Design of a Low Cost Transfemoral Knee Prosthesis with MMG for Developing

Countries: Crus Novus

Group 1: Eleanor Disney, Alice Boo, Alexander Camuto, Cecilia Kan, Rohit Devesar, Johnson Chu, Vekin Virachjarassin, Gerald Png,Rafael Michali, Yomna Genena.

1

Contents1. Abstract.........................................................................................................................................3

2. Aims...............................................................................................................................................3

2.1 Inexpensive..................................................................................................................................3

2.2 Ease of use...................................................................................................................................3

2.3 Durability.....................................................................................................................................3

3. Specifications and requirements...................................................................................................4

4. Final Design and Analysis...............................................................................................................6

4.1 Mechanical components....................................................................................................6

4.2 Hardware and Software...................................................................................................15

5. Device Testing..........................................................................................................................18

5.1 MPU6050 tests................................................................................................................18

5.2 MMG tests.............................................................................................................................21

5.2 Testing the PCB......................................................................................................................23

6. Manufacturing.............................................................................................................................23

6.1 Mechanical components..................................................................................................23

6.2 PCB.........................................................................................................................................26

7. Conclusion and Discussion...........................................................................................................28

7.1 How the prosthesis matched our aims............................................................................28

7.2 Improvements.......................................................................................................................28

7.3 Conclusion.............................................................................................................................28

8. Appendix......................................................................................................................................28

8.1 Appendix A (Commented Arduino Code).........................................................................28

8.1 Appendix B (Risk Analysis and Ethical Considerations)....................................................31

8.1 Appendix C (Business Case and Targeted Consumer)......................................................39

8.2 User Manual....................................................................................................................39

8.3 Appendix C (Group Working)...........................................................................................41

8.6 Appendix D (Initial designs)...................................................................................................42

8.7 References.......................................................................................................................48

8.8 Acknowledgments.................................................................................................................49

2

1. Abstract

2. Aims

2.1 Inexpensive

The project’s focus was to develop a versatile and robust below-knee leg prosthetic, controlled by mechanomyography (MMG), to be used in developing countries. The MMG needed to reliably detect muscle impulses and filter surrounding noise and residual muscle vibrations that occur after contraction. Because the target user was in developing countries a maximum budget of 500£ for the production of the prosthetic was set. If mass produced, the cost of components would be reduced considerably.

2.2 Ease of use

Once muscle signals were filtered the MMG needed to control the locking and unlocking of the knee prosthetic to allow the wearer to manoeuvre with ease. The interface would need to be simple enough for health technicians to set up rapidly and for users to interact with easily. The design aimed to be as simple as possible, both in the electronics used for the MMG and the mechanism to lock the knee so that maintenance and repair of the prosthetic could be done with minimal equipment.

2.3 Durability

Emphasis was also placed on the robustness of the prosthetic. The materials and structure of the prosthetic would need to bear the forces repeatedly inflicted upon the leg during walking and standing and could protect the electronics used for the MMG (give values of forces in next sentence). It would need to be durable enough to be used in developing countries where uneven terrain and unpaved roads are prevalent. To be versatile enough to be used in an array of conditions the prosthetic and the MMG (particularly any sensors used for the MMG) would be housed in a waterproof case to protect from sweat and rain. For ease of use any power supplies used would need to power the MMG and any motor used in the knee for 16 hours continuously and would need to be rechargeable. Additionally if power were to be lost during the use of the MMG, the prosthetic would need to lock in an extended position (stiff) to give the user manoeuvrability. Combined these elements aimed to provide the wearer a gait that would outperform that granted by a passive mechanical knee prosthetic in speed, manoeuvrability and comfort.

3

3. Specifications and requirements

Requirement Type:

Requirements: Comments: Design Specification:

Functional The knee must lock at the beginning of the swing and stance phases of the gait cycle.

Required so the knee does not buckle in the stance phase and has adequate support. The knee must bend enough in the swing phase so that the foot of the prosthesis doesn’t scrape the ground.

The knee hinge joint is locked by altering the tension in the steel cable around the barrel. The signals from the accelerometers on the user’s muscles will determine when the actuator pulling the cable should increase or decrease the tension to lock the knee.

Functional The MMG technology must be powered by a long lasting power source. Baseline lifetime must be a day (24 hours), with target of a week

As this is aimed at amputees in developing countries, it’s important that it is powered by a battery that will last at least a few days to a week as they may not have regular access to charge the battery or not be able to afford a new one.

A 9V lithium rechargeable battery for the Arduino (which delivers 5V to the accelerometers). A 12V battery powers the linear actuator. Lithium batteries were chosen because of their longevity.

Form The prosthesis must have a natural weight

If it’s too heavy it will make it very difficult for the amputee to walk with it and tire them quickly.

Lightweight material with most of the weight focused around the knee.

Form Simple design that can be worn subtly under clothing

Ensuring that the prosthesis looks natural as possible under clothing improves the way the user relates to the product.

The cylindrical body of the prosthesis with adjustable shaft base can be enhanced with a detachable cover to imitate the natural shape of a lower leg.

User The Prosthesis must use parts that can be easily fixed/replaced. (i.e. no specialist parts).

In developing countries it’s unlikely the user will have access to specialist facilities so the parts need to be easy to replace and fix.

All of the parts are easily replaceable except the linear actuator. The shaft is an aluminium body formed of four rods, three plates and a hinged section with horizontal steel supports.

User The battery/electronic components must be easily accessible.

This is in case the battery needs replacing or the electronic components need repairing.

Ideally there would be a removable plastic cover on the shaft of the leg which will allow the user to replace the 9V (Arduino) and 12V (Linear Actuator) battery.

User The Crus Novus must be Water Resistant.

This is to allow the user to wear the prosthesis in all weather and prevent

Ideally there would be a cover on the shaft. This covers the electronic components and is

4

damage to the electronics. waterproof.

User The Crus Novus knee must be usable for a range of weights and heights.

This is a low cost knee that should be usable for many above knee amputees. Therefore there should be ways of extending the shaft for taller users and it should support people in a range of weights.

At the bottom of the shaft there is a plate with a thread where an extension shaft can be screwed in to make the leg longer for taller users. The Crus Novus supports a maximum weight of 100kg.

User The Crus Novus knee must be lined suitably to minimize sweat build up and friction between the skin and accelerometers.

This allows comfort for the user during repetitive daily wear. At minimum, on par with those made available by competitors with products in our price bracket.

Using a sweat wicking material to draw moisture away from the skin and have the accelerometers embed into pockets in the lining.

User The Crus Novus has patient - gait cycle matching

This allows the motion of the knee to match the natural pace of the patient to support a more natural feel for the user.

The precoded software can be adjusted to match patient preferences (c.f user manual).

Cost The Crus Novus Knee must cost less than £500 to manufacture.

The prosthesis must be affordable for users in developing countries whilst still being a high quality knee that is durable and improves the user’s quality of life.

The knee costs less than £370 to build.

Cost The battery must be cheap.

This is so the user can buy a new one if it fails without sparing too much expense.

Cost of the battery - depends on which battery we use.

5

4. Final Design and Analysis

4.1 Mechanical components

4.1.1 Knee Frame The mechanical framework is equivalent to bones in the human leg, and allows the functionality of the prosthetic knee. It must be strong enough and designed to carry out its main functions:

1. To allow the user to balance on it while standing;

2. To support the load of the user;

3. To contain the components of the MMG sensors and locking system;

4. To provide the tension required for the locking mechanism; and

5. To allow bending and swinging motions as in a natural human knee.

The following table shows how the features of the frame must comply with the requirements of our project as a design aiming to reach the developing world.

Feature Advantages

Simple overall structure

Easy to assemble

Allows easy adjustment to adopt different strength and size of linear actuator and MMG components

Simple component structure

Lower cost of manufacture

Minimum number of components

Lower cost of material

Lower weight of structure

Easy to clean and maintain

Table 1. Features of the framework that aim towards designing a low-cost upper-limb prosthetic knee

The overall structure should allow space for components of the locking system to function and attach. The centre of mass should be close to the centre of the axial plane so that the structure can balance on a flat surface. However the centre of mass should not be in the lower leg as the user would experience a larger moment in the swing phase. A compact structure is required for the comfort of the user, so of maximum 105mm diameter in the axial plane was set, based on the smallest adult knee brace size from knee brace protection sellers1. A minimum 3mm thickness of any part of the structure was set. The initial drawing of the framework was prototyped using Meccano. In this model, elastic bands were used to mimic the spring response of the device. The prosthesis includes the lower leg to utilize the extra space for placing the locking system.

A brief description of the main components that form the skeleton of our framework design is given below:

6

Structure Purposes Design requirements

Barrel on horizontal axis

This acts as a hinge joint for the attachment of the lower limb prosthesis to the upper limb. The rotation of the barrel around its axis mimics the swinging motion of a knee.

The width needs to be 12mm for a 4mm ø wire rope to wrap 2.5 loops around the barrel. There must be smooth rotation between the barrel and its axis, and the cable wrapped around it. The axis must be able to withstand bending from 1.64kN shear force from the tension of the cable.

Barrel clip The standard socket piece for lower limb prostheses to the upper limb is screwed onto the barrel clip. The clip is fixed to and rotates with the barrel.

The clip must provide enough surface area for the attachment of the standard socket piece, the springs, the barrel, without colliding with other components of the prosthesis. It must withstand bending from a maximum 980N shear force from the load of the user.

Vertical supports

These are rods that support the load of the user.

They must support the weight of the user, which is a maximum 980N axial load without

7

Figure 1. The initial CAD drawing of the frame in front view. The wire ropes ends were to be fixed to the bottom bar by crimping

buckling. They have to be arranged to allow space for components of the locking system.

Flat plates

These are plates that hold the vertical supports in place. They may also provide means for the cables and springs to attach, as well as the standard socket piece for lower limb prostheses to artificial feet.

They must allow space for components of the locking system, on top of the vertical supports. The plate with attachment of cables and springs should withstand bending from 1.64kN shear force due to the tension of the cable.

Table 2. Basic components of the designed prosthetic knee

The restrictions to the framework design were mostly geometrical. Difficulty in manufacturing, and considerations for ease of assembly also restricted the design. Before the choice of locking system parts and MMG parts were finalised, the required dimensions of the frame were also unknown due to uncertainty in the size of the linear actuator and the barrel. The progression the frame is detailed below:

Barrel

The diameter of the barrel was increased from 30mm in the initial planning stage to 60mm after detailed force analysis was made for the locking system. The location of the barrel was moved to above the top flat plate as it could no longer fit within the curvature of the flat plate without the structure becoming too bulky.

Barrel clip

A barrel-clip was used to fix the barrel and its pin along its main axis (to prevent rotation). The details of the clip are detailed in the manufacture section.

Vertical supports

The cross sectional shape of the rod was changed subsequently from a thin rectangle, hollow square, to finally a filled circle, to increase the area to length ratio of the shape, as well as for a more compatible shape with the rest of the components.

Flat plate

The initial curved oblong plates were designed to minimise volume. To increase their surface area to accommodate for thicker vertical supports, the ends of the plates were extended into a horseshoe shape. However this shape was harder to manufacturing, leading to the decision of using circular plates. The circular plates also allowed the more even distribution of the supports hence better stability of the structure.

8

The final framework design is shown in figure D. The components include:

30mm ø barrel

8mm ø barrel axis

barrel clip (L 90mm, W 66mm, H 63.5mm)

5mm ø barrel pin

17.4mm ø main vertical support ×2

6.3mm ø additional vertical support ×4

10mm thick, 90mm ø circular flat plats (high ring, low plate, base)

The flat plates are held in position by circlips. There is 150mm spacing between the high and low flat plats, forming a compartment to accommodate a 130mm linear actuator in full extension. The 110mm spacing between the low plate and the base can hold an Arduino Mega (102mm ×54mm ×15.3mm) and a battery (100.1mm ×60.7mm ×26.5mm).

Each end of the wire rope would go through the 5mm ø holes in the high ring. The loading end would loop through the hole in the low plate and be fixed with crimp. The other hole on the low plate was a screw hole for mounting the linear actuator. The barrel clip is fixed to the barrel with the pin. The top of the clip is 15mm to accommodate 12mm deep screw holes for the standard socket piece. Two holes on the barrel clip are intended for wires attached to the springs that would go outside the circular plates. Wires on the lower end of the springs are fixed to the vertical supports by crimping.

Specifications:

Total length: 397.5 mm

Diameter: 90 mm

Minimum permitted angle of rotation of barrel: 160 °

Material: Aluminium alloy except stainless steel barrel axis

Total weight: 1.40 kg

To assess the performance of the framework, force analysis for various components is shown below:

Main vertical supports

Stress¿user body weight= 100×9.810.00852π ×2

=2.16×106

9

Figure 2. CAD drawing of the final design that was sent to manufacture in back view.

Safety factor=Material proof stressStress ¿

user body weight ¿= 260×106

2.16×106 =120

Material proof stress taken from Aalco Aluminium Alloy 6082 - T6 data sheet 2

Low plate

To analyse the behaviour of the low plate under the tension from the rope wire, we may use Kirchhoff–Love plate theory for circular plate with clamped edges as the vertical supports help resist bending. 3 The in-plane stress and the bending moment due to the tension in the wire rope is given by:

σ rr=−3Fz32h3 [ (1+v )a2−(3+v )r 2 ]=30.6 kPa

σ θθ=−3 Fz32h3 [ (1+v )a2−(1+3v ) r2 ]=11.6 kPa

M rr=−F16

[ (1+v )a2− (3+v ) r2 ]=2h3

3 zσ rr=0.19Nm

M θθ=−F16

[ (1+v )a2−(1+3v )r2 ]=2h3

3 zσθθ=0.51Nm

where σ is the in-plane stress, F = 1.64kN is the tension from the wire rope, M is the bending moment due to force F, the Poisson ratio v = 0.33 (from trend in 5000 series aluminium alloy) 4, the maximum radius of the plat a = 0.045 m, h = 0.005 is half the thickness of the plate, z = h and r = 0.048 are the point of application of force F in the z and r directions respectively.

Safety factor=Material0.2%Yield strengthσ rr

= 115×106

30.6×103 =3756

Safety factor=Material0.2%Yield strengthσθθ

= 115×106

30.6×103 =9885

Material 0.2% Yield strength taken from Kastal 300 Data Sheet, Smiths Metal Centres 5

For the breaking load of the wire rope where F = 8.826kN,

σ rr=0.164MPa M rr=2.75Nm Safety factor=698

σ θθ=62.6 kPa M θθ=1.04Nm Safety factor=1837

The final product of the framework has been tested out and it functions as intended. The structure is over-designed as the safety factor is at least 100 times higher than typical values. 6 The dimensions of components were set under unknown strength requirement, and was over-

10

engineered to account for these uncertainties. For example, the linear actuator used has a minimum length of 130mm and hence could not fit into the assigned compartment when extended. The hole intended for wire rope was drilled larger, and then the outer rim sawed off entirely to allow space for the linear actuator to extrude. However, the safety factor is still extremely high in the light of such issues. The diameter of the vertical supports and the thickness of the flat plates can be reduced to lower manufacturing cost and reduce the weight of the prosthesis.

4.1.2 Final Locking Mechanism:

The designed above-knee prosthetic has to inhibit the rotation of the knee, in the sagittal plane, upon muscle contraction sensed by MMG sensors. Controlling the rotation of the prosthetic knee improves the Gait Cycle of the patient, but it also allow the use of their leg for a larger range of actions such as climbing up the stairs, or sitting down on a chair. The locking mechanism has two modes of actions: locking the knee and freeing the knee. The prosthetic knee has to be locked during the stance phases of the Gait Cycle and when the knee is in full extension to prevent the prosthetic leg from buckling under the applied load. The knee is freed during the swing phase of the Gait Cycle so it can swing back into its full extension position. The mechanism can also be locked at different angles of flexion to perform some specific tasks: 83 degrees to climb up the stairs and 90 degrees to descend the stairs. Walking requires an angle of flexion of 67 degrees. Our design had to be able to lock and unlock easily without using too much energy in order to save some battery life. In order to achieve this goal, the knee was designed such that it would be in the locked position when no power is supplied to the system. This presents the advantage to lock the device when it is out of power and therefore to prevent the patient from falling if his or her device was running out of batteries.

The final locking mechanism design was implemented by using the friction between an aluminum alloy barrel and a stainless steel wire rope. The wire rope prevents the rotation of the barrel when it does not slip. We used the Capstan Equation (1) to determine the minimum allowable value of the tension in each part of the cable so that the cable does not slip.

11

Figure 3. The Final Locking Mechanism of the Above Knee Prosthesis

Aluminum Alloy Barrel

Stainless Steel Wire Rope

The Capstan Equation:Fhold=F load×e

μϕ1

ϕ: The wrapped angle around the barrelFhold: The force acting on the tensed side of the wire

F load: The external brake actuation forceR: The barrel’s radiusT: The braking torque as a function of F load that tightens the wire rope around the barrel

T=(Fhold−Fload ) R+μvisc×ω 6

μ: The contact friction coefficientμvisc : The viscous friction coefficientω: The barrel’s angular velocity

The system locks such that ω = 0, so:

T=(Fhold−Fload ) R

∴T=F load (eμϕ−1 )R

Knowing that the maximum patient’s weight requirement is 100kg, the diameter of the stainless steel cable is 4mm, the barrel’s radius is 30mm, μ for aluminum against steel is 0.35, ϕ is 5π or 2 turns and a half around the barrel, and the following knee torques 7:

T knee joint(walking) 0.31 Nm/kg

12

Figure 4. The Capstan Model of the Locking Mechanism

R

Drum rotation

ϕ

F loadFhold

T knee joint(stair up) 0.49 Nm/kgT knee joint(sit−¿−standwith awalker ) 0.72 Nm/kg

T knee joint(sit−¿−standwithout constraints) 1.20 Nm/kg

To allow the patient to climb up the stairs, the tension in each part of the wire rope have to reach the following values in order to lock the device:

F load=6.7N

Fhold=1.64kN

The torque at the knee joint is changing through the Gait Cycle

This means that the

tension applied by the linear actuator should fluctuate with respect to the Gait Cycle. However to simplify the task, we have decided to only account for one value of the external applied load in the end of the cable subjected to the tension provided by the linear actuator. It was found for the greatest moment at the knee joint for a 100kg patient climbing up stairs.

The stainless steel cable used for the locking mechanism was a wire rope 7x7 construction. It has 7 strands and 7 wires per strand. The cross-section of the cable is not perfectly circular, but it can be approximated to be 2.83x10-3 m2. Its breaking load is approximately 0.9 tones or 8.826 kN. The tension in the holding end of the cable being 1.64 kN, our design factor for the wire rope is approximately 5.4.

Safety Factor=Stainless SteelWire RopeBreaking LoadResisting End of Wire Rope Load

=5.4

The holding cable is attached to an aluminum alloy plate designed to resist bending under the applied tension during the locking state of the device.

13

Figure 5.The Knee Moments with Respect to the Percentage of the Gait Cycle in a Healthy Patient and in a Patient Equipped with a Prosthetic Knee 8

Moments in a prosthetic kneeMoments in a healthy patient’s knee

The maximum allowable patient’s weight, mmax, is accountable for Fhold = 8.826kN, so using the formula derived from the Capstan equation:

mmax=R×Fhold

(τ+ τeμϕ−1 )

6

mmax(kg)msafety (kg )SF=8.5

msafety (kg )SF=5.4

msafety (kg )SF=3.7

msafety (kg )SF=2.2

Walking 850.6 100 157.5 230.0 386.6

Stair up 538.1 63.3 100 145.4 244.6

Sit-to-Stand with a Walker 366.2 43.1 67.8 100 166.5

Sit-to-Stand without Assistance 219.7 25.8 40.7 59.4 100

The variable msafety was calculated using the safety factors 5.4 and 2 9. SF = 5.4 corresponds to the design factor previously found by comparing the tension in the resisting cable and the breaking load of the 4mm stainless steel cable. In this table, we have calculate the safety factors required for a person weighting a 100 kg to person three different tasks: climbing up the stairs, standing up from a sitting position with and without a walker. The knee torque required to perform those tasks is fluctuating: it is lowest to walk and it is the highest to stand up without requiring assistance. As a result, the safety factor increases as the required knee torque increases. The initial requirement of our device was to allow the patient to walk according to a Gait Cycle as close as possible to the natural Gait Cycle. This would lead to a safety factor of 8.5 for the stainless steel wire rope. Such a factor is within expected range for a wire rope according to the Engineering Toolbox Website10. However, such a safety factor would not allow a 100 kg patient to climb up stairs or to stand up from a sitting position. Keeping a safety factor of 8.5 to adapt the locking mechanism for a 100 kg patient to perform a sit-to-stand action would lead to an increase of the wire rope diameter. Increasing the diameter of the wire rope would cause a change in the barrel design; its width, currently 13mm, would have to be widened. Those changes would critically increase the weight of the locking mechanism and disturb the patient’s Gait Cycle. Keeping a safety factor of 8.5 without changing any other variable than the radius of the barrel would give us the following results. The new radius of the aluminum alloy barrel would be 116 mm compared to 30mm, assuming that mmax=850 kg and the torque τ=1.20Nm/kg for the sit-to-stand without assistance action.

R=

mmax×τ (1+ 1eμϕ−1 )

Fhold

14

Figure … Maximum and Safety Weights for a 100 kg Patient Performing Different Tasks (SF = Safety Factor)

The new radius would be 4 times bigger than the original radius of the barrel. This would not only increase the weight of the part but it would also compromise the design which would become bulkier. It justifies the use of a safety factor of 2 as it would allow to keep the design of the locking mechanism as light and compact as possible. As a final result, a safety factor of 2.2 would allow a 100 kg user to perform all the tasks mentioned in the table.

The inconvenience of this locking mechanism is that the aluminum alloy material is softer than stainless steel as its shear; tensile and tangent moduli are approximately three times larger than the ones of aluminum alloy 6061. The density of stainless steel is around 7480-8000kg/m3 while the density of aluminum 6061 is around 2700kg/m3. As the stainless steel wire rope slides around the surface of the aluminum barrel it will produce wear that will affect the longevity of the device.

4.2 Hardware and Software4.2.1 Hardware

The final circuit was designed with the following components:

2 separate power sources.◦ 12V to power the linear actuator◦ 7V to power the Arduino, H-bridge and

5 connections between the linear actuator and Arduino◦ 1 pair of wires controlling the movement of the linear actuator.◦ 1 pair of wire as voltage reference for the positional feedback built into the linear

actuator◦ 1 wire which sends positional feedback to the Arduino

2 connections between the accelerometer and Arduino

After more testing, more components were added as their need became apparent.

In order to reduce the effect of noise from the power supply and voltage changes from the Arduino, 4 decoupling capacitors were added between the power supplies, Arduino outputs and the ground. The capacitors absorb any voltage fluctuations, reducing noise on the rest of the circuit components.

An H-bridge was used to control the Firgelli L16-50-150-12-P linear actuator (capable of withstanding 250 N, above the requirements of the leg of 6.7 N cf.Section 4.1.2) to the Arduino 4. The function of the H-bridge is to act as a switch between the power source and linear actuator. This allows us to control the voltage provided to the pair of wires on the linear actuator using a smaller signal, thus controlling both speed and direction of operation of the actuator from the Arduino. This allows us to reduce the amount of code needed and only requires two 2 outputs. The H-bridge was linked to 4 diodes to protect the H-bridge as instructed by the Texas Instruments user manual of the SN754410 Quadruple Half-H Driver (the H-bridge used) 11. The final circuit design is shown below.

15

Figure 6: Breadboard diagram of final circuit

Figure 7: Final circuit diagram for the circuit used for the PCB board manufacture.

16

4.2.2 Software

The program for the Arduino was written in C, using the MPU-6050 library found at http://www.i2cdevlib.com/devices/mpu6050 to interface with the MPU-6050 accelerometer we used.

The signal from the accelerometer is processed using a combination of high and low pass 5 th

order Butterworth filters taken from http://www.schwietering.com/jayduino/filtuino/index.php, which automatically calculates the coefficients we need for the filter. The filters are centred around 10-15 Hz, which matches muscle frequencies, allowing us to detect muscle activation.

2 outputs control the signal (high/low) to the H-bridge and then to the linear actuator. This is accomplished using a simple digitalWrite command.

In the initial setup, various variables to be used are initialised. The outputs are set such that the linear actuator retracts till the positional feedback reads 0.

This ensures that the actuator starts from the same position after every reset, allowing the user to manually reset the knee position.

The code then enters a constant loop, where it runs the signal from the accelerometer through the band pass filter at roughly 50 Hz. Muscle activation is detected when the signal hits an experimentally tested threshold for 3 or more consecutive loops. The outputs are then changed such that the linear actuator extends roughly 2cm, which loosens the knee for rotation. The linear actuator is then set to retract to its original position to lock the knee back in place. This completes the cycle of unlocking/locking the knee during a single stride.

For the purposes of the project, the code runs as expected, allowing us to control the locking of the knee using muscle activation. However, while the filters proved to be reasonably robust when filtering noise, it is still possible to improve it through more rigorous testing

to determine the optimal filtering coefficients.

The code can be further improved by designing a sinusoidal extension/retraction routine, which would make the locking of the knee smoother. The activation condition (more than a certain threshold for a certain number of cycles) can be made more accurate if given more time for extensive testing. Implementing the bandpass filter using hardware instead of software may also be more dependant when filtering noise.

The entirety of the code can be found in APPENDIX A.

17

Figure 8: Setup for Arduino code

5. Device Testing

5.1 MPU6050 tests

To ensure the MPU6050 accelerometer could act as an MMG device, we needed to collect data from muscle contractions using the MPU6050. To do so the MPU6050 was connected to an Arduino UNO and the targeted muscle in the following manner:

Figure 9: Built using Fritzing. MPU6050 detailed to the right in red and powered by the Arduino. The Arduino UNO is detailed in blue. The MPU6050 was strapped the subject’s arm using muscle tape.

The actual experimental setup is detailed in the following photo:

Figure 10: Experimental setup of diagram 1. The pink strap holding the MPU6050 to the subject’s arm is muscle tape. The illuminated red dot is the powered MPU6050. Note: the Arduino itself is connected to a laptop.

Data was then collected using the “serial print” function of the Arduino and analysed using an array of Matlab functions. The first step was to see if there was a clear correlation between

18

MPU6050 activity and muscle contraction. The test subject was asked to contract his forearm at regular intervals. This resulted in the regular amplitude spikes of the MPU6050 data seen in the figure below. Each spike corresponds perfectly with the onset of contraction. The plateau phase seen after each spike corresponds to the when the subject sustained contraction for 1 to 2 seconds.

Figure 11: Amplitude against the sample number for regular forearm contractions. Note: The amplitude are just relative values given by the serial.print function of the Arduino and are thus unit-less.

To determine whether these spikes were simply due to the displacement of the accelerometer and were not due specifically to muscle activation we conducted another array of tests. After contacting Ben Greer of the University Of Colorado (who conducted a similar MMG project) we decided to use the Matlab wavelet functions to further analyse our data. Our results are plotted in the figures below.

19

Figure 12: Figure 1 data juxtaposed to the wavelet transform of the same sample data, plotted with the wavelet scale against the sample number (for the same sampling frequency). Yellow shades correspond to greater amplitudes in the

wavelet domain. Blue shades correspond to depressions in amplitude.

Plateau phases of contraction correlate with spikes in the 4-7 wavelet domain as seen by the regular yellow shades in the wavelet scale of figure 2. Using the Matlab scal2frq function we determined that for this particular wavelet transform the spikes in the wavelet scale corresponded to the 10-15 hertz frequency range. We thus determined the MPU6050 could be used as an MMG device.

To improve the legibility of our results and to remove oscillations from our MPU6050 readings we used a smoothing function in our Arduino code (the code for which is detailed in Appendix A of this report and was provided by Dan Greer). Using the same setup as in diagram 1 and asking the test subject to contract his forearm at regular intervals we obtained the figure below. This function increased the legibility of our results and significantly decreased noise in MPU6050 readings as seen in the figure below.

Figure 13: Smoothing function test with amplitude and plotted against sample number (with a 333Hz sampling rate). Note: Amplitude is mapped to a 0 to 1.2 relative scale

To improve MMG signals, frequencies outside the 10-15hz range needed to be filtered using the Arduino UNO. Jürgen Schwietering’s website (http://www.schwietering.com/) provides Arduino code for 4th and 5th order Butterworth filters, both of which we tested (the code for the filters is detailed in Appendix A of this report). Using the same setup as in diagram 1 and asking the test subject to contract his forearm at regular intervals we obtained the figures below.

20

Figure 14: 4th Order Butterworth Filter with Wavelet Scale against sample number (sample frequency of 333hz)

Figure 15: 5th Order Butterworth Filter with Wavelet Scale against sample number (sample frequency of 333hz)

The 5th order Butterworth provides greater filtration of signals outside the 4-7 wavelet scale range (10 to 15hz) as seen by the net decrease of noise outside this range (seen by the decrease of yellow tones outside the 4-7 wavelet scale range) when compared to the 4th order Butterworth and figure 2. Figures 4 and 5 have the smoothened MPU6050 data plotted at their base to reinforce the correlation between contraction and amplitude spikes in the 10 to 15 hz frequency domain.

5.2 MMG testsOnce MPU6050 signals were accurately filtered using a low step and high step filter that

accurately replicated the results of the 5th order Butterworth we wanted to determine whether our current MMG device could accurately control a DC motor and later a linear actuator. Using the following circuit and Arduino loop code we used muscle contraction to accurately control a low power servo motor (that did not require a motor shield).

21

Figure 16: Arduino UNO connected to MPU6050 (itself strapped to a test subject’s forearm) and a low power servo motor powered by the Arduino.

void loop ()

{

sensor.getMotion6(&ax, &ay, &az, &gx, &gy, &gz); // read mpu data

mag = highstep(lowstep(ax/3276.8)); // scale ax and feed to filter low and high pass filters

if (millis()-lastmax < 100)

{

if (fabs(mag) > maxval) maxval = fabs(mag); // find max value in current time window

} else {

currval = maxval;

maxval = 0.0;

lastmax = millis();

}

smoothval = 0.9*smoothval + 0.1*currval;

val= smoothval;

val = map(val, 0, 1023, 0, 179); //Maps smoothval from 0 to 1023 bits to 0 to 180° angles

myservo.write(val); // sets the servo position according to the scaled value

delay(15); // waits for the servo to get there

SoftwareServo::refresh();

}

Code 1: void loop code to test servo activation from user contraction. The code for filters (highstep, lowstep), void setup and other functions are detailed in Appendix A

22

The test subject was again asked contract his forearm at regular intervals. Each Contraction resulted in a proportional rotation of the servo motor (the greater the amplitude of the MPU6050 the greater the angle of rotation).

The final circuit design was tested using the final implementation of the Arduino code (detailed in the design section). Our final design was robust against noise (when the subject was hit or shaken the linear actuator did not activate), generated linear actuator action and caused the leg to unlock whenever the subject contracted his bicep. The results can be seen in the following YouTube video http://youtu.be/hWD8yTxIZLk.

5.2 Testing the PCB

The PCB designed and manufactured (detailed in the manufacture section) was rigorously tested when received and when components were soldered. Every connection on the PCB was tested pre and post solder using a voltmeter, thus ensuring no connections were accidentally connected to the ground plate of the PCB. Despite this rigorous testing, the PCB caused the Arduino to short circuit whenever operated. The reasons for this are unknown.

6. Manufacturing

6.1 Mechanical components

6.1.1 Instructions for manufacture

The knee is mostly made of aluminium alloy 6082-T6, There are some small components made of stainless steel (i.e. barrel pins) and copper (i.e. bushing). The outer case is made of PET plastic.

Firstly, the barrel clip on the top of the Knee is manufactured. The barrel clip is separated into three parts, two sides and a middle part. At the centre of the top surface of the middle part, a shallow circular notch with 38mm diameter is cut. Four threaded holes with 8mm diameter are drilled near the notch. Two small holes with four mm diameter are drilled at the position 7mm away from the front edge and 12mm from the side. Two holes, one with 6mm diameter and the other one with 14mm diameter are drilled on two sides. The smaller hole is located at 31mm down from the top surface and the larger one is located at 53.5mm down from the top surface. These holes are for the pins to go through.

Secondly, a barrel with 60mm diameter is manufactured. The barrel has a 0.5mm thin wall on the edges, and two holes on the side of it. One hole with 6mm diameter is drilled at 19mm away from the centre, and the other hole with 14mm diameter is drilled at the centre. These two holes are for the pins to go through and they are aligned with the holes on the side of the barrel clip. The pins are made of stainless steel. Both pins are 100mm long but one is with 5mm diameter and the other one is with 8mm diameter.

23

Next step is to manufacture three ring plates. The upper ring plate has eight holes. One hole with 35mm diameter is drilled at the centre of the plate. Two holes with 17mm diameter are drilled at 31mm away from the centre hole. Another two holes with 5mm diameter are drilled at 31mm away from the centre and are perpendicular to the centre. Four holes with 7mm diameter are drilled at the four corners of the plate. Each one is 21.9mm away from the centre. The middle plate is similar to the upper plate but without the centre hole. The bottom plate is similar to the middle plate but without the 5mm holes. Moreover, the holes of the bottom plate are threaded.

The fourth step is to manufacture six aluminium bars to support the structure. Two of them are 350mm long and with 17.40mm diameter. There is a 12mm diameter hole drilled through at 15mm down from the top surface. At the bottom surface of these two bars, there is a hole drilled 30mm into the bar which is for a M8 screw. The rest of the bars are 288mm long. There are caps, which are 8mm high and with 12mm diameter on the top of those bars. The diameter of these bars is 6.3mm. At the bottom surface of the bars, there is a hole drilled 30mm into the bar which is for a M4 screw.

To assemble the knee, put all six support bars through the upper ring plate. Using circlips fix the plate by clamping both the larger bars to the bottom of the plate. Next fit the middle plate using circlips. For the middle plate, the circlips are clamped at the top and bottom of the plate. Thirdly, hold the bottom plate in position and screw the bars onto the plate.

After fixing the plate in position, align the holes on the top of the bars, the hole at the centre of the barrel and the holes on the sides of the barrel clip. To assemble these three components place the bushing (made of copper) inside the holes and put the 8mm diameter pin through. The 5mm diameter pin is to fix the barrel and barrel clip together.

After assembling the main structure of the knee, place the linear actuator in position on the middle plate. A 4mm diameter cable is fixed on one end on the middle plate. The cable then goes through the upper plate and is roped one and half times around the barrel. It then goes through the other end of the plate and is fixed on the top of the linear actuator. To fix the cable on either ends feed the cable through the plate or the top of the linear actuator and then clamp the cable.

24

Figure 17: Final deconstructed CAD diagram of the Crus Novus

25

6.2 PCB

6.2.1 Introduction to PCB manufacturing

A Printed Circuit Board (PCB) is generally used to minimize the path lengths of electrical current between electrical components and provides a convenient platform to arrange the components in a compact way. This allows the circuit to take less space in devices. In addition, the location of the electronic parts is fixed and therefore simplifies component identification and enables the circuit to be placed in a moving knee prosthesis.

6.2.1 Gerber files and process

Software called Fritzing was used to transform breadboard design into schematics and PCB Gerber files. The PCB was designed to maximize the number of parallel copper tracks and to keep connections between nodes as short as possible. This was to reduce noise from the copper tracks that could have interfered with the MMG function. The PCB circuit matches the circuit detailed in the design section. The PCB is the same size as the Arduino UNO, 48.72 cm2 (7.73cm x 6.3cm), therefore the PCB will fit exactly on top of an Arduino UNO.

Figure 18 – Both layers of PCB

Black lines are the silkscreen on the PCB. Ground plates grounded both layers of PCB, allowing all grounds and unused nodes to be grounded as can be seen from the red holes. The green holes show that a node is connected to another node. The yellow and orange copper tracks are the connections on the top layer and the bottom layer respectively.

On the left, there are 4 diodes, D1, D2, D3 and D4. J1 and R2 represent the connections for the linear actuator where J1 is for VCC, GND and positional feedback and R2 is for positive and negative positional feedback references. In the middle, there is a L293DNE H-bridge. Next to it is the MPU6050 accelerometer.

26

Figure 19 – Top layer of PCB

Figure 20 - Bottom layer of PCB

Figure 2 and 3 show connections on top and bottom layers respectively and drilling holes in orange.

These Gerber files were sent to the company PCBtrain who manufactured 4 copies of the PCB, excluding the silk screen.

27

7. Conclusion and Discussion

7.1 How the prosthesis matched our aims

The prototype successfully completed our main objectives. We constructed the first MMG controlled knee prosthesis and did so for minimal costs. The knee locking mechanism held the leg in any position the leg was to be locked in. The filters used were reasonably robust to noise and the MMG consistently recognized muscle contraction. With additional time and research we could have resolved issues regarding the PCB board, perfected filtering, strengthened the locking mechanism and added a waterproofing case for the electronic circuit and the knee prosthesis. Our product would additionally require clinical trials to test its robustness, durability and comfort in real life situations.

7.2 Improvements

Future developments would include a waterproofing case, reducing the Arduino to its ATmega328 chip or even developing a custom built microcontroller to replace the Arduino (both would reduce costs). Developing a linear actuator solely for the knee would also have improved the fitting of the linear actuator within the knee. Additionally more emphasis could have been put on the user experience. A battery life indicator, an easier charging mechanism and perhaps improved aesthetics would all improve the ease of use and maintenance of the prosthesis for little additional costs. Ordering single components to build the prototype increased costs significantly: mass producing the device would reduce costs significantly to better target developing countries.

7.3 Conclusion We have proved that a reliable, cheap, robust and active prosthetic is feasible. Our simple design

means the knee would be easy to maintain, an important feature for developing countries where time and health technicians are lacking. A cheap, active prosthetic could drastically improve the lives of amputees in developing countries who typically need to go through 15-25 prosthetics in the course of a lifetime13.Further research into such a prosthetic could very quickly produce a market ready prosthetic and improve many lives.

8. Appendix

8.1 Appendix A (Commented Arduino Code)

#include "Wire.h"// requires I2Cdev library: https://github.com/jrowberg/i2cdevlib#include "I2Cdev.h"// requires MPU-6050 part of the I2Cdev lib: https://github.com/jrowberg/i2cdevlib/tree/master/Arduino/MPU6050#include "MPU6050.h"'#include <Servo.h>

#define NUMLEDS 7#define LEDPIN 13

int ii;

28

//ServoServo myservo;int angle = 0;

const int relay1Pin = 3; // the number of the Realy1 pinconst int relay2Pin = 4; // pins connected to actuatorconst int originalposition = 1024;const int sensorPin = 0;

int CurrentPosition;int goalPosition;int counter=0;int counter2=0;boolean VAL=false;// sensorMPU6050 sensor;int16_t ax, ay, az;int16_t gx, gy, gz;float gain;

// filteringfloat v[9], w[9];float currval, maxval, smoothval, mag;float smootharray[3];uint32_t lastmax;

//analogoutconst int analogout = 5;

//

void setup (){ Wire.begin();

myservo.attach(10); // initialize the filter for (ii=0; ii<9; ii++) { v[ii] = 0.0; w[ii] = 0.0; } currval = 0.0; maxval = 0.0; smoothval = 0; smootharray[0]= smootharray[1]= smootharray[2]=0; lastmax = millis(); Serial.begin(9600); // set up the MPU sensor.initialize(); myservo.write(0); delay(15); gain = 4.0; // higher -> more sensitive

// initialize the relay pin as an output: pinMode(relay1Pin, OUTPUT); pinMode(relay2Pin, OUTPUT);

//reset linear actuator position

while(analogRead(sensorPin)>0) //sets { digitalWrite(relay1Pin, LOW);

29

digitalWrite(relay2Pin, HIGH); }

digitalWrite(relay1Pin, LOW); digitalWrite(relay2Pin, LOW);

}

void loop (){ // read mpu data sensor.getMotion6(&ax, &ay, &az, &gx, &gy, &gz);

// scale ax and feed to filter mag = highstep(lowstep(ax/3276.8));

// find max value in current time window if (millis()-lastmax < 100) { if (fabs(mag) > maxval) maxval = fabs(mag); } else { currval = maxval; maxval = 0.0; lastmax = millis(); }

smoothval = 0.9*smoothval + 0.1*currval; //Smoothval courtesy of Ben Greer smootharray[counter]= smoothval;

Serial.print(smoothval); Serial.print("\t"); Serial.print(millis()); Serial.print("\t"); CurrentPosition = analogRead(sensorPin); //reads potentiometer values from actuator //gives current position. Serial.print(CurrentPosition); Serial.print("\t"); Serial.println(VAL); if(VAL==false && smootharray[counter]>0.7 && smootharray[counter-1]>0.7 && smootharray[counter-2]>0.7 ) // the array is used for coincidence detection, if three // successive values are above threshold the actuator is activated { VAL=true; goalPosition = 200; }

if (VAL== true && goalPosition > CurrentPosition) //if the leg hasn’t extended to the goal it //extends

{ digitalWrite(relay1Pin, HIGH); digitalWrite(relay2Pin, LOW); Serial.println("Extending");

}

else if (VAL == true && goalPosition <= CurrentPosition) //if the leg has extended to the goal //it then retracts to LOCK THE LEG by reversing voltage sent to actuator

{ goalPosition =0; digitalWrite(relay1Pin, LOW); digitalWrite(relay2Pin, HIGH); } if (VAL==true && CurrentPosition==0)

30

{ VAL = false; //if the leg has retracted fully and there is no additional user input the leg //stops. } delay(1);}

// given a new value x, step the filter forward and return the newest filtered value// code generated by http://www.schwietering.com/jayduino/filtuino/// 5th order Butterwoth filter centered on 13 Hzfloat highstep(float x) //sampling rate to change to 50Hz after adding motor{ v[0] = v[1]; v[1] = v[2]; v[2] = v[3]; v[3] = v[4]; v[4] = (7.346926241632e-1 * x) + ( -0.5397732726 * v[0]) + ( 2.4906544831 * v[1]) + ( -4.3389153270 * v[2]) + ( 3.3857389040 * v[3]); return (v[0] + v[4]) - 4 * (v[1] + v[3]) +6 * v[2];}

float lowstep(float x) //class II{ w[0] = w[1]; w[1] = w[2]; w[2] = w[3]; w[3] = w[4]; w[4] = (1.774798045630e-3 * x) + ( -0.2883883056 * w[0]) + ( 1.5067203613 * w[1]) + ( -3.0219304176 * w[2]) + ( 2.7752015933 * w[3]); return (w[0] + w[4]) +4 * (w[1] + w[3]) +6 * w[2];}

8.1 Appendix B (Risk Analysis and Ethical Considerations)

8.1.1 Scope

This document fulfils the requirements laid down in the Quality Procedure ‘Risk Analysis & Management ‘relating to the initial identification of hazards and the risk classification of those hazards. The document is prepared according to ISO 14971:2007 ‘Medical Devices – Application of Risk management to Medical Devices’. This version is based on analysis at the commencement of the design process after risk reduction to ensure that risks have been minimised and that design changes have not introduced new unacceptable risks.

31

8.2.2. Intended Purpose and Identification of Characteristics of Device

1) What is the intended use and how is the medical device to be used?

AThe above-knee prosthesis is designed to join to the remaining limb and be used for walking on flat surfaces.

2) Is the medical device intended to be implanted?

A No

3) Is the medical device intended to be in contact with the patient or other persons?

AYes, the socket will contain MMG sensors. This component will remain in contact with the patient’s skin for the duration of use.

4) What materials or components are utilized in the medical device or are used with, or are in contact with, the medical device?

ASilicon inner sock lines the socket and is in contact with the skin.

5) Is energy delivered to or extracted from the patient?

A To be determined.

6) Are substances delivered to or extracted from the patient?

N/A No.

7) Are biological materials processed by the medical device for subsequent re-use, transfusion or transplantation?

N/A No.

8) Is the medical device supplied sterile or intended to be sterilized by the user, or are other microbiological controls applicable?

N/A No.

9) Is the medical device intended to be routinely cleaned and disinfected by the user?

ACleaned by user to appropriate degree after use to maintain hygiene of skin in contact with prosthesis

10) Is the medical device intended to modify the patient environment?

ADesigned to improve the patient’s quality of life by (re-)enabling walking.

11) Are measurements taken? A

The MMG sensors detect muscle vibrations within the remaining limb to stimulate aided motion of the prosthesis following the patient’s Gate Cycle.

12) Is the medical device interpretative? AYes, with Arduino. The MMG signal is processed.

32

13) Is the medical device intended for use in conjunction with other medical devices, medicines or other medical technologies?

AProsthesis requires the addition of a compatible prosthetic foot that is not provided with the product by design.

14) Are there unwanted outputs of energy or substances?

AEnergy loss will occur through the motion of the motors that drive the damping and locking system of the knee

15) Is the medical device susceptible to environmental influences?

AProsthesis need to be designed to be dust and water resistant

16) Does the medical device influence the environment?

N/A

17) Are there essential consumables or accessories associated with the medical device?

A Prosthetic Foot

18) Is maintenance or calibration necessary? A

Yes, each prosthesis must have its system calibrated to the patients’ Gait Cycle and the socket must also be moulded to fit the remaining limb. Length of staff (tibia) must be set to an appropriate length.

19) Does the medical device contain software? A Yes, (Arduino) physical programming

20) Does the medical device have a restricted shelf-life?

A

Product should have a long shelf life, though specific components may need replacing if they fail. Environment dependant life span.

21) Are there any delayed or long-term use effects?

A

Wear will occur between the aluminum barrel and the stainless steel wire rope constituting the locking mechanism.

Product has delay time to response of C.N.S which will require the patient to adapt to this and modify walking slightly.

22 To what mechanical forces will the medical device be subjected?

A

Weight of Mass of body (100kg)

0.49Nm/kg (Torque generated while climbing up the stairs)

33

23) What determines the lifetime of the medical device?

A

Environmental conditions, i.e time worn for each use, terrain used when wearing prosthesis, weight, intensity of activity during use.

24) Is the medical device intended for single use?

N/A Designed as a long-term walking aid

25) Is safe decommissioning or disposal of the medical device necessary?

AYes, electronic components and battery must be disposed of correctly.

26) Does installation or use of the medical device require special training or special skills?

A

Yes, the socket must be moulded to the patient, this requires specific training for the moulding process. This is only required to be done once at a clinic. All home maintenance and training will also be provided at the clinic during fitting.

27) How will information for safe use be provided?

A User Manual and Clinic Training.

28) Will new manufacturing processes need to be established or introduced?

N/AUse of Tegris® to replace expensive carbon fibre socket.

29) Is successful application of the medical device critically dependent on human factors such as the user interface?

AYes, incorrect Gate Cycle analysis can lead to malfunction of the device /prosthesis

29.1) Can the user interface design features contribute to use error?

N/ANo user interface and patient does not have access to the software.

29.2) Is the medical device used in an environment where distractions can cause use error?

AYes, change in ground stability/ human etc. can cause failure of the prosthesis or loss of balance of the user.

29.3) Does the medical device have connecting parts or accessories?

A Prosthetic Foot

29.4) Does the medical device have a control interface?

N/A No.

29.5) Does the medical device display information?

N/A No

29.6) Is the medical device controlled by a menu?

N/A No.

29.7) Will the medical device be used by persons with special needs?

A Yes.

34

29.8) Can the user interface be used to initiate user actions?

N/A No user interface

30) Does the medical device use an alarm system?

N/A No.

31) In what way(s) might the medical device be deliberately misused?

A Weapon. Carrying Device.

32) Does the medical device hold data critical to patient care?

A Yes,

33) Is the medical device intended to be mobile or portable?

A Yes. Literally AIDS mobility

34) Does the use of the medical device depend on essential performance?

ARequired strong upper body and core to help balance and direction using prosthesis.

8.1.3 Initiating events & circumstances

35

36

8.1.4 Identification of hazards and estimation of risks

3. Evaluation of Risk Acceptability

8.1.5 Risk Acceptability

37

8.1.6 Skin Damage

The patient should read the instruction booklet on how to clean their transfemoral prosthesis to ensure preservation of their skin. They should ensure to keep their sock clean and dry at all time to avoid damaging the MMG sensors. The socket would ideally made of Tegris as it is a cheaper alternative to carbon fibre. The patient’s socket would have to fit him or her specific dimensions to ensure a proper backpressure on the end of the stump. This will limit the skin damages due to lack of terminal pressure in the end of the limb. The phenomenon can be explained by the lack of muscle that will pump blood away from the distal tissues. The socket has to distribute the pressure evenly along the stump to decrease the chances of oedema.

38

Figure 21: Causes of skin damage in transfemoral amputees

8.1 Appendix C (Business Case and Targeted Consumer)

8.2.1 Business case The target market for our prosthetic is amputees in developing countries. Currently there is no

affordable active prosthetic available, and such a prosthetic would have a large consumer base. The most advanced knee prosthesis available for amputees in developing countries is the passive Remotion knee that is currently retailed at 80$12 . Because many of these amputees are in rural areas, data is lacking and the number of amputees in need of prosthetics is unknown (the Remotion project estimates 24 million people are in need of a modern prosthetic)12. However land mines alone are responsible for 26,000 amputations every year13. Our active prosthetic would thus offer an advanced and reliable alternative to often crude, home-made prosthesis and the passive Remotion8

for thousands of people.

8.2 User Manual

8.3.1 Setting up the Prosthetic ° Once the knee is fitted and sealed, power the leg using the two switches fitted to the lithium batteries.

° Allow the actuators to set.

° Once the leg is set allow the health technician to perform a battery of tests to set:

a. Threshold values for actuator activation.

b. The duration of actuator activation and the extent of knee relaxation.

These parameters will depend on the user’s personal preference and physique.

To set these parameters connect the Arduino to a personal computer using the USB cable provided. Open the Arduino software downloadable at: http://www.arduino.cc/en/Main/Software. The software file provided with the Arduino should be modified in the following manner. To change the threshold for actuator activation alter values highlighted in red (a higher value means a greater contraction is needed to activate the actuator). To change the extent knee relaxation alter the value highlighted in blue (a greater value increases relaxation).

Figure 22: Code to tailor to user (found at lines 107 to 114)

39

8.3.2 Charging° To charge, use the provided lithium battery chargers, alternatively plug a voltmeter into the battery (+ pole to - pole and vice versa) and provide a voltage slightly higher than that indicated on the battery packaging.

8.2.3 Safety and maintenance° Ensure no rust accumulates around hinges. Clean the inner socket regularly using a cloth and adequate cleaning material as given by the socket manufacturer.

° If there is substantial accumulation of moisture in the socket or within the prosthesis, power off, remove from leg and allow to dry.

° Power off when not in use to ensure battery longevity.

° Do not use in large bodies of water.

° Ensure any waterproofing textiles are replaced when needed to protect any electronics they house.

° Visit your health technician as prescribed for maintenance.

8.2.4 Troubleshooting If the MMG experiences repeated issues with muscle contraction detection, actuator

activation or any other issues please hit the reset button on the Arduino:

Figure 23: Arduino Diagram with components labelled 14

If the problem persists please contact your health technician.

40

8.3 Appendix C (Group Working)

8.3.3 Team Organization

The group was divided into two teams of five. One worked on the structure of the knee the other worked on developing the MMG device.

8.3.4 Schedule

The group met weekly with Dr. Southgate to discuss new ideas and bring up problems faced by the group. Each group worked on an independent schedule in separate labs.

Figure 24: Timeline of the project with time unit in weeks

41

Period Highlight: 1 Plan Actual % Complete Actual (beyond plan) % Complete (beyond plan)

PLAN PLAN ACTUAL ACTUAL PERCENTACTIVITY START DURATION START DURATION COMPLETE PERIODS

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

Locking Mechanism Idea 1 11 1 13 100%Frame Idea 1 4 1 15 100%November Presentation PowerPoint 9 3 10 2 100%CAD Design 4 12 5 21 100%MMG Testing 4 20 4 22 100%Orders of Materials 5 1 6 3 100%Electrical Component Order 3 15 4 20 100%Batteries Order 16 2 18 2 100%Force Analysis of the Frame 5 19 5 4 100%Force Analysis of the Locking Mechanism 11 5 11 27 100%Poster 22 4 24 2 100%Poster Printing 24 1 25 1 100%Report Writing 26 12 32 6 100%

Project Planner

8.6 Appendix D (Initial designs)

8.5.1 Initial frame designs

Figure 25. Framework design with hollow square vertical supports and horseshoe shaped flat plates. The bar intended for attachment of wire rope ends was to be removed and the wire rope would go into the

holes on the lowest flat plate and be fixed in place with bolts.

Figure 26. Framework design with hollow cylindrical vertical supports and rings as flat plates. Note the two marked pieces on the barrel clip that needed to be welded on if the design was to be manufactured. The wire

rope would go through the hole on the flat plats.

42

8.5.2 Teeth Locking Mechanism:

Our early designs involved a cable rotating a pulley or a barrel to induce a knee rotation. One end of the cable would be attached to the middle aluminium plate of the prosthesis, while the other end would be connected to a spring and to a linear actuator connected in series. The spring was conveniently chosen to smoothen the response of the device. As in our final design, the linear actuator would apply or release tension is the cable to cause the flexion or the extension of the knee. A multiple tooth lock would lock the device to prevent any rotation in the sagittal plane.

A circular plate would be free to rotate under the influence of the cable and the linear actuator. A quarter of a circle plate is placed on the same axis about which the free to rotate plate is rotating, such that it is no rotation is induced by the linear actuator. The static plate is equipped with cylindrical teeth placed at regular intervals along its curved edge. The rotating plate has holes along its edge spaced by regular intervals equivalent in distance to the ones on the static plate.

43

Figure 27: Diagram of the Teeth Locking Mechanism

Extension

Flexion

Aluminum Plate

Stainless Steel Cable

Linear Actuator

Spring

Aluminium Barrel

Torque

Nut Static Plate with Holes

To preserve the battery’s life, the device would be locked when the linear actuator is not building up or releasing tension in the cable. In the locked position, the static plate would prevent the rotation of the free-to-rotate plate, as the cylindrical teeth are engaged in the holes. It was decided that pushing the two plates towards each other with springs would not be reliable enough and that the teeth might not lock properly, which would induced a failure of the locking procedure. The two plates would be brought against each other with the mean of electromagnets or electromechanical solenoid. An electric current induces the magnetic field of the electromagnets, so they need to be continuously connected to a power supply while the mechanism is unlocked. The electromagnets would induce a force of attraction on the plates that would dislocate the teeth from the holes and allow rotation of the main plate. The magnetic field, induced by the electromagnets, needed to dislodge the two plates can be calculated thanks to Ampere’s Law, considering the material used for the magnetic core of the magnet.

Ampere’s Law for an electromagnet:

¿=B (Lcore

μ+Lgap

μ0)15

44

Nut Static Plate with Holes

Figure 29: Schematic of the Relative Motion between the Stating and the Rotating Plates

Plate with Teeth

Plate with Holes

Where, N is the number of wire turns, I is the current going through the wires of the electromagnet, B is the magnetic field in the core, Lcore is the length in the core, Lgap is the length in the air gaps, μ is the magnetic permeability of the core, and μ

0 is the permeability of free space.

The force generated by the electromagnet thanks to its magnetic field can be calculated using the following formula, where A is the cross-sectional area of the core:

F=B2 A2μ0

15

The strength of the magnetic field can be controlled by the diameter of the core material, and by the number of wire turns around the core. Increasing the current I can also increase the strength of the magnetic field but when I is doubled, the heat generated by the system will increase by a factor of 4. Indeed, according to the Joule’s law of heating, where I is the current passing through the conductor, R is the resistance of the conductor and t is the time during which a current I is going through the conductor:

Amount of heat produced=H=I 2×R×t 17

45

Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16Figure 29: Structure of an Electromagnet 16

So if we double the current for a given resistance R and a given time t:

if I '=2 I

H 'H

=(2 I )2

I 2 =4 I 2

I 2 =4

When the current is doubled through a conductor, the heat produced is increased by four times. This could cause damages in the prosthesis or an uncomforting feeling for the patient. The current I would therefore not be a controllable variable.

This mechanism has the advantage of precisely lock the rotation of the knee at certain angles of interest and to be easily controlled by adjusting the value of the magnetic field. Nevertheless the precision to be attained in order to engage the teeth in the holes would need to be high, and the group did not think that such a level of precision could be reached. Moreover, the electromagnets require their own power source, distinct from the battery powering the Arduino board. The Arduino would have to control both the activation of the electromagnets to unlock the mechanism and the linear actuator that will cause the rotation of the plate with the holes. This would add a level of complexity to the Arduino code, the group therefore decided to look for an easier design for which the Arduino board would only have to control one output: the tension applied in the cable by the linear actuator.

8.4.3 The Locking Ring Mechanism:

46

Linear Actuator

SpringTorque

The Locking Ring Mechanism was one of our earliest ideas for the locking mechanism of our above-knee prosthetic. It would be implemented thanks to the use of a locking ring, a cable and a linear actuator. The spring was designed to smoothen the swing response of the prosthesis during the Gait Cycle. The linear actuator would be controlled by the Arduino to apply tension in the stainless steel cable. As the tension builds-up in the cable, the locking rings would be subject to an increase pressure on its surroundings. This increase in pressure would cause the ring to lock the rotation of the knee by entering in collision with a hardstop. The Locking Ring Mechanism would require the system to use power to lock the knee, and therefore the knee would be in an unlocked state when the power is off. The group decided that when the battery is out of power, an unlocked prosthesis could jeopardize the safety of the patient. The patient would not be able to stand up right if his or her device was running out of battery because the locking mechanism could not assure the full extension of the knee. The design had to be modified to allow the knee to be locked in its full extension position when no power could be supplied to the linear actuator.

8.7 References

1. Asterisk Cell Knee Brace Sizing Chart. http://www.asteriskbrace.com/sizingchart.html. Last accessed 16th June, 2015

2. Aluminium Alloy 6082 - T6 Extrusions, Aalco Metal Ltd. Last revised 03rd December 2013. 3. Bending of plates 3 Wikipedia. https://en.wikipedia.org/wiki/Bending_of_plates. Last accessed

47

4. Overview of materials for 5000 Series Aluminum Alloy, MatWeb Material Property Data . http://www.matweb.com/search/DataSheet.aspx?MatGUID=c71186d128cd423d9c6d51106c015e8f Last accessed 16th June, 2015

5. Kastal 300 Data Sheet, Smiths Metal Centres 6. Factors of Safety, The Engineering Toolbox http://www.engineeringtoolbox.com/factors-safety-fos-d_1624.html. Last accessed 16th June, 2015

6Engineering Mechanics Volume 1, STATCS, J. L. Meriam, J. Wiley & Sons, pp 301-302, 1978.

7 Proceedings of the World Congress on Engineering and Computer Science, (2009), Vol Io WCECS 2009, October 20-22, 2009, San Francisco, USA, pp758-788.

o Available at: http://www.iaeng.org/publication/WCECS2009/WCECS2009_pp785-788.pdf

8Ava D. Segal, Michael S. Orendurff, Glenn K. Klute, Martin L. McDowell, Janice A. Pecoraro, Jane Shofer Joseph M. Czerniecki, (2006) Kinematic and kinetic comparisons of transfemoral amputee gait using C-Leg® and Mauch SNS® prosthetic knees, Journal of Rehabilitation Research and Development, Volume 43 Number 7, November/December 2006, Pages 857 — 870

9 Paul DeVita, Tibor Hortobágyi, (2003), Obesity is not associated with increased knee joint torque and power during level walking, Journal of Biomechanics , Volume 36, Issue 9, September 2003, Pages 1355–1362

o Available at: http://ac.els-cdn.com/S0021929003001192/1-s2.0-S0021929003001192-main.pdf?_tid=c4f258ce-14f3-11e5-ae7b-00000aab0f02&acdnat=1434547471_7d523cf23dfaf8d82cfc52d034d138d8

10 Toolbox.com, Factors of Safety: FOS are important in engineering design.

o Available at: http://www.engineeringtoolbox.com/factors-safety-fos-d_1624.html

11 Texas Instruments (1986), SN754410 Quadruple Half-H Driver , SLRS007C –NOVEMBER 1986–REVISED JANUARY 2015.

o Available at: http://www.ti.com/lit/ds/symlink/sn754410.pdf

12 D-rev.org, Remotion kneeo Available at http://d-rev.org/projects/mobility/

13Erin Strait (2006), Prosthetics in Developing Countries, January 2006.

o Available at (http://www.oandp.org/publications/resident/pdf/DevelopingCountries.pdf).

14 Christopher Stanton, Getting to Know Arduino: Part 1: Hello, World!, Posted on Element14.com,

o Available at: http://www.element14.com/community/groups/arduino/blog/2014/03/28/getting-to-know-arduino-part-1-hello-world

15 Feynman, Richard P. (1963). Lectures on Physics, Vol. 2. New York: Addison-Wesley. pp. 36–9 to 36–11, eq. 36–

26.

16 Wikipedia.com (N.a), Electromagnet

o Available at: https://en.wikipedia.org/wiki/File:Electromagnet_with_gap.svg

17 Massachusetts Institute of Technology (2012), Circuits and Electronics : Joule’s Law, Last modified: Mar 23, 2012, 03:37 AM. Available at : https://6002x.mitx.mit.edu/wiki/view/JoulesLaw/

48

8.8 Acknowledgments

The project would not have been possible without the guidance of Dan Greer (University of Colorado) and Paschal Egan’s invaluable knowledge of PCB design and circuit design. Thank you to Satpal Sangha for his invaluable knowledge in manufacturing and for his help building the prosthesis. We would like to thank Dr. Southgate for his weekly advice and support, the Imperial Department of Bioengineering for allowing access to their laboratories and Rio Tinto for their funding and work space.

49