• No results found

Measurement Uncertainty

5 Results and Discussion

6 Uncertainty Analysis of Physical Properties

6.1 Measurement Uncertainty

The outcome of a measured property depends on the measuring system, the measurement procedure, the skill of the operator, the environment and other effects [207]. As a result, no measurement is exact and measurement uncertainty is stated as a statistical parameter that discusses the possible fluctuations of a measured property [208]. An average value of the dispersion of a number of indication values carries the information of an estimation of the true quantity value and that would not generally be adequate to consider as a final value for a measured property.

Measurement error refers to the deviation of measured quantity from its true value or reference. There are two types of errors involved in measurements known as systematic and random errors. Systematic error is a quantifiable measurement error (offset), which remains constant in replicate measurements. The random error varies in an unpredictable manner in replicate measurements.

The corrections that are made for the measurement errors are often performed prior to the uncertainty analysis as described according to the Guide to the Expression of Uncertainty in Measurements (GUM)[209, 210]. The GUM method provides a different approach rather than express the results of measurement by giving the best estimate of the measurand along with the information about random and systematic errors. The GUM approach enables us to express the results of measurement as the best estimate of measurand together with associated measurement uncertainty.

6.1.1 Uncertainty of Viscosity Measurements

The uncertainty of viscosity measurements is raised due to the uncertainty sources involved in the measuring method. Several uncertainty sources were identified for both aqueous and CO2 loaded solutions. Figure 6.1 illustrates the considered uncertainty sources for the viscosity measurement of aqueous amine solutions.

The mathematical model for a double-gap viscometer as shown in Eq (147) provides the relevant uncertainty sources for the uncertainty of viscosity measurements. There are additional uncertainty sources involved in viscosity measurement that are not shown in the model. Such sources are added to the model equation as shown in Eq (148). The uncertainty analysis was done by following the guidelines provided by GUM and QUAM (Quantifying Uncertainty in Analytical Measurement)[210].

___

82

Figure 6.1: Cause and effect diagram for uncertainty analysis of viscosity measurement [211].

πœ‚ = 𝑇

4πœ‹πΏπœ”π‘…2( π‘˜1 2 π‘˜12βˆ’1+π‘˜22π‘˜3

2 π‘˜32βˆ’π‘˜22)

(147)

Here, T is torque, πœ‚ is dynamic viscosity, L is the liquid height, R is the radius of the inner fixed cylinder, Ο‰ is angular velocity, 𝑅1= 𝐾1𝑅 , 𝑅2= 𝐾2𝑅 and 𝑅3= 𝐾3𝑅.

πœ‚ = 𝑇

4πœ‹πΏπœ”π‘…2( π‘˜12 π‘˜12βˆ’1+π‘˜22π‘˜3

2 π‘˜32βˆ’π‘˜22)

π‘“π‘π‘“π‘‘π‘“π‘€π‘“π‘Ÿπ‘’π‘ (148)

Where 𝑓𝑝 is purity of MEA, 𝑓𝑑 is temperature, 𝑓𝑀is weight measurement and π‘“π‘Ÿπ‘’π‘ is repeatability. Those factors are added to the original viscosity expression to consider uncertainty sources, which are not shown in Eq (147).

In the GUM method, the propagation of uncertainty based on the first-order Taylor series approximation is considered in the uncertainty evaluation. A Gaussian distribution is assumed as the probability distribution and that is also used to define confidence intervals.

Consider a measuring system as described in Eq (149).

𝑦 = 𝑓(π‘₯1, π‘₯2, … , π‘₯𝑁 ) (149)

Where 𝑦 is the measurand and π‘₯1, π‘₯2, … π‘₯𝑁 are the input quantities. The propagation of uncertainty according to the Taylor series expansion of 𝑦,

𝑒2(𝑦) = βˆ‘ (πœ•π‘“

πœ•π‘₯𝑖)2𝑒2(π‘₯𝑖) + 2

𝑁𝑖=1 βˆ‘ βˆ‘ πœ•π‘“

πœ•π‘₯𝑖

πœ•π‘“

πœ•π‘₯𝑗𝑒(π‘₯𝑖,π‘₯𝑗)

𝑁𝑗=𝑖+1

π‘βˆ’1𝑖=1 (150)

___

83 In Eq (150), (πœ•π‘“ πœ•π‘₯⁄ 𝑖) gives the partial derivatives (sensitivity coefficients) [212, 213], 𝑒2(𝑦) is the variance of the measuring result, the variance of the input quantity π‘₯𝑖 is given by 𝑒2(π‘₯𝑖) and the covariance between π‘₯𝑖 and π‘₯𝑗 is given by 𝑒(π‘₯𝑖,π‘₯𝑗) [213].

π‘ˆ(𝑦) = π‘˜π‘’(𝑦) (151)

In general, calculated combined uncertainty from Eq (151) is defined as the standard uncertainty with the coverage factor π‘˜=1 which covers 68% of the interval. It is common to represent uncertainty as an expanded uncertainty with π‘˜ =1.96 or 2, which covers approximately 95% of the interval.

Table 6.1 lists the considered uncertainty sources with corresponding distributions and values. The calculated uncertainties for the viscosity of different MEA solutions are given in Table 6.2. The attached Article N and Article O provide more information about the analysis.

Table 6.1: Uncertainty sources with corresponding distributions and values[202].

Input quantity Xi Probability

Distribution Uncertainty U(xi)

Torque (𝜏) Triangular 0.082 πœ‡π‘π‘š

Level (𝐿) Gaussian 0.45 π‘šπ‘š

Angular velocity (πœ”) Triangular 0.01 π‘Ÿπ‘Žπ‘‘ Β· π‘ βˆ’1

Radius (𝑅) Triangular 4.1 πœ‡π‘š

Purity Rectangular 2.886x10-3

Temperature Triangular 2.45 x10-4

Weight measurement Rectangular 8 x10-6

CO2 loading Gaussian 0.013

Repeatability Gaussian 0.00348

The uncertainty sources in the viscosity measurement was calculated as follows.

Torque (𝜏):

The Anton Paar user manual [214] provide the accuracy of torque measurement as, Torque accuracy: max. (0.2 ΞΌNm; 0.05%)

Then a triangular distribution was considered to calculate the uncertainty of torque.

Standard uncertainty of torque (Ο„) measurement 𝑒(𝜏) =0.2

√6 = 0.082

ΞΌNm

___

84

Angular Velocity (πœ”):

All the viscosity measurements were performed under a shear rate of 1000 1/s. for this analysis Β± 1 1/s of accuracy was assumed to determine the uncertainty of angular velocity. Using Eq (152), Eq (153) and triangular distribution, the standard uncertainty of angular velocity (Ο‰) was determined as,

𝛾̇ =πœ‹π‘›

30.1+𝛿2

𝛿2βˆ’1 (152)

πœ” = 2πœ‹

60𝑛 (153)

𝑒(πœ”) = 0.01radβΈ³s-1 Level (𝐿):

The liquid was transferred using a 10 mL syringe with an accuracy of Β±0.1 mL. Using a triangular distribution, the uncertainty of volume measurement was determined as Β±0.4 mL. Then using the geometry of the double-gap rheometer, the standard uncertainty of level was calculated as 0.45 mm.

Radius (𝑅):

For the analysis, Β± 0.01 mm of accuracy was assumed to determine the uncertainty of the radius of the cylindrical geometries. Assuming a triangular distribution for the uncertainty, Β± 4.1 ΞΌm was obtained for the standard uncertainty of radius.

Purity (𝑃):

The purity of the monoethnaol amine is 99.5%. Considering a rectangular distribution for the standard uncertainty 𝑒(𝑓𝑝) was calculated as,

𝑒(𝑝) =0.005

√3 = 0.0029 = 𝑒(𝑓𝑝)

The 𝑒(𝑓𝑝) vary with the material; here considered only monoethano amine for the aqueous monoethanol amine mixtures.

Temperature (𝑇):

The temperature accuracy of the instrument is Β± 0.03 K. A triangular distribution and an average temperature of the measuring range (293.15 K – 393.15 K) were assumed to calculate the uncertainty of temperature 𝑒(𝑓𝑇).

Standard uncertainty of temperature, 𝑒(𝑇) =0.03

√6 = 0.012 𝐾

___

85 𝑒(𝑓𝑇) =0.01225

50 = 0.000245 Weight measurement (𝑀):

The linearity of the balance is considered as Β± 0.2 mg. A rectangular distribution was assumed for the uncertainty of weight measurement. Accordingly, for two weight measurements for a binary mixture:

The standard uncertainty of weight measurement, 𝑒(𝑀) = 2. (0.2

√3)2

=

0.16mg Considering a sample volume of 20g

𝑒(𝑓𝑀) =0.16 Γ— 10βˆ’3

20 = 0.000008 Repeatability

Standard deviation of 0.0076 mPaβΈ³s was observed for the measured viscosity of 2.179 mPaβΈ³s. Accordingly,

𝑒(π‘“π‘Ÿπ‘’π‘) =0.0076

2.179 = 0.0035

Table 6.2: Measurement uncertainty of viscosity for different solutions

Property Solution Uncertainty

(95% confidence at k=2) Viscosity

Pure Β± 0.0162 mPaΒ·s

Aqueous MEA Β± 0.0162 mPaΒ·s

CO2 loaded aqueous MEA Β± 0.0353 mPaΒ·s

6.1.2 Uncertainty of Density Measurements

A similar approach was adopted as explained in section 6.1.1 to evaluate uncertainty in density measurement. The uncertainty sources of material purity, weight measurement, temperature, CO2 loading and repeatability were considered during the uncertainty analysis. The influence of uncertainty in temperature on density measurement was determined by calculating the gradient πœ•πœŒ πœ•π‘‡β„ of density against temperature. A similar approach was used to calculate the influence of uncertainty in CO2 loading on density measurement in which gradient of density against CO2 loading, πœ•πœŒ πœ•π›Όβ„ , was found from measured densities at different CO2 loadings. Finally, (πœ•πœŒ πœ•π‘‡β„ )⸳𝑒(𝑇) and (πœ•πœŒ πœ•π›Όβ„ )⸳𝑒(𝛼) were determined to find standard uncertainties of density measurement due to the temperature and CO2 loading. For CO2 loaded aqueous monoethanol amine mixtures,

___

86

the gradient πœ•πœŒ πœ•π‘‡β„ of density against temperature was found as 0.73 kgβΈ³m-3βΈ³K-1 and the corresponding uncertainty in 𝜌 that is (πœ•πœŒ πœ•π‘‡β„ )⸳𝑒(𝑇) was calculated as Β±0.009 kgβΈ³m-3. The gradient of density against CO2 loading, πœ•πœŒ πœ•π›Όβ„ , was found as 334 kgβΈ³m-3 and the corresponding uncertainty in 𝜌, (πœ•πœŒ πœ•π›Όβ„ )⸳𝑒(𝛼) was calculated as Β±1.67 kgβΈ³m-3. The standard combined uncertainties for density measurement 𝑒(𝜌) for different monoethanol amine mixtures are listed in Table 6.3.

Table 6.3: Measurement uncertainty of density for different solutions

Property Solution Uncertainty

(95% confidence at k=2) Density

Pure Β± 7.1 kgβΈ³m-3

Aqueous MEA Β± 7.1 kgβΈ³m-3

CO2 loaded aqueous MEA Β± 7.8 kgβΈ³m-3