Linear Guide Shafts and SEA Spring Modules

Individual Progress

Linear Guide Shafts

Figure 1: CAD model showing series elastic actuators mounted on front and back of leg

After consulting the open-source SEA developed by Yobotics, I decided that our intended SEA design was insufficiently robust under bending moments. Under compression from each end, the actuator would have buckled, damaging the ball screw. Therefore, I added steel linear shafts (Figure 1) to serve as guides for the ball nut and spring module, and to support the end of the ballscrew. These shafts do add weight to the design, so I am looking at replacing them later on with aluminum or ceramic-coated aluminum linear shafts, although those are likely to be expensive.

SEA Spring Module

Figure 2: CAD model showing spring module and linear potentiometer

The spring module of the SEA (Figure 2) incorporates two pairs of two springs. One of the two pairs is under compression when the actuator is subject to compressive force. The other pair is under compression when the actuator is subject to extensive force. When not compressed, each pair of springs floats between plates, secured in the center by a shaft or spacer. A linear potentiometer is mounted to measure the deflection of the central element of the module relative to the outside plates. The force exerted by the SEA can be calculated based on this deflection and the spring constant.

Part Fabrication and Assembly

I continued fabricating parts using the Makerbot Replicator 2X. I cut the linear shafts using the abrasive chop saw at Techshop. I found instructions for cleanly cutting threaded rod and created a jig to hold the rod while cutting it. I greased the bearings, and inserted the brass thread protectors provided with the bearing nuts. Then, I assembled the ball screw mounting hardware and tightened the bearing nut to hold the assembly together. The brass thread protector prevents the bearing nut set screw from marring the thread on the ball screw. Finally, I mounted limit switches to prevent the ball nut running off the ball screw in each direction. These limit switches will be hooked up to a relay module that will cut power to the motors if the travel limits are reached. These limit switches are not triggered during normal operation.

Challenges

Short Ball screw Stroke

When selecting the ball screws, we chose a stroke length equal to the expected range of motion of the moment arm attachment at the joint. When we were making this choice, I didn't know that the easiest way to mount springs in an SEA is apparently to mount two pairs of springs in line with the ball nut. One set of springs sticks out to each side of the ball nut, so this design necessitates a ball screw stroke that is longer than my estimate by about twice the spring length (Figure 3). In retrospect, we should have just selected ball screws with a long enough stroke to accommodate this design, as the knee ball screw, which we had the option of choosing the length of, has a low mass/length. The ankle ball screw was ordered from a secondhand vendor and we had a limited choice of lengths available.

Figure 3: Open-source SEA developed by Yobotics and provided by IHMC

Actuator CAD Updates and Part Fabrication

Progress 11/05/2014

Actuator CAD

Contracted

Extended

Figure 1: Leg shown with motors and ball screws mounted

Since the last progress review, I designed the motor and ball screw mounting hardware, selected appropriate bearings and couplers, and designed the torque couplings at the knee and ankle (Figure 1). The ballscrew transfers the axial load to a mounting plate through two thrust bearings, which sandwich the mounting plate (Figure 2). Nestled between the thrust bearings is a ball bearing that handles any small radial load on the ball screw. Finally, the torque is transmitted from the motor to the ball screw through a high-torsional-rigidity, low-backlash coupler. The motor side of each actuator transmits force to the joint that is being actuated. The ball nut side of each actuator attaches through a rotary joint to the shank of the leg (Overview in Figure 1).

Figure 2: Knee actuator assembly

After completing the design of the actuator mounting, I fabricated the necessary components using the Replicator 2X 3D printer. I cut several joint shafts from rod stock using a horizontal band saw and ground the ends on a grinder wheel so that the bearings slide on smoothly. I cut a demonstration foot out of ABS plastic and drilled mounting holes.

Challenges

Knee Joint Design

Figure 3: Knee dual shaft holder (highlighted in blue)

In order to maintain sufficient spacing between the actuator assembly and the shank of the leg, some creative design of the knee joint was required (Figure 3). I set the angle between the two shaft holders in the knee component at 45 degrees, halfway through the full range of rotation of the knee joint. This helps to keep the joint moment arm relatively constant, and also serves to space the actuator assembly away from the shank.

Joint Component Assembly and Actuator Conceptual Revisions

Progress 10/22/2014

Conceptual Revisions

Based on further research into existing designs and on conversations with our sponsor, I eliminated the thigh link component of the leg. This allows attachment of a pyramidal adapter (standard prosthetic mounting component) as close to the knee joint as possible. Following the design choices made in the Vanderbilt transfemoral powered prosthesis, I will design the ankle actuator to mount on the back of the calf section and the knee actuator to mount on the front.

Figure 1: Joint with encoder assembly

I also explored the concept of using coil springs with a linear potentiometer instead of leaf springs with a strain gauge in order to measure torque at the leg joints. These springs would mount in-line with the hollow tube that the ballscrew travels inside, allowing design of the series elastic actuator in a single assembly intended to be mounted on rotational joints at each end. The other advantage of this design is that the linear potentiometer provides a much cleaner signal than the strain gauge. However, finding springs designed to operate in both tension and compression has so far proved difficult. Also, reliably mounting the spring will likely be difficult.

Finally, I identified and obtained two motors (AM Equipment 214 Series Gearhead Motor) and a suitable driver (Sabertooth 2X25) from the MRSD inventory that will allow us to integrate and test our position control algorithm early on, before the ballscrews and motors have arrived. These motors will mount directly to the joint shafts, bypassing the actuation system. With no load, these motors operate at about 65 RPM (390 degrees per second). This nearly satisfies the requirement for the joint rotational speeds, which at their maximum are about 450 degrees per second. Unfortunately, if these motors are attached to our intended ball screws, which have a high reduction ratio, the speeds at the joints will be very slow. Therefore, the torque control testing procedure would be time-consuming and probably ill-suited as a fallback for the final validation experiment.

Leg Component Assembly

After the joint components arrived from Misumi, I revised several rapid prototyped parts that had incorrect tolerances, and then assembled the knee and ankle with encoders (Figure 1).

Challenges

Knee Actuator Proximal Mounting

The proximal mount of the knee actuator must be cleverly designed to allow for 90 degree knee flexion, which happens often during daily life (sitting, driving, etc). Basically, the knee actuator, which is mounted on the front of the calf, must be able to connect to the pyramid adapter mount, which is oriented towards the back of the calf. This can be accomplished by a plate that sits alongside the knee joint and transfers motion parallel (approximately) to the calf to rotary motion on the other side, facing the thigh socket.

Risk of Incorrect Actuator Selection vs. Risk of Late Delivery

There are three main ways in which we could choose an inappropriate actuator system: underpowered, overpowered (and therefore overweight), and incorrect tradeoff of torque vs. speed. The effective gear reduction ratio is determined by the moment arm and the ball screw lead, with the option to incorporate a gearbox at the motor if necessary. The motor power will be supplemented by a brake at the knee and a parallel spring at the ankle, enabling the weight and cost savings that can be achieved with a lower wattage motor.

As we have learned more about all of these options, the risk of incorrect selection has steadily reduced, yet the risk of late delivery has steadily increased. At this point, we feel that the relative magnitude of these two risks has reversed, and we must select and order the actuation components as soon as possible. To somewhat mitigate the risk of late delivery, we will maintain two assembled test-beds. One will incorporate the alternate motors identified above. The other test-bed will allow mounting of the ballscrews and motors as they arrive.