CC BY
0Reliably measuring the condition of mineral-based transfer fluids using a permittivity sensor - practical application to thermal fluid heat transfer
Christopher Ian Wrighta,n, Thomas Bembridge b
CrossMark
a Global Group, Cold Meece Estate, Cold Meece, Staffordshire ST15 0SP, UK b Kinetic Partners, 1 London Wall, London EC2Y 5HB, UK
ARTICLE INFO
Article history:
Received 16 June 2015
Received in revised form
6 September 2015
Accepted 18 September 2015
Available online 21 September 2015
Keywords:
Heat transfer fluid
Thermal fluid heat transfer
Heat transfer fluid condition
Permittivity
Cleanliness
ABSTRACT
This article describes a series of experiments to assess the performance and suitability of a permittivity sensor in the area of heat transfer. The permittivity sensor measures condition index and temperature of a fluid. A series of 5 experiments was conducted. They assessed the reproducibility of the sensor using both clean and dirty fluid samples, and showed the sensor had good reproducibility based on calculations of coefficients of variation. The sensor also detected water contamination, assessed from construction of a stimulus-response curve to step-wise increases in water and from real-life samples where water content was reported to be out of specification. Further experiments tested the association between condition index and both water content and fluid cleanliness in a real-life setting. Results demonstrated the sensor that condition index reflected changes in fluid water and cleanliness and was therefore a measure of fluid condition. The implication of these findings is that the sensor can be used to make rapid and reliable assessments of fluid condition using only small samples (i.e., < 50 ml). The sensor may be of benefit to customers that need to make a lot of regular samples over a large processing site, such as concentrated solar power plants.
© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).
1. Introduction
Understanding and predicting the thermal degradation of heat transfer fluid (HTF) is critical to any planned preventative maintenance program. The process of thermal is a complicated process [1] involving a number of factors including elevations in temperature above the HTFs bulk operating temperature, oxidative stress and contamination within the system (i.e., wear particles) and the build-up of foreign particles (i.e., water). The chemical composition of a HTF can be routinely conducted in a laboratory [2,3] and should be conducted according to International standards [4,5]. Water is routinely measured along with other functional parameters of HTF condition. These include carbon residue, acids, flash components and changes in kinematic viscosity. Manufacturers and insurers suggest that HTFs are sampled and analysed at least once per year when a HTF is operating close to its bulk temperature and biannually if it is more than 20 °C below its bulk temperature [6,7]. This allows the condition of a HTF to be monitored over time which can be communicated to the
Abbreviations: ; HTF, heat transfer fluid; ASTM, American society for testing and materials; CV, coefficient of variation * Corresponding author. E-mail address: chrisw@globalgroup.org (C.I. Wright).
http://dx.doi.org/10.1016/j.csite.2015.09.008
2214-157X/© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
manufacturer or insurer. Likewise, it is also important that HTF condition is monitored prior to any intervention, for example, prior to a system fill a HTF sample is taken to determine if water is present [8]. Indeed, water contamination can occur during storage and it is important that HTF drums are not stored outside as water will inevitably get into the drum and this is a potential entry point of water into a HTF system and a source of potential problems [9]. Moreover, contaminants work to accelerate the ageing of a HTF and can potentially shorten the working life of a HTF [1], so it is important to be able to accurately and reliably assess their appearance and also remove them to maintain the long-term efficiency and safety of the plant.
The presence of water can be detected visually as water and HTFs are immiscible and so water is seen as a clear and separate layer at the bottom of a HTF sample. Testing should be complimented by a standardised, quantitative test to provide accurate assessments and to draw sound scientific conclusions. Chemical testing routinely involves assessments performed in the laboratory, but this approach can be labour intensive if a wide range of parameters is being assessed. This can be frustrating to all involved, especially when water has been visually confirmed. Hence there is a need for fast, reliable and accurate assessments to be conducted onsite and once the HTF sample has been drawn. One possible approach might be to assess HTF condition using a permittivity sensor which may be a valuable addition to HTF condition based maintenance [10].
Permittivity sensors offer the potential to provide a measure of HTF condition and temperature. Our study was designed to test the reliability of measures performed with the Global ThermocareTM Sensor (GTS) which is a permittivity sensor. This sensor can be used to make assessments of a fluid's condition based on permittivity. The current study assessed the functionality of this sensor. This was done by assessing: (a) the reproducibility of measurements from assessments of intra-and inter-group coefficients of variation of both virgin and used HTF; (b) comparison of water contamination recorded using standard laboratory tests and the permittivity sensor; (c) the response of the sensor to water which was assessed in the laboratory by constructing a stimulus-response curve to increasing water content; and, (d) assessing the functionality of the sensor in a real-life setting during the flushing of a newly built HTF system [11]. The results of this study are presented herein and discussed in the context of current and possible future uses.
2. Experimental methods
2.1. Global Thermocare™ Sensor (GTS)
GTS is a dielectric sensor used to measure the capacitance of a HTF, which changes as the oil becomes damaged. The GTS works in a similar manner to a traditional dielectric sensor. It passes an alternating current electric across two electrodes. The change in current flow changes with the state of degradation. The GTS works in a similar fashion but measures permittivity, which is a combined measure of capacitance and conductance. Capacitance is the ability to store an electrical charge. Based on a parallel-plate capacitor, capacitance (C) = er»e0»A/d where A is the area of overlap of the two plates; er is the relative static permittivity or the dielectric constant of the material between the plates; e0 is the electric constant; and, d is the separation between the plates. Conductance is the ease with which an electric current passes through a conductor and is defined as current divided by voltage or 1 / resistance.
Permittivity is a reflection of a material's ability to transmit or 'permit' an electric field and is affected by humidity, temperature and other parameters.
Fig. 1. Screenshot of customised software used with the Global Thermocare™ Sensor.
Table 1
Typical properties for a virgin mineral-based HTF.
Parameter
Property
Examples of mineral-based HTFs
Appearance
Carbon residue, % weight
Strong and total acid number, KOH per gram of sample Water content, ppm Ferrous wear, ppm
Elements (e.g., iron, silicon) < 5 ^m, ppm Operating temperature range, °C Pour point, °C Auto-ignition point, °C Boiling point at 1013 mbar, °C Maximum film temperature, °C Fire point temperature range, °C Open flash point temperature, °C Closed flash point temperature, °C Kinematic viscosity at 40 °C, mm2/s
BP Transcal N #Globaltherm™ M Shell HT S2
Clear-yellow liquid with a mild odour 0
< 0.05
< 100 < 10 0
-10 to 320
Note: #, values are presented for Globaltherm™ M product datasheet (see http://www.globalheattransfer.co.uk/heat-transfer-fluids/high-temperature-thermal-fluid).
2.2. GTS software
The GTS is plugged into a personal computer or laptop computer via a USB port and data is recorded using customised software. A screenshot of the software used is shown in Fig. 1. Using this platform it is possible to measure and record the date of the recording, the type of fluid being analysed and the condition of the fluid and the temperature of the fluid are recorded. Data is then stored as a.csv file and can be opened in Microsoft Office Excel 2007.
The GTS measures HTF quality on an arbitrary scale ranging from - 2 to + 21. Each sensor is calibrated specifically to the HTF being used and that means that each HTF has the same range (i.e., - 2 to + 21). On this scale the value of - 2 reflects the lower limit and reported when the sensor is exposed to air. The upper limit is + 21 and this indicates that the HTF is contaminated or severely degraded. In the current study this index is referred to as the 'Condition Index.' The GTS is used to make concurrent measurements of a HTFs condition (i.e., inferred from the condition index) and temperature.
2.3. Experiments conducted with GTS
A number of experiments were conducted and are summarised below.
2.3.1. Experiment 1 - virgin sample, reproducibility
The GTS was inserted into a 50 ml sample of virgin mineral based HTF (i.e., Globaltherm™ M [(please see http://www. globalheattransfer.co.uk/heat-transfer-fluids/high-temperature-thermal-fluid).] and typical values are presented in Table 1) and 10 back-to-back to measurements were conducted with the sensor. Each measurement took less than 10 s to perform. The data can then be used to calculate within-group reproducibility assessed from calculations of within-group coefficients of variation (CV). This procedure was repeated on 6 separate samples and this allows between-group CV to be calculated. The sensor was exposed to a 50 ml sample of tap water (the upper limit; n = 1) and to air (the lowest limit). This was conducted 21-times-prior inserting the sensor into the 50 ml sample of tap water; prior to each of the measures of virgin mineral-based HTF; and, prior to and at the end of the measurements used to construct the stimulus-response curve to tap water.
Intra- and inter-group CV were calculated for HTF condition index and HTF temperature.
2.3.2. Experiment 2 - used HTF sample, intra-group reproducibility
The reproducibility of the sensor in real-life was assessed in 3 separate samples. This was based on measurements conducted on three separate HTF systems that had been sampled recently. In this scenario, a 500 ml sample is taken from the HTF system (this has been explained previously in Refs. [2,3,13]) and a 50 ml sample is extracted for subsequent testing in the laboratory. In-keeping with experiment 1, the sensor was used to assess 10 back-to-back to measurements and within-group CV was calculated to assess the reproducibility of the sensor using real-life samples.
2.3.3. Experiment 3 - water content of used HTF measured using laboratory test ASTM D6304 and the sensor
Two 500 ml HTF samples were taken from the HTF systems of three HTF systems due for routine sampling. This enabled
the comparison of water measured in the laboratory using test ASTM D6304 and to compare this with values measured using the GTS.
The sampling technique has been reported previously [2,3,13]. Once cooled, the following test parameters are performed according to international standards (i.e., ASTM International and International Petroleum). Tests include: appearance (coded according to colour); carbon residue (IP14); total acid number (IP139); strong acid number (IP177); closed flash point (ASTM D93); open flash point (ASTM D92); fire point (ASTM D92); kinematic viscosity (IP71); ferrous wear debris (PQ Analex Method); and, elements (ASTM D5185) [13].
2.3.4. Experiment 4 - the response of the sensor to step-wise increases in tap water only
The response of the sensor was assessed in the laboratory by constructing a stimulus-response curve to increasing water content of a virgin mineral-based HTF (Globaltherm™ M). Tap water was added in a step-wise fashion at an ambient temperature of 25.0 + 0.7 °C and in volumes to achieve the following percentage water composition (parts per million; ppm): 0.125 (1250), 0.25 (2500), 0.5 (5000), 0.625 (6250), 0.75 (7500), 0.875 (8750), 1.0 (10,000), 1.125 (11,250), 1.25, (12,500) 1.5 (15,000) and 1.75% (17,500 ppm) tap water. Once tap water had been added, the sample was shaken to ensure the water was mixed with the HTF. This dilution range is expressed as a percentage and achieved using a volume-to-volume dilution. Measurements with the sensor were performed prior to dilution, at each dilution and after dilution. At each dilution, 10-repeated measurements were performed and HTF condition index and temperature recorded. These recordings with the sensor were bracketed by a measurement conducted whilst the sensor was exposed to air. This was done to ensure that no water remained on the GTS and reduce the potential for erroneous recordings.
2.3.5. Experiment 5 - the real-life functionality of the sensor assessed during the flushing of a newly built plant
Real-life measurements were performed during a recent commissioned job to flush a newly built heat transfer plant in Scandinavia designed to hold 100 metric tons of a synthetic HTF. Prior to filling the system, it was flushed with a flushing and cleaning fluid to remove environmental (e.g., water) and build contaminants (e.g., welding slag) [11]. Flushing was combined with HTF filtration whereby particles greater than 15 ^m in diameter were filtered from the HTF.
In the current case, 19 samples were extracted from the HTF system during flushing and cleaning. These samples were then analysed to assess: condition index using the GTS; the water content according to test ASTM D6304; and, ISO cleanliness according to ISO 4406:1999 [3]. ISO cleanliness quantifies particulate distribution (per millilitre) based on three sizes-4, 6 and 14 ^m.
This experiment was used to assess:
1. The correlation between condition index and water content and particle size measured using ISO cleanliness. Linear
relationships were assessed form the results of Pearson correlation coefficient (r-value) combined with the corresponding
P-value. If P > 0.05, then any relationship, irrespective of strength, was considered not significant.
2. The association between condition index and water content and particle size. This was assessed in two phases.
a. Comparison of low and high condition index groups. By defining the median for condition index and generating two groups-low and high condition index groups. This division was used to divide water content and particle sizes. Groups were then compared using Student t-tests that were unpaired and assumed unequal variance.
b. Determining the association between condition index, water content and particle size. Using the groups defined in the above point. An additional step was to define the median for water content and particle sizes. Then to define these as high (above the median) or low (below the median). Their frequency distribution was then assessed using chi-square test to determine if there was a statistically significant relationship between condition index and water content and particle sizes defined according to the ISO cleanliness test.
2.4. Engineering analysis
2.4.1. Coefficients of variation (CV)
Recorded data was used to calculate intra- (i.e., within recording) and inter- (i.e., between recording) CV for HTF condition index and temperature. Intra-CV was calculated as the standard deviation (SD) divided by the mean and presented as a percentage. Inter-CV was calculated as the square root of the group average of the subject variance divided by its mean [12] and expressed as a percentage.
2.4.2. Presentation of data
Data from individual systems is reported as absolute values and grouped data is shown as means, means + standard deviation (SD) or medians unless stated otherwise. This case study focused on the analysis of HTF condition index and temperature. All data was analysed using Microsoft Office Excel 2007.
2.4.3. Statistical analysis
Statistical significance was taken as a P-value less than 0.05. Comparisons were conducted using a Student's unpaired t-
Table 2
Recordings conducted in air, water and in a mineral-based HTF.
Medium Units Mean + SD Intra-group CV% Inter-group CV%
Air (n=21) Condition index, arbitrary -1.0 7 0.4 0.0 0.0
Temperature, °C 24.2 7 2.2 0.5 0.7
Tap water (n=1) Condition index, arbitrary 21.0 7 0.0 0.0 0.0
Temperature, °C 19.6 7 0.7 3.38 0.0
Mineral-based HTF (n=6) Condition index, arbitrary 1.67 1.6 3.4 5.2
Temperature, °C 21.4 7 1.9 0.7 0.8
Note: n, number of samples assessed.
test that assumed unequal variance. Linear relationships were assessed using a Pearson correlation coefficient (r-value). The frequency distribution of data was assessed using chi-square testing.
3. Results
3.1. Experiment 1 - virgin sample, analysis of CV
Estimates inter- (between) and intra- (within) group CV for both HTF condition index and temperature are presented in Table 2. When exposed to air and tap water, HTF condition index was -1.0 + 0.4 and 21.0 + 0.0, respectively, and temperature was 24.2 °C + 2.2 and 19.6 °C + 0.7, respectively. In air, the intra- and inter-group CV was less than 1%, with intra-group CV tending to be lower than inter-group CV (0.5% versus 0.7%, respectively). The intra-group CV for tap water tended to be somewhat higher, but this was only from one sample.
When placed in HTF, the condition index and temperature were 1.6 + 1.6 and 21.4 + 1.9 °C, respectively. The relationship of intra- to inter-group CV was consistent with that recorded when the sensor was exposed to air with inter-group CV being slightly higher than the intra-group CV. When placed in virgin HTF intra-group CV was 3.4% and inter-group CV was ~5% (see Table 2).
3.2. Experiment 2 - used HTF sample, intra-group reproducibility
Table 3 shows the results from 3 systems and the intra-group CV for HTF condition index ranged between 0.64 and 4.84%. For temperature the range was lower and smaller (i.e., 0.23-0.51%).
3.3. Experiment 3 - water content of used HTF measured using laboratory test ASTM D6304 and the sensor
Table 3 shows the test results for these three systems for which water content was identified as being higher than expected and therefore out of specification. This real-life data shows that GTS was able to detect the content of water in a HTF, although the results, whilst not linearly related, compare favourably with those recorded with those in the laboratory. The GTS was used to assess the condition index of these HTFs. Data reveals the HTF condition index was high in all cases with values ranging between 4.91 and 7.33 (Table 3). The current population was restricted to three systems and indicates that the GTS detected the elevated water content. Indeed, based on the condition index, Fig. 2 would indicate that water content to be between 1 and 1.5% (10,000 and 15,000 ppm). Values that are slightly higher than recorded in Table 3, but not too dissimilar based on a small population.
Table 3
Test results obtained in the laboratory from three HTF systems using mineral-based HTFs.
System Water content (ppm) HTF condition index (mean 7 SD, in-tra-CV%) HTF temperature,°C (mean 7 SD, intra-CV%)
Measured using standardised laboratory testing ac- Measured using GTS Measured using GTS
cording to ASTM D6304
System 1 > 10,000.00* 5.947 0.29, 4.84 22.967 0.12, 0.51
System 2 2200.00* 4.91 7 0.03, 0.64 23.1 7 0.09, 0.38
System 3 9800.00* 7.33 7 0.13, 1.82 23.357 0.05, 0.23
Note: Test results revealed that for system 2, ferrous wear (145 ppm) and elements (146 ppm) were out of specification and above pre-defined parameters for a mineral-based HTF (see Table 3). For the above parameters, standardised testing indicated that this parameter (*) was out of specification and above pre-defined values. CV, coefficient of variation.
Fig. 2. Stimulus-response curve: changes in HTF condition index (filled circles) and temperature (open circles) to step-wise dilution of the HTF with water (percentage dilution, (A) parts per million, (B). (A) Dilution of HTF with water expressed as a percentage. (B) Dilution of HTF with water expressed as a percentage. Note: HTF temperature was 25.01 + 0.790C.
3.4. Experiment 4 - the response of the sensor to step-wise increases in tap water only
A stimulus-response curve was constructed to dilution of a virgin mineral-based HTF with tap water (see Fig. 2). Prior to tap water being added, baseline condition index and temperature were 0.0 + 0.0 and 21.8 °C + 0.03, respectively. Fig. 2 shows that HTF condition index increased gradually between 0 and 1.5% dilution and then rapidly increased at a dilution of 1.75% tap water. Between 0 and 1.5% dilution, the insertion of a linear correlation reveals a strong relationship between dilution and HTF condition index (r=0.912; y=3.2188x+0.6636). This analysis revealed that HTF condition index increased by 0.80 for each 0.25% increase in water (or for every 2500 ppm). In contrast, the HTF condition index increased by 14.69 for the final dilution from 1.5 to 1.75% (see Fig. 2) (Table 4).
3.5. Experiment 5 - the real-life functionality of the sensor assessed during the flushing of a newly built plant
3.5.1. The correlation between condition index and water content and particle size
Data for this group analysis is presented in Table 5 with mean values presented for condition index, water content and particle sizes are presented in Table 5. Linear comparisons revealed no significant correlation (P > 0.05) between condition index and water or ISO cleanliness scores.
3.5.2. Comparison based on low and High condition index groups
Subgroup analysis was conducted to assess the relationship between condition index and the other parameters. Table 6
Table 4
Severity rating for water in mineral-based HTF and based on results from test ASTM D6304.
Severity rating based on the water content in a HTF
Satisfactory
Caution
Action
Serious
Parameter Typical
values
Water content, < 100 ppm
No action required Z 100 to < 300
Start planning your planned preventative activities Z 300 to < 500
Initiate your planned pre-ventative activities Z 500 to < 759
Urgent and immediate action required Z 759
Table 5
Mean values for condition index, water content and ISO cleanliness scores for the HTF.
Population Condition index, arbitrary aWater, ppm ISO cleanliness score, microns
4 6 14
19 bPearson correlation coefficient (r-va-lue), P-value 14.2 7 3.6 Not applicable 207.5 7 302.7 r—0.352, P=0.140 22.47 1.4 20.87 1.4 r= 0.088, P= 0.722 r=0.319, P=0.183 17.07 2.9 r= 0.209, P=0.391
Note: Data is presented as mean 7 SD from 19 samples. a Measured using test method ASTM D6304.
b Values were obtained by comparing condition index with all other parameters.
divides data based on low and high scores for condition index, which was defined using the group median. Numerically higher values were obtained in the higher group than the lower group, but pairwise comparisons between respective groups revealed only a significant difference between low (3.8 + 0.5) and high condition index groups (12.2 + 5.2; P=0.0006, Student's t-test).
3.5.3. The association between condition index and water content and particle size
Chi-square analysis was used to assess the distribution of values based on the calculated median value for water content and ISO particle cleanliness score groups (see Table 7). Interestingly, this analysis showed a significant relationship (P < 0.05, chi-square test) between the distribution of water content and condition index scores whereby low condition index had a higher number of low water content recordings and the high water condition index group had a higher relative number of high water content recordings. This same relationship (P < 0.05, chi-square test) was observed for ISO cleanliness scores of 6 and 14 ^m, but not for the 4 ^m group (P > 0.05, chi-square test).
4. Discussion
The GTS can be used to make rapid assessments of HTF condition based on the fluid's permittivity. The current study assessed the characteristics of the sensor using virgin and real-life HTF samples. This involved bench-top experiments to assess the reproducibility of the sensor using clean (virgin mineral-based HTF) and dirty fluid (in-use mineral-based HTF). Data showed that intra-group CV for condition index was slightly higher for the used (up to 4.84%) than clean HTF (3.4%) and that inter-group CV, using a clean HTF, tended to be slightly higher (5.2%) than intra-group CV. This was also true for temperature.
This article also explored the stimulus-response to the increasing water content of a clean HTF. Plots showed condition index increased steadily up to 1.5% water (i.e., 15,000 ppm) and then rapidly increased beyond this value, reaching a saturation level at 1.75% tap water. This data was complemented by real-life data and specifically data collected during the flushing of a new HTF system build. This unique dataset showed an association between the condition index recorded with GTS and both HTF water and cleanliness.
New technologies offer new opportunities in terms possible applications but also limitations. The GTS is of value in the area of HTF as is can, in a matter of seconds, provide an estimate of the condition of a HTF. In the current case this was done in the laboratory, but the technology has the capability to be plugged into a laptop via a USB port. This means that analysis can be conducted wherever and whenever the sensor is needed. In the field, this means that a sample can be taken and, once cooled to ambient conditions, it can be analysed to assess the extent of contamination (e.g., water), the presence of wear particles and to assess the extent of early oxidation and fluid cleanliness. In a laboratory or research institute, the sensor could also find a use in assessments of HTF quality and this could be done, for example, as HTFs arrive and depart a
Table 6
A comparison of water and cleanliness scores based on low and high condition index scores.
Condition index score Condition index, arbitrary Water, ppm ISO cleanliness score, microns
4 6 14
Low condition index score, n=9 High condition index, n=10 P-value 3.87 0.5 12.27 5.2 = 0.0006 46.67 52.5 173.657 276.7 0.186 22.1 7 0.8 22.37 1.3 0.709 19.37 1.9 20.47 1.7 0.224 14.87 3.4 16.37 3.0 0.321
Note: Data is presented as mean 7 SD from 19 samples. Paired-wise comparisons were conducted using a Students' unpaired t-test assuming unequal variance. Data was divided into two groups based a low ( < 4.3) and high ( >4.3) condition index score and this split was based on the median condition index score.
Table 7
Frequency distribution of water (A) and ISO cleanliness scores (B-D) based on low and high condition index scores. Data presentation: values are presented in the following form: 'observed value' (expected value) [Chi-square value] for the Chi-square test.
A. Median water content
Analysis Water, < 64 ppm Water, Z 64 ppm Row totals (n)
Low condition index score 6 (3.79) [1.29] 3 (5.21) [0.94] 9
High condition index 2 (4.21) [1.16] 8 (5.79) [0.84] 10
Column totals, n 8 11 n=19 (Grand total)
Chi-square statistic (P-value) 4.2318 (P=0.039673)
B. Median iso cleanliness score of 4 |im
Analysis Score, <23 Score, Z 23 Row Totals (n)
Low condition index score 6 (4.26) [0.71] 3 (4.74) [0.64] 9
High condition index 3 (4.74) [0.64] 7 (5.26) [0.57] 10
Column totals, n 9 10 n=19 (Grand Total)
Chi-square statistic (P-value) 2.5544 (P=0.109984)
C. Median iso cleanliness score of 6 |im
Analysis Score, <20 Score, Z 20 Row totals (n)
Low condition index score 6 (3.79) [1.29] 3 (5.21) [0.94] 9
High condition index 2 (4.21) [1.16] 8 (5.79) [0.84] 10
Column Totals, n 8 1 n=19 (Grand Total)
Chi-square statistic (P-value) 4.2318 (P=0.039673)
D. Median iso cleanliness score of 14 |im
Analysis Score, < 15 Score, Z 15 Row totals (n)
Low condition index score 6 (3.79) [1.29] 3 (5.21) [0.94] 9
High condition index 2 (4.21) [1.16] 8 (5.79) [0.84] 10
Column totals, n 8 1 n=19 (Grand total)
Chi-square statistic (P-value) 4.2318 (P=0.039673)
warehouse.
The current study was designed to assess the reliability of measurements. Experiments 1 and 2 assessed the reprodu-cibility of the sensor from calculations of intra- and inter-group CV for clean (laboratory) and dirty (real-life) HTF samples. Such measurements are critical as they indicate the reliability of measurements and need to be reproducible-able so the customer can be informed about the condition of their asset [9].
Results show that measurements made using a small 50 ml sample in a temperature controlled room had good re-producibility with coefficients of variation being roughly equal or less than 5%. Indeed, HTF condition index had an intra-group CV of 3.4% and an inter-group CV of 5.2%. The intra-group CV was also similar to that achieved using a dirty (real-life sample). This indicates that within group analyses are more reliable than those made between groups. This is not surprising as intra-group CV% was based on ten measurements whereas inter-group analyses were based on 6 measurements. Furthermore, it is normal for within-group CV to be lower than between-group comparisons and this was also confirmed from measurements of temperature (see Table 2).
In experiment 4, the relationship between water content and HTF condition index was assessed and is presented in Fig. 2. HTF condition index increased gradually up to a water content of 1.5% with a strong linear relationship (R2=0.8321; y=3.2188x+0.6636) being shown. Data shows that for condition index increased by 0.80 for each 0.25% increase in water content. Above 1.5% water content, the HTF condition index increased much more steeply and saturation was reached (i.e., condition index of 21) when water content reached 1.75%. In these experiments tap water was added to a virgin HTF and shaken to mix the fluids. In real-life this may be the case when sampling HTFs that have been stored and a useful quality standard. In such cases, the HTF should be sampled from the lowest point, e.g., using an IBC tap. In such cases the presence of water may be visibility detected. In real-life cases, this may not be so apparent and presented in experiments 3 and 5 below.
The relevance of laboratory tests needs to be discussed in terms of real-life applications as was done in experiments 3 and 5 and this was driven by data presented by experiment 3. Indeed, in experiment 3 water content was compared with condition index score for 3 real-life samples (see Table 3). At the present time, water content is performed in the laboratory and according to ASTM D6304. Once a value has been defined for a HTF, test results rated according to a pre-defined rating system and judged as being satisfactory (within specification) or not satisfactory (out of specification). The specific ratings are presented in Table 4. In experiment 3, the results water content was judged to be severely out of specification according to the values presented in Table 4. The laboratory results were compared with those recorded with the GTS and for the limited number of samples (n = 3) there was no clear linear relationship, although data did suggest that condition index was scored to be around 5 and indicates that the sensor was detecting the change in condition of the HTF.
The relationship between water and condition index was explored further in experiment 5. This data confirmed that there was no linear relationship (P > 0.05 for all inserted correlation coefficients) between water content of a HTF and its condition index score. Furthermore, this experiment confirmed this was also the case for cleanliness scores of a mineral-based flushing and cleaning fluid. This begs the question as to whether there is an association between condition index and both water and cleanliness. Subgroup analysis was performed to determine if high and low condition index scores were related to high and low scores of water content and cleanliness. Whilst values were numerically higher based on this split, statistical comparisons showed no difference. To understand these numerical differences, data was analysed to determine the frequency of distribution of water content and cleanliness. Again, this was done by scoring these parameters as high or low, based on their group median. This analysis shows that here is an association between condition index scores and water content scores, with the statistical test results demonstrating a positive association i.e., a high frequency of condition index scores is associated with a higher frequency of water content scores. This means that as the count of higher condition index scores increases so does higher scores of water content and vice versa. The same association was also found for cleanliness scores, particularly particle sizes 6 and 14 ^m. Taken together this suggests that GTS is a measure of condition per se of the fluid as opposed to reflecting simple one parameter.
5. . Conclusions
The GTS can be used to make rapid assessments of HTF condition based on the fluid's permittivity. This can be done repeatedly and tests showed good reproducibility (roughly 5% or less) based on calculations of coefficients of variation. Experiments confirmed that the sensor responds to changes in water and also fluid cleanliness. Stimulus-response curves showed that increases in water were detected below a 1% level and rapidly increased between 1.5 and 1.75% water. The sensor was also used to show that in a real-life setting, during the flushing of a newly built facility, condition index reflected changes in fluid water and cleanliness. The advantages of the sensor are that measurements can be conducted using small samples of HTF ( < 50 ml), they can be performed on-site at the point of fluid sampling. This has potential applications for the quality management of high value goods such as HTFs and may be a simple approach for customers that need to make a lot of regular samples over a large processing site, such as solar farms.
Acknowledgements
The authors would like to thank the engineering and technical support provided by Global Heat Transfer and the writing support provided by Red Pharm communications.
References
[1] O. Walter Wagner, Heat transfer technique with organic media, 2nd ed., Maria-Eich-Strape, Graefelfing, Germany, Chapter 2, Heat Transfer Media, 1997, pp. 4-58.
[2] C.I. Wright, Thermal HTF problems following a system flush with caustic and water, Case Stud. Therm. Eng. 2 (2014) 91-94.
[3] C.I. Wright, A. Burns, C. Jones, Spling hot heat transfer fluids: simple insights for gaining a representative sample, Int. J. Eng. Innov. Technol. 3 (2013) 202-204.
[4] ASTM Standards D92 Standard Test Method for Flash and Fire Points by Cleveland Open Cup Tester. Link: <http://www.astm.org/Standards/D92.htm>. accessed: 03.09.14.
[5] ASTM Standards D93 Test Methods for Flash Point by Pensky-Martens Closed Cup Tester. Link: <http://www.astm.org/Standards/D93.htm>. accessed: 03.09.14.
[6] Solutia. Liquid phase systems design guide: A design, operating low-pressure heat guide for low-cost, and maintenance transfer systems. Publication number 7239128D. Source: <www.therminol.com/pages/tools/therminol_liquid_phase.pdf>. accessed: 24.04.14.
[7] Heat transfer by organic and synthetic fluids, in: Factory Mutual 7-99. Property Loss Prevention Data Sheets 12-19. Source: <ftp://cable-129-140-83. b2b2c.ca/sda1/Basement/Mes%20documents/Reconnaissance%202011/Codes_Normes/Factory%20Mutual/DS/7-99.PDF>. accessed: 24.04.14.
[8] Water in thermal oil systems; detection, removal and prevention: Source: <www.paratherm.com/resources/usersguide/water-detection-removal-and-prevention>. accessed: 14.09.14.
[9] T. Ennis, Safety in design of thermal fluid heat transfer systems, Hazards XXI Symp. Ser. Number 155 (2009) 162-169.
[10] A. Torres Pérez, M. Hadfield, Low-cost oil quality sensor based on changes in complex permittivity, Sensors 11 (11) (2011) 10675-10690.
[11] C.I. Wright, E. Picot, A case study to demonstrate the value of a system flush and clean prior to filling a plant with virgin heat transfer fluid, Heat Transfer Eng. http://dx.doi.org/10.1080/01457632.2015.1067061.
[12] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurement, Lancet 1 (1986) 307-310.
[13] American society for testing and materials. Source: <www.astm.org>. accessed: 14.09.14.
