Biotransport Processes (BMEN30007)

Viscosity Laboratory Report

Caroline Quartermain (1080564)

Claire Esler (995559)
Michael Coghlan (912138)
Vincentia Janice (1067686)
I. Abstract

Viscosity is the resistance of a fluid to flow. It is quantified by determining the

relationship between the shear stress on the fluid causing the flow, and the shear rate.
To find this, the fluids in question were allowed to flow from a triangular reservoir of
known dimensions through a capillary tube, and the height of fluid in the reservoir at one
minute intervals recorded. Expressions for the pressure drop across the capillary tube
and the volumetric flow rate of the fluid through the tube were combined to formulate an
equation relating the height of fluid in the reservoir to time, from which the value of
viscosity could be determined. Plots of the reservoir height against time for each of the
three test fluids yielded two linear graphs and one curved function. This demonstrated
that the viscosities of pure water and the solution of glycerol in water (50 wt% glycerol)
are independent of shear rate and are thus Newtonian fluids, whereas the viscosity of
polyacrylamide in water (0.8g/L polyacrylamide) did vary with shear rate, meaning it was
non-Newtonian and needed to be described by the power law model. From the
gradients of the former two graphs, the absolute viscosities of pure water and the
glycerol-water solution were determined to be 1.024 mPa​·​s and 5.876 mPa​·​s
respectively, and the kinematic velocities were 1.026 mm​²​/s and 5.274mm​²​/s
respectively. Thus the glycerol solution is more viscous than pure water. From the
curved graph, the power-law parameters governing the viscosity of the polyacrylamide
in water were estimated to be n = 0.1339 and k = 0.5011.

1
II. Introduction

Viscosity is an essential property to consider when measuring the flow of liquids and
gasses as without understanding the rheological relationships involved it is impossible
to predict the behavior of fluids. The various scientific and industrial applications of
viscosity data are innumerable, ranging from determining the effects of blood viscosity
on endothelial functioning [1], to determining the effect of acidic fluid viscosity on the
erosion of tooth enamel [2]. Additionally viscosity standards are used by many
manufacturing companies to ensure ideal consistency in food products [3]. And, when
designing any industrial level pump or piping system it is necessary to know the
rheological properties of the fluid being handled as high viscosity fluids require more
power to pump than a low viscosity fluid [4].

This practical will evaluate the experimental viscosity of two Newtonian fluids, water and
a glycerol solution that is 50% water and 50% glycerol. Additionally, this practical aims
to determine the power law parameters of a non-Newtonian fluid, a polyacrylamide
solution that has a concentration of 0.8g of polyacrylamide per litre of water. Different
methods were used to analyse Newtonian and non-Newtonian fluids due to their
fundamentally differing properties. A Newtonian fluid is a fluid with a viscosity that does
not depend on the shear stresses being applied to it. The viscosity of Newtonian fluids
is not necessarily constant as it can depend on other factors such as temperature and
pressure [5] however these experiments were performed under constant conditions so
the viscosity of the Newtonian fluids can be assumed to be constant. Newton’s Law of
Viscosity can be derived by imagining a fluid flowing through a small gap between two
large parallel plates with area, A.

2
 Figure 1​: Velocity profile for a fluid between two plates [5].
The top plate is stationary but the bottom plate is being moved forward by a force, F. If
we define the gap between the plates as Y and the velocity of the bottom plate as V
then the equation for F can be derived from observation alone.

Equation 1​: Relationship between Force and Plate Velocity [5]

Since shear is defined as force per unit area we can replace F/A with . Additionally

V/Y can be replaced with . This results in an equation for Newton's Law of Viscosity:

Equation 2​: Newton’s Law of Viscosity [5]

According to Newton’s Law of Viscosity the shear stress being applied to the Newtonian
fluid is equal to its viscosity times its shear rate. Non-Newtonian fluids cannot be

3
assumed to have constant viscosity as their viscosity is a function of their shear rate.
The rheology of some non-Newtonian fluids can be defined by the power law model:

Equation 3​: Power Law Model (equation 12 in Viscosity Laboratory Manual)

Instead of viscosity it is the values of k and n that must be solved for to determine the
rheological properties of the fluid.

4
III. Materials and Methods

EXPERIMENTAL EQUIPMENT:

The experimental equipment consisted of three triangular reservoirs, which were used
to measure three fluids. Each reservoir had a connected ruler to measure the height of
the fluid, a horizontal capillary tube which attached to the bottom of the reservoir for the
fluid to flow through, and finally a container at the other end of the capillary tube to act
as a fluid collector.

The capillary tube inner diameter, as well as the triangular reservoir depth, D, were
measured using a digital vernier caliper (DVC). The length of the capillary tube as well
as the varying height of water throughout the experiment were measured using a
standard ruler, and the apex angle of the triangular reservoir was measured using a
protractor.

3 fluids were used, each assumed to have a temperature of 20 degrees celsius during
measurement:

1. Pure water, with a density of 998.2kg/m^3

2. Solution of glycerol in water (50%wt glycerol) with a density of ​1114 kg/m^3 (as
calculated in section ​IV.II​)
3. Polyacrylamide in water (0.8g/L of Polyacrylamide), which can be assumed to
have an equal density to water, 998.2kg/m^3

5
SETUP:

Figure 2​: Setup of experiment (adapted from the viscosity laboratory manual)

EXPERIMENTAL PROCEDURE:

A rubber stopper was initially used to block the discharge end of the capillary tube in
order to prevent fluid flow. The fluid of interest was then poured into the triangular
reservoir until near full. The beaker was placed beneath the discharge end of the
capillary tube and the rubber stopper was removed, starting the flow of fluid. Initially,
the fluid expelled at a varying rate, but after approximately 30 seconds, the fluid flow
reached steady-state. At this time, t0, the stopwatch was started and the first fluid

6
height measurement, h1, was recorded using the ruler attached to the reservoir. Height
measurements of the fluid were taken every 60 seconds for approximately 8-10
minutes, or until sufficient data is collected.

It should be noted that for increased accuracy of results, measurements should be

taken at 30 or 60 second intervals, for a longer total period of time. This results in a
smoother height vs time graph, corresponding to a more accurate dh/dt value, and thus
a more accurate value of viscosity.

Repeat the same procedure for the other two fluids. Once completed, refill each
reservoir with its designated fluid.

CALCULATIONS:

For this experiment, students did not complete the lab component individually, but were
instead provided with raw data of the height of each fluid over time. Physical equipment
measurements were also given such as the radius, height, angle and depth of
components. The density of each fluid was also provided.

In MATLAB the raw height vs time data was plotted, the absolute viscosity​, ​μ​, calculated
from the gradient using the equation ​dh/dt = - (​ρ​g​π​R^4)/(8​μ​tan​α​DL), and the kinematic
velocity, v , found from this using v =​μ/ρ​. For the non-Newtonian polyacrylamide
solution, the viscosity cannot be described using the same equation, since the fluid
does not follow a linear Δh vs t relationship. Instead, a variation of the power law model
n
is used, with the equation h(t) = [( 2n−1
n )( c1 t + c2 )]]
2n−1
. Using the Matlab curve

fitting app, and inserting this custom function (as seen in Appendix 2), values for n and k
can be found.

7
IV. Results and Discussion

Measured values:
● R ​= 1.2 mm = 0.0012 m
● L ​= 1.1 m
● D ​= 10 mm = 0.01 m
● α ​= 15​°

Given values:
● ρ ​(water, 20​°C) ​= 998.2 kg/m^3
● ρ ​(glycerol, 20​°C​) = 1261 kg/m^3

8
IV.I Pure Water

From the data provided in table 1 (see appendices, table of values 1), we can plot the
following graph 1.

Graph 1​: Plot of ​the instantaneous height of pure water in the reservoir, h(t) (m),
over time, t (s).

Graph 1 shows that h(t) and t are linearly dependent as h(t) decreases with t. Hence,
water is a Newtonian fluid.

9
Its absolute viscosity value was calculated to be 1.024 ​mPa⋅s, and its kinematic
viscosity was 1.026 mm​2​/s (see appendices, MATLAB script 1).

IV.II Water/Glycerol Solution

From the data provided in table 2 (see appendices, table of values 2), we can then plot
the following graph 2.

Graph 2​: Plot of ​the instantaneous height of water/glycerol solution in the

reservoir, h(t) (m), over time, t (s).

10
Graph 2 shows that h(t) and t are linearly dependent as h(t) decreases with t. Hence,
the water/glycerol solution is a Newtonian fluid.

Evaluating the density of the water/glycerol solution, ​ρ​(w/g),

​ρ​(w/g) = [m(water)+m(glycerol)]/[v(water)+v(glycerol)]
= (2*m)/[(m(water)/​ρ​(water))+(m(glycerol)/​ρ​(glycerol))
= (2*m)/[(m*(​ρ​(water)+​ρ​(glycerol)))/(​ρ​(water)*​ρ​(glycerol))]
= [2*​ρ​(water)*​ρ​(glycerol)]/[​ρ​(water)+​ρ​(glycerol)]
= (2*​998.2*1261)/(998.2+1261)
= 1114 kg/m​3

Using this value of density, along with the measured values, the absolute viscosity was
calculated to be ​5.876 mPa⋅s and the kinematic viscosity was ​5.274 ​mm​2​/s (see
appendices, MATLAB script 1). This means that the water/glycerol solution is more
resistant to flow than pure water.

IV.III Water & Polyacrylamide Solution

From the data provided in table 3 (see appendices, table of values 3), we can then plot
the following graph 3.

11
 Graph 3​: Plot of ​the instantaneous height of polyacrylamide solution in the
reservoir, h(t) (m), over time, t (s).

Graph 3 shows that h(t) and t are non-linearly dependent as h(t) decreases with t.
Hence, the polyacrylamide solution is a non-Newtonian fluid.

To solve for the Power Law parameters n and k, the curve fitting tool in MATLAB was
used.

First, the height (in metres) and time (in seconds) data for the polyacrylamide fluid were
loaded into MATLAB as matrices. We can then generate a function and use these

12
arrays as data points for plotting.

Using the following ​Equation (20) ​from the handout:

n
h(t) = [( 2n−1
n )( c1 t + c2 )]]
2n−1

1
πR3 ρgR
Where c1 = − 1 (
( n +3)tan(α)D 2Lk
) n

2n−1
n(0.654) n
​ c2 = 2n−1

And where:
○ R ​= 1.2 mm = 0.0012 m
○ L ​= 1.1 m
○ D ​= 10 mm = 0.01 m
○ α ​= 15​°
○ ρ ​(water, 20​°C) ​= ​ρ (polyacrylamide solution) = ​998.2 kg/m^3

Once all of the above values are substituted into h(t) for each variable, we get the
following function which can then be loaded into MATLAB (see appendices, MATLAB
script 2).

h(t)=(((2*n-1)/n)*(((-(0.00000000542)/(((1/n)+3)*0.002679))*((11.7508/(2.2*k))^(1/
n)))*t + (((n*(0.654^((2*n - 1)/n)))/(2*n - 1)))))^(n/(2*n - 1))

After inputting the above function, h(t), into the ‘custom function’ section, a graph 4 is
outputted:

13
 Graph 4​: Curve fitting plot of h(t) over t for polyacrylamide solution.

Values for the Power Law parameters n & k that fit the function are given as follows:
- n = 0.1339 with 95% confidence bounds of (0.498, 0.5042),
- k = 0.5011, with 95% confidence bounds of (0.1317, 0.1362)

These values are considered to be quite accurate, as an R squared value of 1 was

found, as well as the tight confidence intervals around each value as seen above.

Since the Power Law parameter n = 0.1339 ＜ 1, polyacrylamide is a shear-thinning

fluid. The viscosity of this fluid decreases as shear stress increases, so will flow more
freely under greater stresses. This is why the height vs time graph is not linear, but
decreases at a faster rate initially when there is a greater stress applied due to gravity,
and then slows down as the stress is relieved.

14
IV.V Error Analysis
The measured values were obtained using either the vernier calipers or a ruler.

Since the calipers were digital and gave a reading to two decimal places, it was
reasoned that the measurement uncertainty was ±0.01mm.

Measurements using a ruler were done by eye, therefore measurements would have
been rounded to the nearest millimetre marker making the uncertainty ±0.5mm.

The maximum error for the absolute and kinematic viscosities of each Newtonian fluid
was calculated using MATLAB (see appendices, MATLAB script 1).

For water these were found to be ±0.0942 mPa⋅s and ±0.0944 mm​2​/s respectively.
For the glycerol solution they were ±1.022 mPa⋅s and ±1.024 mm​2​/s respectively.

IV.V Effect of Reservoir Shape

A triangular reservoir was used for this experiment as it has a cross sectional
area that is dependent on h(t).

(equation 7 from Lab Manual)

This results in an equation for volumetric flow rate that is dependent on h(t) as
well:

15
 (equation 6 and 8 from Lab Manual)

The volumetric flow rate, Q, can also be given by this equation:

(equation 4 from Lab Manual)

And pressure is also a function of h(t) so there are two equations for Q that are
functions of h(t). By letting them equal each other h(t) can be cancelled out.
Resulting in dh/dt being equal to a constant and hence h(t) being a linear
equation.

(equations 9-11 from Lab Manual)

16
An alternative apparatus for the measurement of the viscosity of the Newtonian
fluids could employ a rectangular rather than a triangular-shaped reservoir. A
rectangular reservoir would have a cross sectional area of B*W. Hence for a
rectangular reservoir, the discharge rate, Q ( ​m3​​ /s) is:

This will cause dh/dt to not be constant throughout the discharge through the
capillary tube. As seen in step 6 of the following calculations dh/dt is actually a
function of h(t).

This is due to the cross sectional area of the reservoir not being a function of h(t).
Hence in step 5 there is no cancellation of h(t) as there was for a triangular
reservoir.

17
The final expression for the height of the fluid in the reservoir over time is a
negative exponential. This means that if a rectangular reservoir were used, h(t)
would not decrease at a constant rate (for constant viscosity) and it is for this
reason that a triangular reservoir was used for this experiment, as it allows for a
constant rate of decrease for h(t). This provides a linear expression for h(t),
which allows for simpler calculations to find the rheological properties of the
Newtonian fluids.

IV.VI Areas of Improvement

In order to improve the accuracy of the results the experiment should be

repeated in order to minimize the effects of human error.

Additionally, more precise results could be achieved if a more accurate value for
the density of the polyacrylamide solution was obtained instead of assuming it is
the same as water. This is because pressure, which is used in the
Hagen-Poiseuille equation, is dependent on density.

18
V. Conclusions

Water and the water/glycerol solution are proven to be Newtonian fluids as the plot of
h(t) over t were linear for both. From the plot, we derived that the absolute viscosity of
water, ​μ​(water), is 1.024 mPa⋅s whereas the absolute viscosity of the water/glycerol
solution, ​μ​(w/g), is 5.876 mPa⋅s. As such, we then calculated the kinematic viscosity for
water and the water/glycerol solution to be 1.026 mm^2/s and 5.274 mm^2/s,
respectively. Thus, the water/glycerol solution is more resistant to flow than pure water.

On the other hand, the polyacrylamide solution is proven to be a Non-Newtonian fluid as

the plot of h(t) over t was non-linear. The value of the power law parameters n & k that
were derived using curve fitting is ​0.1339 and 0.5011​, respectively. Since the value for n
is less than 1 we can determine that the polyacrylamide solution is shear thinning.

19
VI. References

1. Melkumyants A, Balashov S, Khayutin V. Endothelium dependent control

of arterial diameter by blood viscosity. Cardiovascular Research.
1989;23(9):741-747.
2. Aykut-Yetkiner A, Wiegand A, Bollhalder A, Becker K, Attin T. Effect of
Acidic Solution Viscosity on Enamel Erosion. Journal of Dental Research.
2013;92(3):289-294.
3. Mann L, Wong K. Development of an Objective Method for Assessing
Viscosity of Formulated Foods and Beverages for the Dysphagic Diet.
Journal of the American Dietetic Association. 1996;96(6):585-588.
4. Şahin A. The effect of variable viscosity on the entropy generation and
pumping power in a laminar fluid flow through a duct subjected to constant
heat flux. Heat and Mass Transfer. 1999;35(6):499-506.
5. Bird, R.B., Stewart, W.E., and Lightfoot, E.N. (2002), Transport
Phenomena, 2nd Edition, John Wiley & Son, New York, pp. 11-21, 67-68,
261.

20
VII. Appendices

1. Table of values

Table 1​: Measurements of the instantaneous height of pure water in the reservoir
at time t, h(t), in cm as well as the respective values converted to m over a period
of time 0 ​≤ t ≤ 240 s.

Table 2​: Measurements of the instantaneous height of water/glycerol solution in

the reservoir at time t, h(t), in cm as well as the respective values converted to m
over a period of time 0 ​≤ t ≤ 600 s.

21
Table 3​: Measurements of the instantaneous height of Polyacrylamide solution in
the reservoir at time t, h(t), in cm as well as the respective values converted to m
over a period of time 0 ​≤ t ≤ 1800 s.

22
23
2. MATLAB scripts

MATLAB script for graphs, viscosity calculations and error calculations:

%IMPORTING DATA

%tables saved as .csv files that are imported

dir=uigetdir;
n1=uigetfile('*.csv','polyacrylamide');
n2=uigetfile('*.csv','glycerol');
n3=uigetfile('*.csv','water');
f1=fullfile(dir,n1);
f2=fullfile(dir,n2);
f3=fullfile(dir,n3);

%data as matrices
poly=csvread(f1,1,0);
glycerol=csvread(f2,1,0);
water=csvread(f3,1,0);

%time (s)
tp=poly(:,1);
tg=glycerol(:,1);
tw=water(:,1);

%height (m)
hp=poly(:,2);
hg=glycerol(:,2);
hw=water(:,2);

%constant values (in SI units)

R=1.2/10^3;
L=1.1;
D=10/10^3;
alpha=15; %degrees
g=9.81;
pw=998.2;
pg=1114; %the calculated value
pp=pw;

%--------------------------------------------------------------------
%PLOTTING GRAPHS
figure(1)
plot(tw,hw)
title('Pure water')
xlabel('time (seconds)')
ylabel('fluid height in reservoir (metres)')

figure(2)
plot(tg,hg)
title('Glycerol solution')
xlabel('time (seconds)')
ylabel('fluid height in reservoir (metres)')
figure(3)
plot(tp,hp)
title('Polyacrylamide solution')
xlabel('time (seconds)')
ylabel('fluid height in reservoir (metres)')

%--------------------------------------------------------------------
%CALCULATING ABSOLUTE AND KINEMATIC VISCOSITIES FROM DATA (question 2)

C=-pi*g*R^4/(8*tand(alpha)*D*L);

%water
W=polyfit(tw,hw,1);
wgrad=W(1);
uw=C*pw/wgrad;
vw=uw*10^6/pw;

%plot line of best fit

LW=length(tw);
bestfithw=zeros(1,LW);
for i=1:LW
bestfithw(i)=wgrad*tw(i)+W(2);
end
figure(1)
hold on
plot(tw,bestfithw,'r--')

%glycerol
G=polyfit(tg,hg,1);
ggrad=G(1);
ug=C*pg/ggrad;
vg=ug*10^6/pg;

%plot line of best fit

LG=length(tg);
bestfithg=zeros(1,LG);
for i=1:LG
bestfithg(i)=ggrad*tg(i)+G(2);
end
figure(2)
hold on
plot(tg,bestfithg,'r--')

%--------------------------------------------------------------------
%ERROR ANALYSES ON NEWTONIAN FLUID EXPERIMENTS

%setting +- error values:

%calipers: digital, and measures to 2 decimal places, therefore error

is +/-0.01mm
%ruler: +/-0.5mm because reading by eye it is rounded to the nearest
millimetre marker, so biggest deviation would be 0.5mm

Re=0.01/10^3/2; %R error (/2 because diameter measured)

he=0.5/10^3; %height error
ae=1; %angle error
De=1/100/10^3; %D error
Le=0.5/10^3; %L error

%max gradient will be when LHS points are lowest and RHS points
highest
%pure water
hw1=hw; %higher gradient line (pre-allocating)
hw2=hw; %lower gradient line
for i=1:LW
if i<LW/2
hw1(i)=hw(i)-he;
hw2(i)=hw(i)+he;
elseif i>LW/2
hw1(i)=hw(i)+he;
hw2(i)=hw(i)-he;
end
end
%glycerol solution
hg1=hg;
hg2=hg;
for i=1:LG
if i<LG/2
hg1(i)=hg(i)-he;
hg2(i)=hg(i)+he;
elseif i>LG/2
hg1(i)=hg(i)+he;
hg2(i)=hg(i)-he;
end
end

%find these bigger and smaller possible gradients

W1=polyfit(tw,hw1,1);
W2=polyfit(tw,hw2,1);
wgrad1=W1(1);
wgrad2=W2(1);
G1=polyfit(tg,hg1,1);
G2=polyfit(tg,hg2,1);
ggrad1=G1(1);
ggrad2=G2(1);

%pure water absolute viscosity

%biggest
uw1=-pi*pw*g*(R+Re)^4/(8*wgrad2*tand(alpha-ae)*(D-De)*(L-Le));
%smallest
uw2=-pi*pw*g*(R-Re)^4/(8*wgrad1*tand(alpha+ae)*(D+De)*(L+Le));
 %glycerol solution absolute viscosity
%biggest
ug1=-pi*pw*g*(R+Re)^4/(8*ggrad2*tand(alpha-ae)*(D-De)*(L-Le));
%smallest
ug2=-pi*pw*g*(R-Re)^4/(8*ggrad1*tand(alpha+ae)*(D+De)*(L+Le));

%calculating the max error (in kg/m/s)

%water
if (uw1-uw)>(uw-uw2)
werror=uw1-uw;
else
werror=uw-uw2;
end
%glycerol solution
if (ug1-ug)>(ug-ug2)
gerror=ug1-ug;
else
gerror=ug-ug2;
end

%max error for dynamic viscosities (in mm^2/s)

vwerror=werror/pw*10^6;
vgerror=gerror/pw*10^6;
%--------------------------------------------------------------------
%PRINTING ANSWERS
fprintf('\nThe absolute viscosity of water is %f kg/m/s, which is %f
mPa.s\n',uw,uw*1000)
fprintf('Max error +/- %f mPa.s\n\n',werror*1000)
fprintf('The kinematic viscosity of water is %f mm^2/s\n',vw)
fprintf('Max error +/- %f mm^2/s\n\n',vwerror)

fprintf('The absolute viscosity of glycerol is %f kg/m/s, which is %f

mPa.s\n',ug,ug*1000)
fprintf('Max error +/- %f mPa.s\n\n',gerror*1000)
fprintf('The kinematic viscosity of glycerol is %f mm^2/s\n',vg)
fprintf('Max error +/- %f mm^2/s\n\n',vgerror)

Command window output:

The absolute viscosity of water is 0.001024 kg/m/s, which is 1.023683 mPa.s

Max error +/- 0.094184 mPa.s

The kinematic viscosity of water is 1.025529 mm^2/s

Max error +/- 0.094354 mm^2/s

The absolute viscosity of glycerol is 0.005876 kg/m/s, which is 5.875640 mPa.s

Max error +/- 1.021908 mPa.s

The kinematic viscosity of glycerol is 5.274363 mm^2/s

Max error +/- 1.023751 mm^2/s
MATLAB code for finding k and n for power law fluid:

% Set symbolic variables n and k

syms n k

% Load height data as a matrix

h=[0.654
0.628
0.604
0.582
0.56
0.54
0.52
0.501
0.482
0.465
0.447
0.43
0.415
0.399
0.384
0.369
0.356
0.343
0.329
0.316
0.304
0.293
0.282
0.271
0.261
0.251
0.242
0.233
0.225
0.216
0.209];

% Load time data as a matrix

t = [0
60
120
180
240
300
360
420
480
540
600
660
720
 780
840
900
960
1020
1080
1140
1200
1260
1320
1380
1440
1500
1560
1620
1680
1740
1800];

% C1 definition
c1 = ((-(0.00000000542)/(((1/n) + 3)*0.002679)) *
((11.7508/(2.2*k))^(1/n)));

% C2 definition
c2 = (((n*(0.654^((2*n - 1)/n)))/(2*n - 1)));

% Function of h(t), with independent variable t, and unknown

constants, n and k.
h= (((2*n-1)/n)*(c1*t + c2))^(n/(2*n - 1));

Curve fitting output:

General model:
f(t) = (((2*n-1)/n)*(((-(0.00000000542)/(((1/n) + 3)*0.002679)) * ((11.7508/(2.2*k)
)^(1/n)))*t + (((n*(0.654^((2*n - 1)/n)))/(2*n - 1)
))))^(n/(2*n - 1))
Coefficients (with 95% confidence bounds):
k= 0.1339 (0.1317, 0.1362)
n= 0.5011 (0.498, 0.5042)

Goodness of fit:
SSE: 2.091e-05
R-square: 1
Adjusted R-square: 1
RMSE: 0.0008492