P14029: Robotic Fish
/public/

Subsystem Level Build & Test

Table of Contents

Subsystem Level Build & Test

Construction of the individual parts required for the subsystems began during the Intersession period between MSDI and MSDII.

The BOM was updated to reflect any changes or additional parts throughout the construction, which can be seen here.

Main Head Box

The acrylic for the head "box" was cut out in the Brinkman Lab with John Bonzo and his watercutting table. The first step of this process was creating DXF files of the individual parts. A DXF file is simply a sketch, converting any flat surface into a line drawing that the watercutter's software can determine the cutting path from. The part outlines were drawn onto the stock pieces, and a DXF file was made of the stock piece with the part outlines in it. This was to visualize how the pieces were going to be cut out, in order to accelerate waterjetting and get the most from our materials.

Watercutting acrylic
Acrylic prior to cutting
Acrylic after cutting

The pieces of the box were then assembled with silicone caulk and the parts placed inside. The caulk was used because only a temporary bond was desired, as the purpose was to make sure there was sufficient room, and that the components were placed exactly as desired before the holes were drilled in the box walls. The extra space was also recorded, so that the floor of the box could be moved up or down in order to change the internal air volume, and hence the buoyancy characteristics, if desired.

Initial, temporary assembly with silicone caulk
Extra space in box assembly
Extra space in box assembly

Muscles

Several ideas were considered for a tensioning system to adjust the amount of slack in the fishing line connecting the muscles and the points being actuated, and the chosen concept was one adapted from our sister team, P14253, the Boeing Underwater Arm Team. The idea was to place a bolt with a hole in it crossways in one of the muscle ends, and rotating the bolt would adjust the tension in the line. The bolt was kept from rotating during operation by a slot on the side of the part that the bolt head fits into. The finished part is shown below. The muscle tubing and mesh is hose clamped around the small end of this piece, and the bolt goes through the hole in the large end.

Muscle end plug and tensioning device
Extra space in box assembly

Tail Subsystem

The tail subsystem was partially built during MSD1 for the proof of concept testing, as a 2-jointed model. This was expanded to a 3-jointed model, a tailfin was cut out of laminated paper, and the original tail testrig was modified to receive six McKibben muscles.

Tail and Muscles
Tail subsystem
Six muscles on test rig

Once bolted onto the testrig, the tail was found to be able to rotate slightly in the vertical plane. The cause was that the joints between tail segments were able to act like 4-bar linkages until they contacted adjacent pieces, permitting considerable slop. The cylindrical black pieces shown below were originally designed to make the fishing line pass through the axis of the joint, in order to keep muscles from affecting joints that it wasn't supposed to actuate, but when implemented, also eliminated the slop and the unwanted degree of freedom.

Tail Joint Stiffener
Black nylon joint stiffener

Electronics & Fluid Subsystems

Most of the circuit board components came unassembled from SparkFun and required soldering miniature circuit boards, header pins, and components together. The pieces were then tested at the component level, making sure they were functional and that the Arduino Mega 2560 was capable of interacting with them. From there, pieces were integrated together until the major components of the electronics subsystem were all working in unison to turn the pump and solenoids on and off.

Solenoids and Pump Control Subsystem Assembly and Testing
public/Electronics_unassembled.png
public/Powderdriver_1.png
public/Powderdriver_2.png
public/Pumprelay_1.png
public/Pumprelay_2.png
public/Mega2.png

Integrated Testing of Fluids & Electronics Subsystems

Being that unprotected electronics were being used to control pumps, with high pressure water lines right next to them, additional safeguards had to be taken to prevent damage in the event of a water leak or line blowing off. The pump and solenoids were first secured to a piece of plywood in order to keep things from vibrating and moving around, and the Arduino and Power Driver Shield, relay, power switch, and other circuit components were placed inside a plastic enclosure with a makeshift lid. The evolution of the waterproofing improvements and the final testing setup are shown below.

Integrated Testing of Fluid & Electronics Subsystems
Waterproof box take 1
Waterproof box take 1
1st
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing
Testing

Integrated Tail, Electronics, and Fluid Subsystems Testing

The culmination of the previous testing was using the electronics subsystem to control the solenoids and pump and move the tail. The pictures below show the system, and videos of it are shown further down under Subsystem Demo.

Integrated Testing of Fluid System, Electronics, and Tail Subsystem
Testing
Testing
Testing

Heat & System Runtime Testing

The last set of tests conducted during this second phase of MSDII were aimed at making sure that heat buildup would not be an issue during operation. This was done by putting all of the heat generating components in the box and sealing it shut. The lid couldn't be installed for the test, as there would have been no way for the fluid lines to pass through the box, and drilling the holes in the walls couldn't be done until after heat testing. This was in case heat management (such as a heat sink, fan, or aluminum heat plate) needed to be incorporated into the design, which might require relocating some of the waterline penetrations. As a result, the top of the box was sealed with ClingWrap to prevent air circulation out of the box.

The test was completed with the tail in the water but with the box in the air. This was very much a worst case condition, as heat dissipation out of the box would be much lower in air than submerged in the cool water. This round of testing also served as the maximum runtime test. The system had an accumulated runtime of approximately an hour before this testing began, which lasted over an hour before the battery began to run low (temperature data isn't plotted at the beginning). This meant the 2 hour runtime easily met the minimum 15 minutes that was required, and agreed with the predicted worst-case runtime of 1.6 hours that was predicted before the batteries were purchased. The temperature as measured by the thermistor also flattened out as it approached 34C, well below the maximum allowable temperature. The datasheet may be viewed here and the resulting plot is below.

Integrated Testing of Fluid System, Electronics, and Tail Subsystem
Testing
Testing
Testing
Testing
Testing
Heat testing data
Testing

Subsystem Demo

http://youtu.be/9reZ9TjSB5Y Swim #1

Forward Motion

http://youtu.be/9mpRPKBxhCs Swim #2

Testing

http://youtu.be/OFK7Ubh2Q50 C-Turn #1

Testing

Skin Subsystem

The "skin" for the fish was an area that we knew experimentation would be necessary in order to ensure a an effective decision could be made. The original concept was to simply use a stretchy fabric to cover the fish and provide a somewhat "lifelike" appearance. Ideally, the skin will be one piece, with a Velcro seal on the underside where its presence is least noticeable. In order to avoid ripples or sloppy appearance, the fabric would have to exhibit appropriate strain behavior without failing or restricting the motion of the fish. We expanded on this further, since fabric doesn't exhibit the same tactile properties as one would expect with skin, we decided to treat the fabric with something to give it more "body". The scope of the experiment was combinations of two different fabric compositions, and three different rubber coatings.

The two types of fabric available locally that provided the greatest strain were both spandex blends. The green fabric, samples #1 and #2, is an 82% lycra 18% spandex blend. The blue fabric, samples #3 and #4, is a 90% polyester, 10% spandex blend. Initial impressions were that, while both exhibit decent strain, the lycra/spandex blend definitely strains more, as would be expected with the increased spandex content. The fabric was crafted into 3"x6" samples for the sake of this experiment.

The three types of rubber coatings examined were: Ultraflex, Clearflex, and Ecoflex, all of which are used in either special effects or taxidermy. Ultraflex is a thermoplastic rubber, Clearflex is 2 part urethane based rubber, and Ecoflex is a 2 part silicone based rubber.

The Ecoflex and Clearlex samples were very easy to construct, simply mix the 2 part solutions at the manufacturer recommended ratio, and dip the fabric in. The samples were then laid flat to dry on a piece of scrap lumber. The Ultraflex, on the other hand, is not a mixture, but a thermoplastic which is solid at room temperature. Fort this reason, a hot-plate and aluminum mold were used to bring the material to its melting point to liquify it prior to application. This ended up causing severe problems on the first sample, the lycra/spandex blend. Not only was it hard to mix the fabric in while maintaining the temperature to avoid premature congealing, but due to the relatively high melting point, the fabric ended up melting and getting scorched as well. Due to the catastrophic failure and difficulty involved, we didn't even attempt to apply this solution to the polyester blend.

Skin fabrication
Testing
Testing
Testing
Testing

Next, an unadulterated sample of each material was laid flat, using a tape measurer for reference, and baseline strain behavior for each was characterized. This was done by stretching the material to the maximum strain possible by hand and recording the extended length compared to the original length so that strain could be given on the conventional (in/in) scale. While the method may not necessarily be the most accurate, it more than suffices for the purpose of this experiment which is simply selecting which best suits our needs. The uncoated 90% Polyester / 10% Spandex had a strain of ~0.5 [in/in or 50%, and the uncoated 82% Lycra / 18% Spandex had a strain of ~0.9 [in/in] or 90%.

Strain testing of uncoated fabric
Testing
Testing

Finally, the samples, which cured overnight, were examined using the same method. It was clear that the urethane based clearflex samples wouldn't work for our purposes by simply handling them. It was apparent that both the lycra and polyester versions had lost most of their elastic characteristics due to the treatment, rendering them moderately stiff and useless for our purposes. However, the silicone based Ecoflex samples faired extremely well, and both samples exhibited "skin-like" qualities to the touch, a large improvement on the untreated sample. The "body" of the material closely resembled skin in weight, and additionally, the tactile behavior mimics that of skin quite nicely. Tests for the strain characterization on both Ecoflex samples are shown below. The Ecoflex treated 90% Polyester / 10% Spandex achieved ~0.5 [in/in] or 50% strain, and the Ecoflex treated 82% Lycra / 18% Spandex achieve ~0.75 [in/in] or 75% strain.

Strain testing of rubber coated fabric
Testing
Testing

The polyester/Spandex sample didn't seem to lose any of it's strain characteristics, staying close to 50% before and after, which would make sense if the material itself were the limiting strain factor. The Lycra/Spandex, on the other hand, had strain reduced from 90% to 75%. These results make sense if we assume that the Ecoflex was the limiting strain factor, having an approximate maximum strain of 75% for the force applied by hand. This would mesh well with the results for the polyester as well, since the strain of the rubber would be greater than that of the polyester/spandex blend, it makes sense that maximum strain wasn't reduced by application of the rubber coating.

Additionally, we had concerns about the fabric and rubber finish separating due to different strain characteristics. If there was a large enough variance between the stiffnesses of the coating and the fabric itself, it would lead to a stress between the materials at the mating surface potentially inducing separation. Thankfully, none of this behavior was observed, even at the maximum strain (that could be achieved by hand).

In conclusion, the results of the test clearly demonstrated that both the Clearflex and Ultraflex were unsuitable for our application. The Ultraflex was incompatible with the fabric, and the Clearflex was too stiff. Strain tests demonstrate that if we combine the Lycra/Spandex material with the Ecoflex treatment, we will achieve the desired material characteristics, which maximize strain capabilities as well as "lifelike" feel.

Table of Contents MSD I Home

Project Definition Systems Design Subsystem Design Detailed Design

Table of Contents MSD II Home

Subsystem Level Prep Subsystem Level Build & Test Subsystem & System Level Build, Test, Integrate Systems Level Build, Test, Integrate Verification & Demonstration of Results
MSD - The Postseason