MICRO-COGENERATION HYDRAULIC TESTING FACILITY
E. Bouchera* and I. Beausoleil-Morrison
a
Carleton University, 1125 Colonel By Drive,Ottawa, Ontario, Canada
evan_boucher@carleton.ca
ABSTRACT
A heavily instrumented and highly flexible
hydraulic test facility has been commissioned to
characterise the performance of micro-
cogeneration systems. The facility has the
ability to reproduce typical house space heating
loads and domestic hot water demands suitable
to investigate the operation of units under
realistic operating conditions. It can also be
configured to create and sustain controlled
thermal boundary conditions required to
generate data suitable for model calibration or
performance mapping. The facility provides a
platform to experimentally investigate
appropriate configurations to integrate micro-
cogeneration units into residences through
variations in the balance of plant components
and control strategies.
The test bench consists of a number of fluid
circuits allowing heat transfer between a
cogeneration device, variable volume thermal
buffers, supplementary heating systems, and
thermal loads. Heat addition to the system can
be supplemented by either a 57 kW
instantaneous condensing hot water heater
rated at 98% percent efficiency (relative to the
HHV), or through a variable 20 kW electric in-
line resistance heater. Thermal load simulation
is realized through flat plate heat exchangers
connected to a dedicated roof-top chiller unit.
Heat transfer can be measured across 11 key
locations through the use of positive
displacement oval gear meters and a
combination of thermocouples and 5 junction
thermopiles. Then embedded temperature
sensors were designed, built and calibrated in-
house to reduce overall measurement
uncertainty.
Keywords: micro-cogeneration, experimental,
thermal storage, measurement uncertainty
INTRODUCTION
As a consequence of increased greenhouse
gas emissions and energy use from the
residential sector, there has been renewed
interest in fuel cell and combustion-based
micro-cogeneration systems. A number of
studies have demonstrated that simultaneous
production of heat and electricity can help
meet the needs of a home while reducing
primary energy demand and associated
emissions [1,2]. The potential for reduction
depends on a number of factors including the
thermal energy demand and buffering for heat
and domestic hot water, electricity demand
profile, and the local electricity mix [3]. In order
to inform policy makers and to support the
design and deployment of micro-cogeneration,
continued research must be undertaken to
determine how best to integrate these
technologies into homes. Initiatives including
field trials, experimental testing and simulation
must be carried out to characterise system
level performance. Although field trials can be
extremely informative, they are often costly,
time consuming and can be logistically difficult
to execute. Models such as those created as
part of the International Energy Agency’s
Annex 42 program provide much simpler
implementation and greater flexibility;
however, they require device specific
calibration [4]. To this end, there remains a
need to build and commission facilities in order
to generate data suitable for model calibration
and validation of individual units, and to
experimentally study complete systems.
TEST BENCH CONFIGURATION
A flexible hydraulic test bench was built to
explore suitable balance of plant configurations
and to provide realistic, controlled thermal
loading to micro-cogeneration units. The
facility is comprised of 2 auxiliary heaters, 2
thermal stores, 3 flat-plate heat exchangers,
and 4 centrifugal pumps. Figure 1 shows the
possible thermal energy flows between
components while Figure 8 shows the
complete hydraulic layout of the facility. Heat
transfer can be measured across 11 key
locations spanning all plant components.
A central data acquisition and control system
(DAQ) allows for complete automation of the
pumps, heaters and the plant network’s 11
electronically actuated 3-way ball valves. The
pumps selected for each fluid circuit can be
found below in Table 1. The specifications of
the heat exchangers and auxiliary heaters are
listed in Tables 2 and 3. The inline electric
heater can be remotely controlled to maintain a
specific output temperature or its thermal
output can be directly modulated up to full by the natural gas heater. Direct fluid
power in one percent increments. exchange between tanks is also possible
(shown with a dashed line in Figure 1). A 10
cm thick layer of rigid glass fibre board strips
Electric laminated to an aluminium foil backing serves
Circulation to insulate each tank. A flexible insulated,
Heater Tank A inflatable cover was built to further reduce heat
losses and evaporation through the top of the
tanks. Each tank was fitted with 4 headers
spaced at 90 degree intervals. Each individual
header allows for the addition or removal of
Micro- Natural
fluid at 5 different points along the height of the
Cogeneration HX-1 Gas HX-2 tank. The vertical spacing between inlets is
Device Heater roughly 20 cm. Figure 2 shows the layout of
the tank and header system.

Table 2: Selected Flat Plate Heat Exchangers

HX-3 Tank B Heat Capacity Manufacturer
Exchanger (kW) /Model
SEC/
HX-1 40
M31B-50
Figure1: Possible energy flow paths
SEC/
HX-2 20
The instantaneous natural gas water heater M11DC-50
outlet temperature can be varied in 5-degree
increments to a maximum of 85°C. The unit SEC/
HX-3 30
was retrofitted in-house with circuitry allowing M31B-30
the set-point to be controlled using the DAQ
system. Heat exchangers were preferred over Each tank has also been equipped with a
direct heat transfer because water was used thermocouple rack used to determine vertical
as the heat storage medium while the temperature profiles during experiments. As
cogeneration device and chiller circuits employ many as 16 measurements can be taken along
a glycol solution. The capacity listed for the the centreline of the tank.
heat exchangers are measured values under
typical operating conditions observed during
commissioning.
TC E
Table1: Selected Pumps
Manufacturer/ Inflatable
Pump Capacity Insulated Top TC D
Model
(L/min)
TC C
Pump-1 36.6 Grundfos/
CR1-4
TC B
V =1300L
Pump-2 36.6 Grundfos/
CR1-4 TC A

Pump-3 36.6 Grundfos/

CR1-4 Figure 2: Storage tank and headers
Pump-4 3.79-36.6 Grundfos/
Due to the size of the tanks, a minimum
CR1-3
volume of 100L is required before fluid can be
fed to the hydraulic test bench. The selection
Thermal storage is provided by two 1300L of inlet and outlet heights must be done
open cylindrical tanks having a maximum manually using ball valves located between the
working temperature of 60°C. The tanks can tank and header outlet. Controlled electrical
be charged in series or in parallel by the loading of cogeneration units is also possible
cogeneration device through HX-1 or directly using a programmable AC/DC load. The
 Table 3: Auxiliary Heaters Table 5: Flow Meter Specifications
Power Manufacturer Max.
Heater Flow Flow Manufacturer/
(kW) / Model Accuracy
Meter Rate Model
Tankless (L/min)
Navien
Natural Gas 57
NP 240-A Brooks / 1% of
Heater F-1 3-25
BM-04 reading
Electric
20 Wattco MFLS Brooks / 1% of
Circulation F-2 3-25
BM-04 reading

specifications of the load can be found in Table Brooks / 1% of

F-3 7.6-70
4. The load has been integrated with the DAQ BM-07 reading
system to subject units to user defined Brooks / 1% of
electrical load profiles realized at flexible time F-4 3-25
BM-04 reading
intervals. The unit also comes equipped with
on-board instrumentation to measure current
and voltage. When used in conjunction with Table 6 describes the meters chosen. Under
the hydraulic bench, electrically led micro- typical operating conditions, the warranted
cogeneration system configurations can be accuracy is 1% of reading. The primary dial
3
explored. travels one full rotation per 0.0283 m of flow
through the meter.
Table 4: AC/DC Programmable Load
Table: 6 Natural Gas Flow Meters
Max. Load (W) 4500
Flow
Manufactur
Resolution (W) 1.125W Rate Accuracy
er /Model
(L/min)
Accuracy 0.2% + 0.3% F.S.
95 Elster / 1% of
Voltage Range (V) 50-350 DTM 200-A reading

Max Current (A) 45 95 Elster / 1% of

DTM 200-A reading

INSTRUMENTATION In order to measure the gas flow with the DAQ

system, a rotary encoder with a pulse rate of
In order to take advantage of the flexibility of
1000 pulses per revolutions was mounted to
the hydraulic test bench, a large amount of
the dial of the meter. The combination of the
instrumentation was added at critical locations.
dry test meter and encoder provides excellent
Positive displacement oval gear flow meters
resolution of natural gas volumetric flow.
were selected to measure the flow through
Estimates of the molar flow rate of natural gas
each of the pumps. The meters are factory
can be derived using the measured
calibrated to produce a train of pulses
temperature, pressure, and flow rates
proportional to the measured flow rate. An
observed on the line. An estimate of the
estimate of the flow can be obtained by
heating value of the fuel is needed to compute
reading the number of pulses registered over a
the molar heat rate into a particular device.
given sampling interval. The specifications of
each meter can be found in Table 5. The
A large number of thermocouples and
natural gas line used to fuel the instantaneous
thermopiles have been embedded into the
water heater and cogeneration unit has been
hydraulic test bench and storage tanks to
instrumented to measure temperature,
measure both single point temperatures and
pressure and volumetric flow. Gas temperature
temperature differences between locations.
is measured upstream of the flow meters using
Both thermocouples and thermopiles were
a thermocouple while pressure is measured at
calibrated over the entire operating range of
the inlet of the facility to within 0.25% of
the test bench in order to reduce measurement
reading. Volume flow to the water heater and
uncertainty. Table 7 shows the total
cogeneration device are measured separately
uncertainty for a temperature measurement
using dry test, revenue grade diaphragm
including error sources from both calibration
meters.
and experimental measurement. This includes
sources such as voltage reading errors,
reference temperature bias, cold junction
compensation and calibration uncertainty.

Table 7: Temperature Sensor Accuracy

Temperature Accuracy
Manufacturer
Sensor (°C)

Type T
Omega 0.5
Thermocouple

5 Junction
Built In-house 0.1
Thermopile

In order to evaluate the heat transfer between

components, an estimate of the fluid flow rate
and temperature change must be made.
Compound sensors comprised of a single 5-
junction thermopile and 2 thermocouples were
placed across relevant components such as
heat exchangers, heaters and tanks. The
temperature difference can be read from the
two independent thermocouple readings or
directly from the output from the thermopile.
Figure 3 is a schematic of the compound
sensors embedded into the test bench.
It should be noted that the temperature reading
using the thermocouples is software
compensated and measures the temperature
either at the inlet or outlet of a component.
Thermopiles are a temperature difference
sensor by definition whose output (emf) is
proportional to the difference between the hot
and cold junctions embedded across a
particular component. The 5-junction
thermopile is constructed by creating 5
thermocouple junctions on both the supply and
return sides of a component. The
thermocouple junctions are wired in series to
produce a thermopile producing a voltage
approximately 5 times as large as that of a
single thermocouple wired across the same
temperature difference.
The uncertainty for a temperature
measurement difference made by taking the
difference between two thermocouple
measurements is 7 times greater than that of
the calibrated thermopile. The difference
between these methods can be attributed to
two main factors. Firstly, once calibrated, the
output voltage from the thermopile can be read
without any need for cold junction
compensation. Secondly, the increase in signal
strength greatly reduces the error attributed by
the resolution limit on the hardware used to
read the emf produced.
Figure 3: Compound temperature sensor
SENSOR CALIBRATION AND
UNCERTAINTY ANALYSIS
As previously mentioned, a calibration was
undertaken to reduce the uncertainty
associated with temperature and temperature
difference measurements. Rather than
implementing the coefficients from the National
Institute of Standards and Technology (NIST)
used in typical thermocouple measurements, a
unique calibration data set relating the
thermocouple output voltage to a temperature
was experimentally derived. The calibration
data was generated by immersing the
measuring junction of the thermocouple into a
constant temperature bath and logging the emf
produced. The cold junction temperature was
recorded using a thermistor built into the DAQ
system while the measuring junction
temperature was recorded using a platinum
resistance thermometer with a resolution of
0.02°C. Measurements of both voltage and
temperature were recorded at 5°C intervals
between 0-85°C. A second order polynomial
was fitted to the data.
The error resulting from interpolation within the
data set is significantly smaller than that from
the generic curves because the temperature
range considered is relatively small. The
correlation developed is used to convert the
output voltage to a temperature and to
implement a software compensation to map relating an output voltage to a
account for the cold junction temperature. The temperature difference. For a given cold
individual bias sources contributing to the total junction temperature, a data set of voltages
measurement uncertainty can be found in was created by increasing the hot junction
Table 7. The total bias was calculated by temperature in increments of 5°C until the
taking the summation of the root mean square upper temperature of the test bench was
of each bias source. The largest source of reached. This procedure was repeated for a
error results from the thermistor used to large range of cold junction temperatures to
determine the cold junction temperature. The characterise the thermopile response. Figure 4
manufacturer of the DAQ system warrants a shows the matrix chosen to calibrate the
resolution of 0.2°C for the thermistor and thermopiles. Multivariate regression was used
recommends an additional bias of 0.2°C to to express the temperature difference as a
account for its location within the block rather function of cold junction temperature and emf
than directly on the cold junction. The second produced. The calibration data was separated
largest source of error results from the into 3 individual sections to reduced the
resolution of the instrumentation used to read regression errors.
the thermocouple voltage. This source of bias
occurs during both the calibration procedure Cold
and during experimental measurements. Junc.
Temperature Difference °C
Temp.
Table 7: Thermocouple Measurement Bias (°C)
Bias Source Bias (°C) 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85
5 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Reference junction 0.130 10 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Read error (experiment) 0.200 15 5 10 15 20 25 30 35 40 45 50 55 60 65 70
20 5 10 15 20 25 30 35 40 45 50 55 60 65
Platinum RTD 0.020 25 5 10 15 20 25 30 35 40 45 50 55 60
Thermistor 0.250 30 5 10 15 20 25 30 35 40 45 50 55
35 5 10 15 20 25 30 35 40 45 50
Thermistor location 0.200
40 5 10 15 20 25 30 35 40 45
Regression prediction error 0.080 45 5 10 15 20 25 30 35 40
50 5 10 15 20 25 30 35
Read error (calibration) 0.200
55 5 10 15 20 25 30
Total Bias 0.454 60 5 10 15 20 25
65 5 10 15 20
A “reference junction” error was introduced 70 5 10 15
75 5 10
during the calibration because an average cold
80 5
junction temperature was used to develop the
curve fit while this value varied throughout the Figure 4: Thermopile calibration matrix
calibration experiment. A regression prediction
bias error was also applied to account for the As with the thermocouples, a total bias error
maximum possible variation between the was obtained by taking the summation of the
calibration data and the polynomial fit used to root mean square of each individual bias
represent the data set. Finally, the bias of the source. The sources contributing to the total
platinum RTD used to measure the calibration bias of the thermopile measurement can be
bath temperature was included to complete found in Table 8. The regression prediction
the overall measurement uncertainty. error contributed the largest fraction of the
uncertainty followed by the read error from the
A similar calibration exercise was conducted to voltage measurements occurring during both
calibrate the 5 junction thermopiles. The cold the calibration and experiment. Although the
and hot junctions were placed in two separate use of the thermopiles requires that the cold
baths to create a known temperature junction temperature be measured using a
difference across the thermopile and the thermocouple, the bias from this measurement
voltage produced was measured directly using need not be included in the uncertainty for the
the DAQ system. The bath temperatures were temperature difference measurement. This can
simultaneously measured and logged using be attributed to the fact that the relationship
two platinum resistance thermometers. between the emf produced and temperature
Because the output of the thermopile is difference is linear and that this slope remains
sensitive to both the hot and cold junction virtually constant over such a small change in
temperatures, both of these parameters were the cold junction temperature. The uncertainty
systematically varied to produce a calibration calculation for derived quantities such as
thermal power, and thermal efficiency, requires parameter will change based on the sampling
that the bias from individual measurement interval. Only an integer number of pulses can
sources be propagated into a final result. The be recorded over a particular sampling interval
method developed by the American Society of creating a resolution error for short sampling
Mechanical Engineers was adopted for intervals. As the sampling interval increases,
calculation of experimental uncertainty for all this error diminishes and the bias from flow
derived quantities. measurement becomes a constant for a given
flow rate. As the measured temperature
Table 8: Thermopile Measurement Bias change increases, the contribution from the
Bias Source Bias (°C) temperature bias also becomes less
significant.
Read error (calibration) 0.04
Read error (experiment) 0.04 An uncertainty analysis was repeated to
RTD 1 Bias 0.02 determine the bias on the calculation for the
thermal efficiency of the natural gas heater
RTD 2 Bias 0.02 over its complete operating range and can be
Regression prediction error 0.08 found in Table 10. The bias is reported as a
Total Bias 0.10 percentage of thermal efficiency for different
power levels and sampling intervals. The
maximum flow rate through the unit is
The total bias can calculated by taking the restricted to 18 L/min due to the limitations of
summation of the root mean square of each the pump. A lower heating value of 800
bias source each weighted by a sensitivity kJ/kmol was chosen for natural gas. The
parameter [5]. The sensitivity parameter is heating value was assumed to vary by 0.5%.
obtained by taking the partial derivative of the
derived quantity of interest with respect to the Table 10: Natural Gas Heater Efficiency Bias
bias source. Table 9 shows the bias on the
power measurement of selected circuits on the Efficiency Bias (%)
hydraulic test bench. The bias is reported as a Sampling interval (s)
percentage of the thermal power being Delta Power
measured. 1 60
T (°C) (kW)
5 6.2 25.3 4.4
Table 9: Selected Uncertainty
10 12.6 15.6 4.1
Sampling Interval (s)
20 25.1 11.6 4.0
1 60
30 37.7 10.5 3.9
Flow Thermal
Bias 45 56.5 9.9 3.9
Rate Power Bias (%)
(%)
(L/min) (kW)
The resolution error from the pulse output on
Cogeneration Heat Dump Loop the natural gas meters dominates the
30 2 12.3 12.0 uncertainty for 1-second sampling interval. As
the sampling interval is increased to 60
30 5 5.5 4.9
seconds, the errors attributed to both gas and
30 10 3.6 2.6 water flow are reduced and the error on the
30 20 3.0 1.6 temperature difference measurement becomes
more significant. The bias from the natural gas
30 40 2.8 1.2 temperature and pressure were also included
Tank Loading Loop for completeness although their contributions
18 2 8.2 7.3 are negligible.
18 5 4.9 3.1
An uncertainty analysis was also undertaken to
18 10 4.2 1.8 estimate the bias for the electrical, thermal,
18 20 4.0 1.2 and total efficiency of a 6kW e cogeneration unit
being loaded by the hydraulic test bench. The
bias used for the LHV is the same as
The bias for thermal power measurement
described for the natural gas heater while the
based on fluid flow rate and temperature
flow rate used for heat recovery was increased
change across a component is sensitive to a
to 30L/min. Electrical power was assumed to
number of factors. Although the flow rate in
be measured using revenue grade energy
each loop is constant, the bias for this
meters having a pulse output and current
transducers. The specifications for the energy
meters and transducers can be found in
Tables 11 and 12.
Table 11: Energy Meter Specifications
Manufacturer WattNode
Model WNB-3Y-208-P 120
Nominal line
120
voltage (V)
Frequency 48 - 62 Hz
Output
1333.33
(pulse/kWh)
Accuracy 0.5% of reading
The energy meters require current transducers
having an output ranging from 0-3.33 VDC
which is proportional to the instantaneous line
current. Based on the maximum electrical
power of the unit, 30 A current transducers
were selected. Custom split core units were
chosen to further reduce the bias on the power
measurement. The phase error was not
propagated into the final result because purely
resistive loading is possible using the
programmable load. Ultimately, the uncertainty
on the power measurement was dominated by
the resolution of the energy meter. Although
the units have good accuracy, they are built for
long-term measurements and lack the
precision needed for very short sampling
intervals. Even at 10 second sampling
intervals, the resolution error from the pulse
output is an order of magnitude larger than
other sources of error. Despite this
observation, reasonable estimates of power
can be made provided the sampling interval is
long enough.
Table 12: Current Transducer Specifications
Manufacturer Magnelab
Type Split core
Rated current (A) 30
Linearity error (%) 0.2
Phase error (%) <1
Table 13 shows the extended uncertainty for
the cogeneration device efficiency over a full
operational range. As many of the
measurement transducers produce a pulse
output, the total efficiency uncertainty was
calculated at 10, 30 and 60-second sampling
intervals. As the units electrical output and
sampling interval are increased, the
uncertainty for all measured and derived
quantities is reduced.
COMMISSIONING
A number of experiments were undertaken
during the commissioning phase to access the
performance of the test bench and determine
its limitations. The test bench can be
configured to subject the cogeneration device
to both steady-state and varying inlet
conditions. Heat can be directly added to the
flow stream through the inline electric
resistance heater or rejected to a roof top
chiller unit through a flat pate heat exchanger.
The temperature of the heat transfer fluid
returning to the cogeneration device can be
fixed by diverting a portion of the flow around
the heat exchanger (HX-3). Fine-tuning of
systems controlling the response of the
proportional valve (V-2) and the inline electric
heater was required to achieve optimal
performance. Figure 5 shows how the test
bench can be used to fix the return
temperature. By varying the mass flow rate
and fluid temperature through the micro-
cogeneration unit, a complete performance
map of thermal efficiency can be produced.
Figure 5: Return temperature to cogeneration
device
The hydraulic test bench can also be
configured to reproduce time-varying home
heating and domestic hot water loads in order
to evaluate system level performance under
typical operation. Thermal loading of the
storage tanks is simulated by varying the
volume of flow through HX-2. Another DAQ-
controlled proportional valve (V-1) placed
upstream of the heat exchanger diverts a
portion of the flow in order to dynamically
modulate the amount of thermal power
dissipated. Careful tuning was required to
ensure that the desired load profile was
reproduced. Figure 6 shows a plot comparing
the target load profile with the one measured in
real time. Overall, the system responded
rapidly and maintained good agreement with
the desired loading.
Figure 6: Dynamic tank thermal loading
Under conditions where the storage tank is
unable to meet the given load, the natural gas
heater can be used to supplement the heating
demand. User defined load profiles can be
produced using time intervals as small as 5
minutes. Experiments ranging in length from
several hours to several weeks are possible
and can be run autonomously. The total heat
rejection through the exchanger is limited by
the upper temperature limit of the storage
tanks and the flow rate of the pump. The
hydraulic test bench is well suited to evaluate
the performance of micro-cogeneration
systems operating under typical conditions
over the course of a single day or several
weeks.
Figure 7: Tank temperatures during charging
A flexible tank configuration was chosen to
support a number of charging and discharging
strategies. Figure 7 shows the evolution of
tank temperatures while being charged by the
natural gas heater. The instrumentation within
the tank can be used to determine the level of
stratification developed for different loading
schemes
CONCLUSION
The design, construction and commissioning of
the micro-cogeneration test facility has been
completed. The calibrated instrumentation and
plant level controls have been fully integrated
with the central DAQ system allowing for
almost complete autonomous operation of the
facility. The test bench can be used
characterize the performance of micro-
cogeneration devices and to investigate
suitable balance of plant configurations and
control strategies.
REFERENCES
[1] I. Beausoleil-Morrison and H. Ribberink,
The Potential for Reducing
EnergyConsumption and Greenhouse
Gas Emissions in the Ontario (Canada)
Housing Sector with Solid-Oxide Fuel-Cell
Micro-Cogeneration, Micro-Cogeneration
2008, (Ottawa, Canada), 2008.
[2] Denny Beyer and Nick Kelly, Modeling the
Behaviour of Domestic Micro-
Cogeneration under Different Operating
Regimes and with Variable Thermal
Buffering,Energy Systmes Research Unit,
University of Strathclyde, Glasgow, UK.
2008.
[3] Hajo Ribberink and I. Beausoleil-Morrison,
Realistic Performance Forecast for Stirling
Engine Residential Cogeneration Systems
Applied in Single Detached Houses in
Canada. CANMET Energy Technology
Centre, Natural Recourses Canada,
Canada, Ontario. 2008.
[4] I. Beausoleil-Morrison, An Experimental
and Simulation-Based Investigation of the
Performance of Small-Scale Fuel Cell and
Combustion-Based Cogeneration Devices
Serving Residential Buildings Energy
Conservation in Buildings – Final Report
of Annex 42, Government of Canada,
2008.
[5] R. Moffat, Describing the Uncertaint in
Experimental Results, Experimental
Thermal and Fluid Science 1 (1998), 3-17.
Figure 8: Hydraulic layout of the test bench
 Table 13: Efficiency Uncertainty for the Cogeneration Device
10 Second Sampling 30 Second Sampling 60 Second Sampling
Power Power
(kWe) (kWt) Biasel Biasth Biastotal Biasel Biasth Biastotal Biasel Biasth Biastotal
(%) (%) (%) (%) (%) (%) (%) (%) (%)
1 6.25 5.44 3.36 6.39 2.06 2.82 3.49 1.38 2.70 3.03
2 7.88 4.40 2.95 5.30 1.82 2.49 3.08 1.36 2.38 2.74
3 9.15 3.74 2.62 4.56 1.70 2.23 2.80 1.38 2.14 2.55
4 10.26 3.31 2.38 4.07 1.66 2.04 2.63 1.43 1.97 2.43
5 10.96 3.07 2.21 3.78 1.67 1.92 2.54 1.49 1.85 2.38
6 12.00 2.84 2.09 3.53 1.67 1.82 2.47 1.53 1.76 2.33