P17341: Harris Near-Zero CTE Test Apparatus
/public/

Detailed Design

Table of Contents

Team Vision for Detailed Design Phase

The following table shows the expected tasks to be competed in preparation for the Detailed Design Review, and the team member(s) responsible. Completion is indicated in the furthest column. Many tasks are in a state of in-progress completion, as final satisfactory completion will only be reached toward the end of MSDII.

Detailed Design phase goals

Detailed Design phase goals

Progress Report

Week 15 goal completion

Week 15 goal completion

Individual progress reports can be found here.

Bill of Material (BOM)

The Bill of Materials provides a way for the team to track acquisition of all physical components of the final test fixture. This includes whether or not the components have been fully designed, specified, and purchased, and what supplier will be used to obtain the final component. The BOM also includes non-physical components such as Arduino programs and error analysis documentation.

Example section from Bill of Materials

Example section from Bill of Materials

Due to its size, only a snapshot of the BOM is shown above. The full document can be accessed here.

Error Analysis

The error analysis approach was started with a general, team-wide brainstorming in order to identify possible sources of error, what can be done to quantify them, and whether or not they will be significant in the final analysis.

Sources of error:

The original three error sources were also related back to individual components in order to show their overall effects.

Comonent error sensitivity breakdown

Comonent error sensitivity breakdown

These two sources were combined and streamlined into a more formal document, with sensitivity values, the known load temperature, and the resulting error allocation.

Error allocation

Error allocation

The Root Sum of Squares (RSS) Approach was used to combine the error sources and determine the largest contributor to the final error. The RSS approach is employed to account for the low likelihood of all dimensions occurring at their extreme limits simultaneously. The sum of squares is a mathematical treatment of the data to facilitate the legitimate addition of measures of variability. The RSS method is used to add up tolerance stacks when more than two components are assembled together. Often these assemblies must then fit into another assembly.

The RSS method is used to determine if a functional fit is going to occur between the mating assemblies. A probability of assembly success or failure can be calculated using the standard Z transform for population data, or by using the student-t transform for sample data. The final comparison between error allocations is given below.

Error analysis table

Error analysis table

Link to the live error analysis document here. This document also includes descriptions of how error sources were calculated / determined.

We believe that this document captures all major sources of error. However, it may need to be expanded in regards to breaking some of these error sources down into smaller contributors in order to measure them accurately. Additionally, some of the error sources presented may end up being insignificant to the final total, or they may cancel out or otherwise be accoutned for by changing design. For instance, noise in the sensor singnal can be attenuated via an analog RC filter. Going forward in MSDII, a large focus will be on determining the magnitude of the error sensitivities presented above. While some may be determined theoretically, the majority must be found experimentally.

Thermal Analysis

Structure Simulation

Motivation: Simulate structure expansion due to thermal effects to determine material selection of components and quantify possible error induced during system operation.

Procedure: Import .stp files of prototype assembly into ANSYS Workbench, subject to a uniform temperature change of 50 deg F, and find the change in height between the top and bottom plates, using different combinations of materials for the plates and rods.

Results: Material selection for the supporting rod proved to have much more effect on the height change than the plates.

Deflection results for various material combinations

Deflection results for various material combinations

ANSYS Simulation with exaggerated scale

ANSYS Simulation with exaggerated scale

Conclusions: In order to minimize deflection, Invar should be utilized for the rod supports, and steel for the top and bottom plates.

Next Steps: Determine ways to minimize or counteract thermal effects on the structure.

Clamp Simulation

Motivation: Simulate clamp expansion due to thermal effects to determine material selection of components and quantify possible error induced during system operation.

Procedure: Import .step files of clamp assembly into ANSYS Workbench, subject to a uniform temperature change of 50 deg F, and find the vertical elongation of the clamp for each material.

Results: Clamp experienced a vertical expansion of 1.832e-4 in

ANSYS Simulation of coupon clamp

ANSYS Simulation of coupon clamp

Conclusions: Although the clamp will expand minimally when compared to the whole structure, it is much more likely to experience the same temperature range as the coupon.

Next Steps: Determine methods to counteract thermal expansion effects on the coupon clamp.

Heat Transfer Analysis

Motivation: In order to run a CTE experiment, a method for measuring and maintaining temperature must be established and understood.

Procedure: Use a lumped capacitance analysis of the coupon-shroud system, with a control volume drawn around the coupon. Neglect losses from the top and bottom of the shroud, as well as the losses from the shroud to the surroundings. Neglect effects of the clamp on the coupon. Assume that the system is at steady state, and that the shroud is providing uniform heating.

Results: Due to the shape of the coupon, the Biot number should be well under .1, showing that a lumped capacitance method is valid to show the behavior of the system.

Derivation of temperature expression

Derivation of temperature expression

Conclusions: For long experiment times (very slow rate of heating), the temperature difference between the shroud and coupon is small enough that the temperature can be measured directly from the shroud, provided that the shroud is providing uniform heating.

Next Steps: Determine methods for calculating overall heat transfer coefficient U, calculate dT/dt required for an acceptably small temperature difference, quantify what is meant by uniform heating, and determine the effects of the clamp on coupon temperature.

Mechanical Design

Final test facility assembly

Final test facility assembly

CAD model of assembling test facility

CAD model of assembling test facility

Clamp Prototyping

Motivation: Verify that the clamp design selected at the end of System Design phase is appropriate for the final maglev tester.

Procedure: All designs were mocked up in CAD using Autodesk Inventor. These were then 3D printed using ABS filament and assembled together with the required hardware to provide a prototype of the final design.

Results: Two clamp designs were selected for prototyping: the "movable plate" design and the "screw clamp" design. The screw clamp (below) is the more flexible design, featuring two rounded contact points that meet on either side of the sample. The distance between the points is adjusted and held secure by the nut and bolt running through the two arms. This design can accomodate a number of sample geometries and will always keep the same center of gravity, regardless of arm position.

3D model of Screw Clamp design

3D model of Screw Clamp design

The movable plate clamp (below) features two parallel plates that hold either side of the sample when tightened by a pair of bolts, and features a locating hole to align with a proposed matching hole in the coupon for repeatable loading. This design can securely hold a thin, flat coupon, but the thickness of the coupon must be fixed, since the design is asymmetric and moving the plate too far to accomodate a larger sample will unbalance the assembly.

3D model of Movable Plate Clamp design

3D model of Movable Plate Clamp design

The screw clamp prototype was 3D printed and assembled together with the required hardware. The clamp fucntions as intended, and can hold samples of varying thickness. However, the clamp does not hold the sample in place securely. With only one point of contact on the rounded edge, the sample has a tendency to slide from side to side and not hang straight down. This is troublesome for holding the sample steady above the capacitive sensor in the final design.

3D Printed Screw clamp

3D Printed Screw clamp

The movable plate clamp was also 3D printed to assess its effectiveness. While it can only accomodate samples of one thickness without becoming unbalanced, the sample is held very securely once the two bolts are tightened. The locating hole ensures that the sample is loaded the same each time, and the design constrains 5 of the 6 degrees of freedom of movement. The sample can still rotate around the axis of the locating hole, but the tight clamp ensures that it takes some force in order to accomplish this, and it is unlikely that it will occur spontaneously during CTE test operation. The design can also be changed in order to prevent this movement via a groove in the clamp that fits the sample coupon profile.

3D Printed Movable Plate clamp

3D Printed Movable Plate clamp

Conclusions: The movable plate clamp is superior in terms of sample stability, though it loses out in terms of accomodating multiple sample sizes. However, the regular size of the test coupon combined with the need for stability in order to make an accurate measurement make the movable plate clamp the preferred choice.

Next Steps: Incorporate magnet into clamp design and test it in prototype maglev apparatus. Perform thermal and mechanical analysis to ensure that clamp does not deform unduly under temperature variation.

Feasibility Testing

Hall Effect Sensor Calibration

Motivation: Acquire a calibration curve for the feedback hall sensor. The system reads a digital value based on magnetic field strength (in Gauss), converts it to a voltage and the voltage can be related to a distance. As these curves can vary depending on the device an experimental calibration is desired. Additionally, the sensor transfer characeristic is Gauss to voltage, which is a linear relationship. However, the relationship between Gauss and distance is nonlinear, and this calibration will serve to characterize it. This will allow conversion of sensor parameters in Gauss to distance values, which can be used in the error analysis.
Sensor transfer curve from Honeywell SS495A1 datasheet

Sensor transfer curve from Honeywell SS495A1 datasheet

Procedure: Using a ruler, measure the reading on of the hall sensor using Matlab/Simulink at 0.25 cm increments from the closes location to the hall sensor. Once data points have been collected they can be plotted and the calibration cuvee can be acquired.

Results: Multiple trials were run which all displayed consistent results. The calibration curve depicts a read value and corresponds that to a distance.
Hall sensor calibration

Hall sensor calibration

Conclusions: The experiment was successful, the calibration curve for the specific hall sensor in use was able to be obtained. We can now relate a reading to position/distance.

Next Steps: Find how a magnet will affect the hall sensor readings.

Biased Hall Effect Sensor Calibration

Motivation: Using the already acquired Hall Effect sensor data, we want to find how much the electromagnet will affect the reading. This is important for the control of the system and for the error analysis of the system.

Procedure: Run a very similar experiment as the Hall effect calibration, but bias the sensor with the electromagnet.

Results: The results show that the biased calibration curve has the same trends as the non biased calibration curve. The only difference is that the biased curve is offset by 0.733V biased low compared to the non-biased case. This means that depending if the magnet is on or off, the Arduino will be reading one value that can represent 2 different positions - not good.
Hall sensor calibration

Hall sensor calibration

Hall sensor calibration

Hall sensor calibration

Conclusion: The experiment was successful, we were able to obtain a biased calibration curve using the hall sensor, and identify the difference between it and the non-biased case.

Next Step: The next step will be to figure out a way such that the controller will be able to read 1 value that corresponds to one distance. In other words, a way to decipher the position that is not dependent on the state of the magnet.

Sliding Mode Control Derivation

Motivation: Since the current magnetic levitation system is unstable, a new control system was derived. Sliding mode control is a good candidate controller because variable values do not need to be known, simply the variable range. Sliding mode control is also good for this application because it can handle nonlinear system proficiently. An accurate plant model is also useful to know for future control implementations.

Procedure: Research accurate plant model. Form a state space model of the plant. Define a sliding surface, and solve for the input. Generate a candidate form for the input. Form a sliding condition and verify the candidate from is correct. Then apply the sliding condition to find the gain K.

Results: The derivation is of the correct form. The derivation allows for the use of a sign gain, a constant boundary layer and varying boundary layer. The hold back to this is that the control system does not take into account the PMW signal that is driving the hardware in implementation. The actual hardware implementation that is being used is very difficult to model in simulation. If the physical system can be driven similarly to the simulations the simulation to hardware transition will be smoother. Another limitation to this control scheme is that the system needs to be completely observable which can be a difficult task to accomplish.

For brevity on this page, the full derivation is omitted, but can be found here.

Block diagram representation of sliding mode controller

Block diagram representation of sliding mode controller

Conclusions: The derivation of the control system was successful but also revealed software to hardware implementation problems.

Next Steps: Research methods to implement the hardware closer to how the simulations are ran.

Hall Effect Sensor Noise Reduction

Motivation: Filter the Hall sensor output such that the signal is smoother for the Arduino to read. This may help the control of the system.

Procedure: Experiment with multiple analog filters such that the noise coming from the Hall sensor is minimized. Extract data using Simulink and compare results. In order to compare the results of the filtered and non-filtered data, simply pull out the filter from the hardware half way through the test.

Results: The filters overall had a positive effect with reducing the amount of spikes within the Hall sensor noise.

Figure 1- 10uf capacitor filter

Figure 1- 10uf capacitor filter

Figure 2- 100uf capacitor filter

Figure 2- 100uf capacitor filter

Conclusion: The experiment was successful. We were able to determine that the filter has an overall positive effect on the system. The output signal has less noise than the non-filtered output.

Next Step: experiment and find the optimal filter.

Maglev Tester Update

Motivation: The prototype maglev tester utilized in previous tests did not levitate the test magnet stably. Without a reliable plant model with which to determine an optimal controller analytically, the control scheme parameters had to be tuned by hand. Minor modifications were made to the tester in order to more reliably levitate a test component, since greater repeatibility would allow for more accurate optimization of the controller.

Procedure: Communication between EE team members performing testing and ME team members with access to material resulted in the fabrication/sourcing of the necessary components to update the protoype test facility, without a complete overhaul.

Updated test facility

Updated test facility

Results: First, the top plate design was altered. The previous design was very simple, with a through hole which the electromagnet was press-fit into, with the Hall sensor attached to the bottom of the plate. This resulted in the magnet being at a variable distance to the Hall sensor, and the Hall sensor was vulnerable to impacts from the test sample if it got too close to the magnet and attached itself to it. This was resolved via an updated top plate that featured a built-in channel to hold the Hall sensor, and a circular hole to hold the electromagnet. Rather than pass completely through the top plate, this hole allows the electromagnet to be butted to the plate, holding it at a constant distance from the Hall sensor. The channel for the Hall sensor protects it from impacts and holds it more steady. While this change does constrain the sensor and magnet within the plate, ABS plastic has a similar permeability to air and this does not affect the magnetic operation.

The next change involved the addition of a rudimentary heat sink to the magnet in the form of an aluminum block. Long-term testing with the magnet revealed that its field strength decreased as it heated up, rendering repeated feasibility testing difficult without allowing it to cool first. The aluminum serves as a thermal mass that draws heat away from the magnet through conduction, allowing the magnet to remain in operation for longer periods of time. However, this aluminum does not transfer heat to the surroundings at an appreciable rate, and thus does not truly function as a passive cooling system, merely a postponing of the inevitable. While a heat management system for the magnet will be included in the final design, it must be designed in order to radiate heat away effectively under the vacuum test conditions of the Harris chamber.

The final change was the addition of a long bolt to the levitation test piece. The reasons for this are many. First, the increase in length provides greater repeatibility in tests. When the levitation becomes unstable and the test piece falls to the bottom of the tester, it can easily be picked back up and placed in the active region without significant lateral motion. Second, the long sample with a permanent magnet on top provides a better model of the final test coupon, which is long but narrow with a permanent magnet on the clamp to ease levitation. Finally, the increase in length results in a lower center of gravity, helping to balance the test piece during levitation. If the test piece is picked up at a slight angle, the weight of the bolt acting through the center of gravity serves to return it to vertical, increasing the robustness of the system to disturbances.

As a recap of demonstrations, here is the simple electromagnet constructed before the Systems Design Review:

Electromagnet Feasibility Test

Electromagnet Feasibility Test

The demonstration of the electromagnet switching on and off when integrated with the hall sensor, as shown in the Preliminary Detailed Design Review:

Magnet controlled by sensor input

Magnet controlled by sensor input

And finally, the prototype tester demonstrating true magnetic levitation.

Video of magnetic levitation

Video of magnetic levitation

Conclusions: The modifications to the maglev tester provide greater ease in testing and allow the test piece to remain stable for longer, allowing for finer tuning of the control system. However, the system is still not stable, and further modifications and controller tuning are required.

Next Steps: Continue tuning maglev control system. Incorporate improved aspects of prototype design into final design.

Test Plans

Due to the nature of the design, testing will mostly be functional - whether or not the maglev system can hold the sample steady and produce a viable CTE reading. However, a significant amount of testing will be performed in the error analysis. This is to quantify the errors and uncertainty values that cannot be determined directly from component data sheets, or from engineering analysis.

Creating the appropriate test plans and carrying them out will be a major focus in MSDII. While no test plans have been created yet, the MSD test plan template will be used to devise the plans and ensure the following criteria are met:

1. Define what will be tested (and what will not be tested)

2. Define how tests will be performed: equipment and materials needed, test configurations and procedures, pass/fail criteria)

3. Show team member responsibilities and the approval process

4. Document risks and contingencies

5. The plans will be understandable by an engineer not familiar with the project

Sample Test Plan Document

Sample Test Plan Document

Risk Assessment

Risks were re-assessed following Preliminary Detailed Design review. Risks were updated to remove unnecessary considerations and more closely follow the tasks actually being carried out by the team members.
Risk Management list, updated from previous design review

Risk Management list, updated from previous design review

Risk Explanation

Risk Explanation

Link to the live risk assessment document here. For a link to the previous (now outdated) risk management document presented at previous design reviews, click here.

Plans for next phase

Tasks and Plans for Phase 1 of MSDII

Tasks and Plans for Phase 1 of MSDII

The plans mentioned above will be worked on in the first phase of MSDII. Preparation for the Gate Review next week ensures that the plans are prepared such that tasks may easily be resumed in August at the start of MSDII.


Home | Planning & Execution | Imagine RIT

Problem Definition | Systems Design | Preliminary Detailed Design | Detailed Design

Build & Test Prep | Subsystem Build & Test | Integrated System Build & Test | Integrated System Build & Test with Customer Demo | Customer Handoff & Final Project Documentation