 20C to 0.4 cP at 70C. In figure 1, the experimental values of absolute viscosity taken from Namburu et al. (2007) for 29nm CuO particles with base fluid of EG + Water (60-40% in weight) at a volume fraction of 1% in the given range of temperature are composed with the viscosity values predicted from theoretical and empirical models.

Figure 1: Experimental vs. predicted values of absolute viscosity with increase in temperature at 1% volume fraction for 29nm CuO particles

Figure 1 shows that none of the model except Nguyen et al. (2007) could exactly predict the viscosity values. Einstein (1906) could not predict the values exactly even if the volume fraction is very low i.e. of the order of 0.01. Batchelor (1977) could also not predict the viscosity values although it considers Brownian motion of solute particles and interaction between particles. It is also found that the viscosity values predicted from Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) are so close that these values are overlapping each other.

In figure 2, the experimental values of absolute viscosity obtained from Namburu et al. (2007) for 29nm CuO particles at a volume fraction of 3% in given range of temperature are compared with the theoretical and empirical models.

Figure 2: Experimental vs. predicted values of absolute viscosity with increase in temperature at 3% volume fraction for 29nm CuO particles

In this figure, the experimental values could not be predicted by any model. All the models are highly under predicting the viscosity values. As the volume fraction has been increased here , the error in predicting the viscosity values as also increased. Wang et al. (1999) gives the highest value of prediction. Predicted values of Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) are overlapping each other. As the temperature increases, the value of experimental and prediction comes close to each other. The difference between experimental and predicted values reduces continuously as the temperature increases.

In figurer 3, the experimental values of absolute viscosity obtained from Namburu et al. (2007) for 29nm CuO at a volume fraction of 6.12%are compared with theoretical and empirical values.

Figure 3: Experimental vs. predicted values of absolute viscosity with increase in temperature at 6.12% volume fraction for 29nm CuO particles

In this figure, the Nguyen et al. (2007) gives the highest value of prediction whereas being close to each other Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) gives least values of prediction. As the temperature increases, predicted values of models come close and overlap near 300K temperature.

In figure 4, the experimental values obtained from Kulkarni et al. (2009) for 30nm CuO particles with EG + Water base fluid at a volume fraction of 6% in given range of temperature are compared with theoretical and empirical models.

Figure 4: Experimental vs. predicted values of absolute viscosity with increase in temperature at 6% volume fraction for 30nm CuO particles

In this figure, neither of the models could predict the experimental viscosity values. All the models are under predicting the values. It is difficult to distinguish Einstein (1906), Brinkman (1952),

Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) as these are overlapping each other. As temperature increases predicted values comes close to experimental values and the difference between predicted and experimental values reduces.

In figure 5, the experimental values obtained from Nguyen et al. (2007) for 29nm CuO particles, with Water base fluid at a volume fraction of 1% in given range of temperature are compared with theoretical and empirical models.

Figure 5: Experimental vs. predicted values of absolute viscosity with increase in temperature at 1% volume fraction for 29nm CuO particles

In this figure, since the volume fraction is low, values of all the models come close to each other. Wang et al. (1999) gives viscosity values very close to the experimental values. In the temperature range of 306K to 311K Wang et al.(1999) slightly over predict the values. Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) are giving values very close to each other.

In figure 6, the experimental values obtained from Nguyen et al. (2007) for 29nm CuO particles, with Water base fluid at a volume fraction of 4.5% in given range of temperature are compared with theoretical and empirical models.

Figure 6: Experimental vs. predicted values of absolute viscosity with increase in temperature at 4.5% volume fraction for 29nm CuO particles

In this figure, none of the model could predict the experimental value of viscosity. Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977) and Saito (1950) are highly under predicting viscosity values. Wang et al. (1999) and Nielson (1970) gives the highest and lowest values of prediction respectively. Even at high temperature there is a great difference between the experimental and predicted values.

In Figure 7, the experimental values obtained from Nguyen et al. (2007) for 29nm CuO particles, with Water base fluid at a volume fraction of 7% in given range of temperature are compared with theoretical and empirical models.

Figure 7: Experimental vs. predicted values of absolute viscosity with increase in temperature at 7% volume fraction for 29nm CuO particles

Here all the models are highly under predicting the experimental values. Wang et al. (1999) and Nielson (1970) gives the highest and lowest values of prediction respectively. As the temperature increases the viscosity values decreases. But still these values are very far away from the actual values.

In figure 8, the experimental values obtained from Nguyen et al. (2007) for 29nm CuO particles, with Water base fluid at a temperature between 22 to 25C in given range of volume fraction are compared with theoretical and empirical models.

Figure 8: Experimental vs. predicted values of absolute viscosity with increase in volume fraction at temperature between 22C to 25C for 29nm CuO particles

From figure 8, it is found that at a very low particle volume fraction ( less than 1%) the values predicted from models comes close to experimental values. As the volume fraction, difference between the predicted and experimental values increases. Over all Wang et al.(1999) gives the highest value of prediction. Einstein (1906), Brinkman (1952), Krieger et al. (1959), Lundgren (1972), Batchelor (1977), Nielson (1970) and Saito (1950) gives values very close to each other.