Molding and Casting

Proprioceptive Soft Robot (GRACE)

This week is molding and casting week. The assignment was to design a mold around the stock and tooling we will be using, mill it (rough cut + three-axis finish cut), and use it to cast parts! Since Ozgun and I have a shared interest in soft and stretchable structures, we decided to collaborate this week to realize a soft robot that is augmented with sending element to enable proprioception!

We got inspired by this article recently published article in Science Robotics, where researchers developed a novel type of pleated pneumatic artificial muscle. The fabrication method they utilized was 3D printing the actuators, which allowed a wide variety of sizes. They have named the results GeometRy-based Actuators that Contract and Elongate (GRACE). The GRACEs consist of a single-material pleated membrane and do not need any strain-limiting elements. They can contract and extend by design, as described by a mathematical model, and can be realized at different dimensional scales and with different materials and mechanical performances, enabling a wide range of lifelike movements.

Although 3D printing seems to be one of the easiest methods fot fabricating such structures, we decided to challenge ourselves this week by replicating the structure using a two-part molding approach! Sometimes 3D printing, such as FDM and laser-based SLA consumes more time due to the serial nature of the process. However, molding and casting are not usually an issue since after milling the mold for one time, casting multiple elements will only be limited to the curing time of the material. Thus, each method has different pros and cons depending on the use and resource availability.

CAD Model

As shown in the left photo below, we started with the 2D sketches provided on the referenced paper as a starting point. It is worth mentioning that the paper explored multiple geometries and dimentions; we decided to go with the one that they mentioned performed better in contraction. This structure has narrow ellipses and deep pleats, which can largely expand when the GRACE is inflated.

Following the 2D sketch, Fusion 360 was used to generate the extruded 3D model. Since we are directly milling our negative mold, as seen below, our model consists of two parts. One is the bottom part which will form the outer structure of GRACE. The second is the top part which will form the inner structure of GRACE, ensuring that it will be a hollow shell.

Milled Mold

We tried to cast a variety of materials, including PDMS, Smooth-On EcoFlex 20A), 30A and 50A. We used Smooth-On SIL-poxy rubber silicone adhesive to bond the two casted parts together.

Initial Results

We tried to cast a variety of materials, including PDMS, Smooth-On EcoFlex 20A), 30A and 50A. We used Smooth-On SIL-poxy rubber silicone adhesive to bond the two casted parts together.

GRACE in Action

In order to test this first primitive structure, we used FlowIO soft robotics platform developed by Ali Shtarbanov of the MIT Media Lab. the sequential inflation photos are shown below

Adding Sensing (proprioception)

The future of soft robots is adding both proprioception and perception to the robot so it can smartly react to changing environments. We wanted to add some level of proprioception to our grace structure by embedding a flexible strain sensor.

We used carbon conductive grease from MG Chemicals as the main sensing material for resistive -type of sensing. We created a tape senticl mask to make it easier to pattern the great on top. Then we used thin wires to tether the sensor and finally encapsulated it with a thin layer of silicone.

For the sensor readout circuit, We employed a voltage-divider circuit using a commercial reference resistor and our sensor as the unknown resistor. We used Arduino Leonardo as the microcontroller and the Arduino IDE for the serial plotting.

Failures

We experienced many issues and failures during the fabrication process of GRACE. They are summarized as follows:

Sometimes the silicone would push the top part of the mold away due to the weight. To mitigate this, we sometimes needed to use an added weight to ensure the top part was fixed in place.

Another issue was the rupturing of the structure during demolding. Due to the ratio between the height of the structure and the wall thickness, it was sometimes hard to peel off the structure without it rupturing at some critical points. Applying a release agent helped in this case.

For the 3D printed top part, we used many different materials including PLA using FDM and resin using FormLabs. Although the ones based on the resin are nice and smooth, they become really problematic for casting; they are not cleaned well. If the resin is not completely cured, it inhbitis the curing process of PDMS or Ecoflex. We did experience this, and I think further cleaning will be a solution. Cleaning should involve more UV exposure and het

Another issue was secular bonding the two parts together without any leakage while maintaining the overall shape of the strcutre. We initially used silicone adhesive but then needed to dip the structure in more silicone to ensure final sealing. This added layer is not in the original design, and it is also not uniform, which might cause some variability.

Another issue we faced is difficulty in pattering curvilinear structures. Unlike planar structures, it was a bit tricky to pattern with the conductive carbon since it was a wavy structure. Potential solutions in the future include embedding the channcel/cavity of the sensor within the mold itslef, so filling can be leveraged. Or, using a spraying technique with a flexible conformal stencil mask.

Establishing a secure bond between the wire and the grease was also tricky. In planar structures, people have used copper tape before as the pad connecting the great to the wire. However, it was tricky here because of the curves so copper tape kept coming out of the plane. We found that embdeing a thin wire within the grease was eaiser. It is important, however, to apply top silicone layer for insulating and fixing it in place.