Memo



Meeting Purpose: To review and receive feedback on design details

Meeting Date: 2/13/09

Meeting Location: 09-1129

Meeting time: 12 – 2 PM

Agenda:

|Time |Topic of Review |Team Member |

|12:00 - 12:10 |Project Introduction & Overview |Bryan |

|12:10 – 12:18 |Action Items from System Review |Andy |

|12:18 – 12:28 |Data Collection & Time Studies |Gabriela |

|12:28 – 12:45 |Power Unit Design |John |

|12:45 – 12:50 |Mass Flow/Head Loss Calculations |John |

|12:50 – 1:00 |Piping Layout & Design |Ken |

|1:00 – 1:15 |Electrical System Design |Jon |

|1:15 – 1:30 |Peak Power Control System |Ken |

|1:30 – 1:40 |Labview User Interface |Andy |

|1:40 – 1:50 |Feasibility Analysis |Gabriela |

|1:50 – 2:00 |Action Items & Wrap-Up |Bryan |

[pic]

[pic]

[pic]

Main Action Items from System Review

• Add Specs for Customer Need 9 (CN9) and Customer Need (CN11)

➢ CN9 = Temp measurement (is what I'm measuring what I think I'm measuring)

➢ CN9 Spec: A. viii. Taking measurement does not affect measurement

➢ Yes or No ( Yes, groove analysis refined and tested experimentally

➢ CN11 = Measure mean fluid temperature along power unit

➢ CN11 Spec: A. ix. Measure temperature between zones

➢ # of Measurements ( # of zones plus 1

Refined Groove Analysis

Geometry Representation & Boundary Conditions (Figure 1)

-R2 represents the Thermoelectric Module (k=6W/mK)

-R1 represents the proposed groove filled with a thermal paste (k=2.8 W/mK) [This represents a worse case scenario because the thermocouple would actually be more beneficial to have modeled than the paste. Due to the uncertainty of the actual effects of the paste, the groove was assumed to be entirely filled with paste.]

-CO1 represents the upper portion of the aluminum test unit (k=160W/mK)

-Tc = 30°C or 303K

-Th = 200°C or 473K

-All other boundaries are insulated

Mesh (Figure 2)

-There are 30720 elements

FEA Solution for Previous Model (Figure 3)

-As can be seen, significant issues existed due to the previous model’s boundary conditions. These were changed to the current boundary conditions. The new model also takes into account the actual dimensions of our proposed system.

FEA Solution for Current Model (Figure 4)

[pic]

-The average temperature within the groove is 455 Kelvin. This agrees with the average boundary integral temperature from Figure 5 with only a minimal difference. Most of the difference can be credited to the groove being assumed to be entirely thermal paste rather than the combination of the more thermally conductive thermocouple and paste.

Boundary Integral (Figure 5)

[pic]

Solution – Closer Look (Figure 6)

- The absolute worse case temperature seen is the 438 Kelvin inside the groove. The average temperature throughout the groove is 455 Kelvin which agrees with the boundary integral value of 458 Kelvin meaning only a minimal difference. This being a worse case scenario analysis allows us to believe that the temperature read inside the groove will read close or exactly the temperature that the hot side of the thermoelectric module is seeing.

Experimental Results of Simple Test (Figure 7)

-Test stand was used to test individual thermoelectric module under three different scenarios

-Scenario 1: Thermocouple in the groove in the middle of the hole in the thermoelectric

-Scenario 2: Thermocouple in the groove offset from the hole in the thermoelectric by 9mm

-Scenario 3: No thermocouple and no groove

|Results |Pressure (psi) |Th (degC) |Tc (degC) |Pmax (W) |

|Scenario 1 |130 |200.06 |29.77 |4.55 |

|Scenario 2 |130 |200.12 |29.03 |4.45 |

|Scenario 3 |130 |200.16 |----- |4.54 |

-These results provide evidence that the groove has no known effect on the operation of the system.

Conclusions

-The groove is an acceptable design for taking measurements of temperature on both sides of the thermoelectric.

-More importantly, the groove will not significantly effect the transfer of heat through the aluminum plate to the hot side of the thermoelectric. A relatively even distribution of heat allows for the temperature to be within measuring capabilities of expected values.

Data Collection

The goal for performing tests on the current thermoelectric power unit was to gather enough data to perform statistical analysis to see how reliable and repeatable the data is. With this information we could get a better idea of where the system stands and where it needs improvement.

Four tests were performed, the first two under the same exact conditions (i.e. same assembly, same temperature and flow), and for the second two the unit was disassembled and assembled one more time, to be able to measure the variability that is accountable for the assembly process.

During both assemblies we also performed time studies to understand the process better and to be able to make design improvements to decrease assembly time. The results of the time studies, along with concepts to improve the times are displayed in the following chart:

| | |Test 1 |Test 2 |Improvement |

|Cleaning |Cleaning |1:02 |1:50 |  |

|Bottom Assembly |Applying Thermal Paste to Modules: |  |  |  |

| |Three Modules, 1 side |4:12 |3:46 |Considered thermal pads, not feasible |

| |Three Modules, 2nd side |3:38 |3:16 |Considered thermal pads, not feasible |

| |Placing Insulation |1:04 |1:02 |Marked Places for Modules for easy placement |

| |Setting on Bottom Plate |1:22 |0:52 |Bolts attached to bottom plate |

| |Sub Total |10:16 |8:56 |  |

|Top Assembly |Applying Thermal Paste to Modules: |  |  |  |

| |Three Modules, 1 side |3:51 |3:04 |Considered thermal pads, not feasible |

| |Three Modules, 2nd side |2:47 |2:21 |Considered thermal pads, not feasible |

| |Placing Insulation |0:38 |0:44 |Marked Places for Modules for easy placement |

| |Setting Top Plate |0:28 |0:59 |  |

| |Sub Total |8:44 |7:08 |  |

|Final Assembly |Placing Washers and Nuts |1:30 |1:15 |  |

| |Tightening |3:01 |4:21 |Better Tool Clearance |

| |Outside Insulation |1:26 |1:59 |More space to insert, less pieces |

| |Thermocouples |6:50 |5:54 |Embeded in Power Unit |

| |Sub Total |12:47 |13:29 |  |

|Set-up in Stand |Set on Jacks and Lift |1:47 |2:51 |Block with Appropriate Height in Test Stand |

| |Attaching to Stand |10:29 |7:14 |Better Tool Clearance |

| |Plugging into DAQ |6:54 |6:01 |Color Coding |

| |Sub Total |18:10 |16:06 |  |

|Total |  |49.57 |45.39 |  |

Issues/Solutions on Data Processing

The processing and cleaning up of the data that was collected presented many issues, which also gave place to ideas for improvement.

The data is stored from LabView to a text file that is compatible with Microsoft Excel, but the data appears completely raw, with no labels or headers. This can make the cleanup process slow and allows for many user errors. To improve on this, the output will be formatted before storing into a file, so that the final result contains headers, valuable decimal points, etc.

Another issue noted was that there is no control that prevents the system from taking invalid data, which is usually seen after the test is over, invalidating the stored data. To avoid this, data constrains will be implemented in LabView, with indicators when the data is out of range.

Statistical Analysis

Two tests were performed on the data to try to assess the variability of the different tests: a two-sample t-test and a z-test, using hypothesis testing.

The two-sample t-test is one of the most commonly used hypothesis tests in Six Sigma work. It is applied to compare whether the average difference between two groups is really significant or if it is due instead to random chance. To perform this test, both samples must be normally distributed, so a plot of each measurement was performed. The p-value being less than 0.05 shows that the distributions are normal. The plots also showed a normal curve, as seen in the graph below:

[pic]

The next step was to perform a t-test in MiniTab comparing the measurements made in the two different assemblies. The t-test outputs two important values, a t-statistic and a p-value. If the p-value is greater than 0.05 it is concluded that there is no difference in both samples. Unfortunately, none of the temperature measurements passed this test, and only the voltages and pressures passed it.

The hypothesis tested was Ho: x1-x2=0. (where x represents mean of the population) Namely stating that the mean of the first sample is equal to the mean of the second. To test whether this hypothesis is true, the t-test requires the finding of the t-statistic through the following formula:

where [pic]

The t-value must be compared with the critical value of the t distribution. The critical t-value marks the threshold that – if it is exceeded – leads to the conclusion that the difference between the observed sample mean and the hypothesized population mean is large enough to reject H0. The critical t-value equals the value whose probability of occurrence is less or equal to 5 percent. From the t-distribution tables, one can find that the critical value of t is +/- 2.093.

The t-statistic confirmed that both measurements system are statistically different. This is clearly seen in a box-plot of one of the temperature measurements for both samples:

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

Most aggressive fin:

[pic]

Air flow (m/s)

1.0 2.0 3.0 4.1 5.1

| |

| |

| |

| |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

|[pic] |

| |

Thermal resistance K/w

0.75

.50

.25

0

Rt = 0.39 K/w when cut to the right width and height 4”x1.939”. This is about 20% of the thermal resistance through the module itself, so about 2/3 of the temperature drop should be across the module.

Middle case:

[pic]

Rt = 0.86 K/w at size 4” x 1.939”

Wimpy case:

[pic]

Rt = 1.06 K/w at size. About 1/3 of the temperature drop will be across the module.

Also investigating custom folded fins:

[pic] [pic]

For best case only. Cost, geometric compatibility and lead time will be the main considerations in final heat sink selection.

[pic]

[pic]

Piping Layout

[pic]

[pic]

[pic]

Width = 45” Length = 98”

[pic]

Height = 16”

Peak Power Impedance Matching

&Automated Measurements

The current test stand requires manual measurements of Voc and IL.

[pic]

Original Circuit with Manual Impedance Matching

3

By implementing a shunt resistor (RShunt) we can automate the measurement of the load current.

[pic] (Eq.1)

[pic]

Impedance Matching Circuit for Peak Power

To complete the automation of the impedance matching, for peak power, a power MOSFET can be implemented, where [pic] is the MOSFET resistance, controlled by the software control system.

[pic] (Eq.2)

[pic] (Eq.3)

[pic] (Eq.4)

Peak Power Control System

1) Plant model

[pic]

[pic]

α = .05, ΔT=150, Rint = 4

A controller is needed to control Rload such that P is maximum for any value of α, ΔT, and Rint.

The maxima occurs at Rload = Rint, however Rint changes with temperature and is not well known.

So, we will take the derivative dP/dRload and design a controller to force it equal to zero.

[pic]

[pic]

This equation has no time dependence and therefore cannot be used as plant model for a control system design. A damping term was added to account for the settling time of the system

[pic]

An optimization routine determined that a c value of .25 best corresponded to the open loop settling time of 3 seconds

The control system:

[pic]

the PID controller:

[pic]

The Mosfet converts the controller’s command voltage into a resistance:

[pic]

Notice, this equation has a singularity at Vcommand = Vt

Vt ranges between 2 and 4 volts, which is well within our domain

This singularity causes the control system to seek the dp/dRload = 0 condition at R = ∞

Passing the PID output though the following function corrects the problem:

[pic]

Optimized Response

Conditions: α = .05, ΔT=150, Rint = 4

Numerically optimized gains: Kp = 0.3.62, Ki = 16.21 Kd = -0.87

Initial condition = Rload = 1 ohm

[pic]

[pic]

[pic]

Robustness

α, ΔT, and Rint are all highly variable, however the control gains can only be optimized at one operating condition.

It is therefore important to understand how the gains optimized under the previous operating condition behave under a different condition.

Conditions: α = .05, ΔT=75, Rint = 3

Same control gains

Initial condition = Rload = .5 ohms

[pic]

[pic]

max power = 1.17 watts

[pic]

The system still reaches steady state in less than 2 seconds, with minimal oscillations demonstrating that the gains can be optimized for a single operating point.

Data Acquisition

|Needs |  |

|CN4 |Automate electric data collection |

|CN6 |Reduce setup time |

|CN7 |Improve data display to assess problems during experiment |

|CN10 |Operate at peak power for duration of experiment |

|CN16 |Steady State Indication |

|Total Sensors (Inputs to DAQ) |  |  |  |

|Thermocouples |On Modules |48 |  |

|  |Before/Inbetween/After Zones |5 |  |

|  |Inlet/Outlet of Cold Exhaust |4 |  |

|  |Heater Inlet |1 |  |

|Mass Flows |Air |1 |  |

|  |Water |1 |  |

|Compression Force |Flexi-Force Sensors |8 |  |

|Change in Air Pressure |Pressure Transducers |3 |  |

|Voltages |Load Voltages |4 |  |

|  |Shunt Voltages |4 |  |

|  |  |79 |Total |

|Air Configurations Sensor Needs |  |  |

|Thermocouples |On Modules |48 |  |

|  |Before/Inbetween/After Zones |5 |  |

|  |Inlet/Outlet of Cold Exhaust |4 |  |

|  |Heater Inlet |1 |  |

|Mass Flows |Air |1 |  |

|Compression Force |Flexi-Force Sensors |8 |  |

|Change in Air Pressure |Pressure Transducers |3 |  |

|Voltages |Load Voltages |4 |  |

|  |Shunt Voltages |4 |  |

|  |  |78 |Total |

|Water Configuration Sensor Needs |  |  |

|Thermocouples |On Modules |48 |  |

|  |Before/Inbetween/After Zones |5 |  |

|  |Heater Inlet |1 |  |

|Mass Flows |Air |1 |  |

|  |Water |1 |  |

|Compression Force |Flexi-Force Sensors |8 |  |

|Change in Air Pressure |Pressure Transducers |1 |  |

|Voltages |Load Voltages |4 |  |

|  |Shunt Voltages |4 |  |

|  |  |73 |Total |

Current System User Interface [pic][pic]

New System User Interface

Block Diagram

Interface

B. ii. Peak Thermoelectric Power Indication

B. iii. Steady State Indication

B. iv. Graphs of Calculated Values

[pic]

B. v. Real Time Feedback

• Graphs

• Indicators

B. vi. Automated Post Processing

• All data will be output in text format, preferably excel, with column headers to save in post processing time

B. vii. Automate All Measurements

• Previously, IL and Voc were measured manually using a multimeter

• IL is now measured through the shunt resistor

• Voc is no longer needed due to new MPPT tracking system

Associated BOM

|Description |Approximate Unit Price |Quantity |Cost |Supplier |Estimated Lead Time |

|Sensors and Modules |  |  |  |  |  |

|Melcor TE modules |$26.50 |40 |$1,060.00 |Melcor |2-3 weeks |

|Taihauxing TE modules |$30.00 |40 |$1,200.00 |Taihauxing |2-3 weeks |

|K type Thermocouples (1/16") |$22.00 |72 |$1,584.00 |  |  |

|K type Thermocouples |$22.00 |3 |$66.00 |  |  |

|K type Thermocouples |$22.00 |4 |$88.00 |  |  |

|Thermocouple extension wire |$138.00 |1 |$138.00 |Omega |  |

|Pressure Transducers |$8.55 |2 |$17.10 |Freescale Semiconductor |  |

|Flexiforce sensors (sheet, 8 per sheet) |$110.00 |1 |$110.00 |  |  |

|Op-Amps |$0.34 |16 |$5.44 |Microchip |  |

Feasibility Study

Dresser-Rand follows an established format for gathering the required information that they need when assessing whether a new technology is feasible or not to be implemented. The template requires information on the goals, benefits, risks, possible markets, and other variables that are summarized in the following chart:

|Project Charter |

|Project Name |Thermoelectric Module for Large |Project Lead |Bryan McCormick |

| |Scale Systems | | |

|Date, Revision # | |Project Sponsor |Dresser-Rand: Paul Chilcott |

|Start Date |12/1/2008 |SDT Leader | |

|Completion Date |5/22/2009 |Idea Submitter |Dr. Robert Stevens |

|Current Phase |Detailed Design | | |

|Element |Description |Team Charter |

|Process Definition |The business process in which |This project is aimed at acquiring an improved understanding of the workings of |

| |opportunity exists. |thermoelectric (TE) devices as a means to recovering wasted energy from Dresser-Rand |

| | |turbo machinery exhausts in the form of usable power. |

|Strategic Goal/ |Describe the opportunity as it |A clear understanding of the workings, benefits and fallbacks of thermoelectric devices |

|Business Case |relates to strategic business |is important when deciding whether it is feasible to implement this technology in |

| |goals. |industry, and whether it is the best choice among alternative technologies. |

|Problem |State the significant issue(s) that|To better understand thermoelectrics and their potential applications in industry for |

|Statement |needs to be addressed or |energy recovery, it is important to have a reliable, flexible and efficient test stand to|

| |opportunities to pursue. |validate models. The current test stand requires long set-up times, does not provide |

| | |readily available data and has too many variables that compromise the reliability of the |

| | |acquired data. There are also different configurations that could be tested if the stand|

| | |allowed for more flexibility, which would produce a wider range of data to draw |

| | |conclusions from. |

|Benefits Impact ($) |What are the anticipated business |• Upgraded thermoelectric test rig utilizing interchangeable air-cooling on the cold side|

| |results and when would the results |and capabilities for different configurations. |

| |be realized? |• Upgraded data acquisition system. |

| | |• Improved Power Unit. |

| | |•Documented data on module performance for verification of theoretical results. |

| | |(All these deliverables are expected to be ready by 05/22/09) |

| |What is the preliminary budget |This project counts with a budget of $7500 |

| |estimate for the project cost? | |

|Scope/Boundaries |Describe the project's scope and |Activities within scope include: |

| |boundaries. Describe what is in |1. Implement air-cooling on cold side of thermo-electric array in addition to the |

| |and outside the scope. |water-cooling system. |

| | |2. Be able to experimentally validate thermoelectric system models and enable more |

| | |parameters to be explored. |

| | |4. Improve set-up and shut-down procedures to reduce assembly times. |

| | |3. Improve user interface and data acquisition to allow greater ease of use of the test |

| | |stand. |

| | |The scope excludes: |

| | |- Any study or design of the system that would eventually be implemented in industry. |

|Schedule/Milestones |What are the start and completion |Starting Date: 12/01/08 |

| |dates of the project? |System Level Design:01/02/09 |

| | |Detailed Design:01/30/09 |

| | |Project Completion:5/22/09 |

|Benefit to Customers |Who are the Customers, what benefit|Dresser-Rand: The results of experimentally testing valuable aspects of thermoelectric |

| |will they see and what are their |modules will provide insights on the feasibility of implementing this technology to |

| |most critical requirements? |provide energy recovery capabilities to Dresser-Rand's customers, which would further |

| | |increase the value of the products, specially in applications in remote locations. |

| | |RIT: This knowledge will help bring RIT to the forefront of this emerging technology. |

|Support Required (if any) |Will you need any special | |

| |capabilities, hardware, etc.? | |

|Key Stakeholders |Who has been identified as a Key |Primary Stakeholder: Dresser-Rand |

| |Stakeholder(s) for this project? |Secondary Stakeholder: RIT |

[pic]

-----------------------

T=465K

T=459K

T=438K

T=457K

T=458K

Th = 200°C or 473K

Tc = 30°C or 303K

[pic]

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download