Breathing Assistance Apparatus

Assists a human with breathing malfunction to improve breathing day or night.

Similar projects worth following
Commercial alternatives retail from $1000 to $4000 depending on the software features necessary to help the user. Because of the huge costs of commercial treatment many people do not seek to treat themselves and thus deal with the significant physical and cognitive symptoms. This project will give those people another option.

GPLv2 hardware designs & software. All text/imagery on this project page are GPLv2. The license is in the GitHub repository linked.

3D printable radial centrifugal compressor and BLDC motor. Motor must be wound. Plans will be breadboard based. Arduino controlled.

* Energy efficient, 12v power supply for ease of off grid usage.
* Low noise.
* Low cost, aiming for < $100 materials.

This project will not include a humidifier, tubing, or a respiratory mask. However these features and others may be looked into in the future.

I was working on this for myself and discovered this contest in late July when looking for prior art. I am now using this as an opportunity to share my work on this wonderful platform.

To get some understanding of the unscrupulous status quo in the industry and some nomenclature read this and this. I refuse to use the terms 'CPAP Machine', 'APAP Machine', 'BiPAP Machine' and such in order to distance this apparatus from those marketing inventions.

The challenge this project addresses

There are a variety of breathing function afflictions a person may experience that are relieved with a radial centrifugal compressor, such as:

  • Lung cell oxygen adsorption deficiency [3]. Deficient oxygen is bad. Treatable with higher than atmospheric pressure "constant pressure".
  • Upper Airway Resistance Syndrome.
  • Obstructive sleep apnea. In a deeper state of sleep, muscles relax, and obstructions could occur. Treatable with "auto adjusting pressure". The pressure adjusts into appropriate rhythms until the obstructions stop.
  • Central sleep apnea. This is Ondine's Curse. In a deeper state of sleep, the body "forgets" to breathe.
  • Ondine's Curse during the day!
  • Pleural effusion. "Fluid on the lungs".

Note a person may be afflicted by two or more of the above things. In both forms of sleep apnea, the body notices low oxygen when deeper sleep is entered. First the heart rate will rapidly accelerate from say 50bpm to 100bpm in an attempt to fix the problem- which cannot fix this problem. The body then rises out of the deeper state of sleep in order to breathe again, and the heart rate returns to normal. The body then enters a deeper state of sleep and the cycle repeats again, over and over all night long. Hardly any deep sleep is obtained while exhausting the heart.

In children with sleep apnea, bed wetting may be a symptom [5].

Commercial diagnosis and treatment is out of reach for many people in the world. I've seen quotes of up to $1000-$2000 per sleep study. And $1000-4000 for the breathing device. $4000 for "ASV Ventilators" with algorithms capable of handling Central Sleep Apnea, cheaper devices don't handle that condition, and some people gain central sleep apneas as a side effect of treating obstructive sleep apnea. This is very expensive and out of reach not just for the typical person, but especially for someone with a disabling condition who experiences exhaustion all day every day due to their heart running a marathon for hours every night.

Further, the individuals affected are unable to perform non-trivial mental tasks either, as they experience brain fog and cognitive decline. As they never enter deep sleep, the brain cells never shrink to allow the cerebrospinal fluid (CSF) to flow through the brain to clear out toxic byproducts of the brain’s metabolism [1][2]. There will be a significant cognitive decline at onset, followed by gradual decline over the years affected. Upon using the ventilation device with appropriate settings, the affected’s physical symptoms will be relieved over the next few days where they finally get some sleep. However it takes a few years of usage for the brain to return to normal.

How this project will alleviate or solve this problem

By publishing completely open source designs to a low cost solution, and instructions on how to build it.

How the project might be world changing

Take Carrie Fisher for example. Sleep apnea contributed to her death on an airplane. One issue people who need breathing assistance experience, is if their device is approved or appropriate to use on the airplane or out camping, or even if they have a battery setup for it. Maybe they were able to purchase a device usable at home, but they can't afford or justify spending another $4000 to obtain for one suitable for travel (they aren't so interchangeable, one marketed "brick" may have a battery but no humidifier, and is thus unsuitable for use at home). On top of this, the advice is to buy a new machine every 6-8 years due to the increased...

Read more »

  • 1 × Arduino Uno (ATmega328P) $24.95. May use the Arduino Micro (ATmega32U4) instead if the features are sufficient.
  • 1 × TI DRV10975 (IC) $3.81. 25W, 12v 3-phase sensorless BLDC motor driver. Uses a proprietary sensorless control scheme to provide continuous sinusoidal drive, which significantly reduces the pure tone acoustics that typically occur as a result of commutation.
  • 1 × BMP280 (Adafruit Board) $9.95. Temperature, barometric pressure sensor. 0.16+/-0.2 Pa resolution up to 26.7 Hz.
  • 1 × Half size breadboard $5
  • 1 × Breadboard wire pack $4.95

View all 14 components

  • Status: Modelling Impeller in Blender

    Andrew Smart09/03/2017 at 08:50 0 comments

    To ensure I have enough time to work on the electronics/code by the end of this contest I have decided to put off analysis of the efficiency curve, and work on an impeller with a maximal efficiency point of 88.5% at 4cmH2O design pressure. I've chosen 4cmH2O to get a smaller impeller- to reduce on prototyping costs if I need to change it later.

    5mm radius hub, 2.1233cm radius shroud inlet, 1.5cm blade outlet width, 3.5712cm axial height, 5cm radius impeller, and outlet Beta2=89.63°. 2600rpm for 4cmH2O.

    The above picture is some scratch work, to make sure Blender is sufficient.

    I'm used to working in SolidEdge, Maya, and Autodesk 3ds Max. I have not worked much with Blender. I've spent the past few days evaluating Blender, and NURBS support within Blender appears sufficient for this task [1][2]. I suppose I'll note two agenda items I still need to investigate in Blender: (1) angle measurement w/ NURBS objects, and (2) NURBS control point -> curve translation. I desire these so that I can precisely model the hub/blade/blade-hub/blade-shroud curves instead of evaluating many points by hand (time consuming approximations).

    Perhaps I should instead model the blade in FreeCAD, then import to Blender for the duplication & rotation or for other functionality FreeCAD doesn't have. I'm making sure to use open-source/free tools so that everyone can replicate these steps when working on their own problems!

    Looks like I've found the answer to (2). Hopefully that plugin to FreeCAD is open-source somewhere, but it doesn't look like it. Regardless, that document is very helpful.

    I also have some concern with the Beta2 of the outlet triangle- the pressure rapidly increases as Beta2 nears the tangent of the impeller hub. This is important considering manufacturing tolerances (i.e. huge pressure difference between 89.2° and 89.6°).

    Here is an example of the blade using impeller.m, when varying Beta2.

    Notice that pressure rapidly increases as Beta2 approaches the tangent (180° above). I am concerned about non-uniformity between the 1.08mm thick 3D printed blades. The variation in pressure between the design & print would easily be corrected by the pressure feedback from the sensor, but I believe non-uniformity between the 3D printed blade tips would result in shockwaves (noise).

    Though now that I think about it some more, notches in the blade tips may reduce these shockwaves just like the notches would with the jet/wake shockwaves.

  • Compressor Design Research

    Andrew Smart08/28/2017 at 04:06 0 comments

    Compressor Design

    The impeller and bearing assembly are preferably sized and configured such that the air gap is maintained at a minimum (e.g., less than approximately 0.1524 mm) during rotation of the impeller. In this manner, the compressor assembly minimizes aerodynamic losses such as vane-to-vane losses or losses resulting from parasitic fluid eddies which can reduce the operating efficiency of the compressor assembly [3, p15].

    Doesn't need a fancy axial seal due to low pressure differential, low efficiency lost, and lost air will just cool the motor anyway [10][10] shows that discharge pressure can be relatively constant using speed control on the electric motor.

    Appears that trimming of blades by 120° maximizes efficiency [13].

    Impeller Design

    Requirements incrlude a flow rate up to 120 L/min and pressure 20 cmH2O or higher [3, p7]. May offer multiple compressor designs with peak efficiency at user pressure range [3, p8]. If the compressor is “too large” for the person’s lung capacity, I suspect with the flow necessary for the right pressure it is possible that the extra flow would exceed the mask’s vent capacity, and it would escape the mask by blowing onto the person’s face causing discomfort. This is a common complaint by users. A compressor appropriately sized for the person’s lungs and pressure range might mitigate this issue, this is something I need to research more.

    The rotating speed of a centrifugal compressor is an inverse function of diameter to maintain a desired peripheral speed at the outer diameters of the impellers regardless of the physical size of the compressor [10].

    The leakage of air past the turbine blades can cause a 10-15% loss of turbine efficiency, a closed shroud largely eliminates the issue, along with a running clearance as small as 0.508mm, leakage loss will be ~0.5% with a low hub/tip ratio [11]. An open shroud seems to be better for low flow rates, and closed for higher flow rates [L4, pp17-18]. A 3D printer likely won’t be able to make a closed shroud as a single part [could use expert opinion here].

    Partial blades known as splitter blades widen the choke margin while also reducing slip losses [L4, p15]. Because the relative velocity of the leading edge varies in the inducer between the impeller hub and shroud, modern inlet designs employ inlet blade angles that vary in the spanwise direction to maintain an optimum incidence angle along the leading edge [L4, p15]. 3D printers may have difficulty making these overhanging blade edges.

    It is proposed that optimal axial length is dependent on the impeller diameter and fluid coefficient [L4, p18]

    Inlet Whirl Design

    An inlet valve is used to keep pressure constant for a constant rpm [needs ref]. This could allow people to have their machine operate at the appropriate pressure, without blowing their mask off or print a more suitable compressor.

    Inlet whirl may may decrease or increase the pressure, but there are cons [L4, p20]. Principally the inlet whirl should be used as a noise dampener.

    Alternatively to an inlet whirl, a straight pipe may be used.

    Diffuser Design

    The gas exits the impeller at high velocity and enters a diffuser passage. In the diffuser, the gas velocity decreases and dynamic pressure is converted to static pressure. Diffusers can be either vaneless or vaned [10]. Vaned diffusers are quieter, but vanes only appropriate for system with constant speed [L3].

    Involute curves should be used for the diffuser, formulas here [8]. Diffuser length formula here [L3], note vaned diffuser can be much shorter.

    The pressure sensor would probably be best placed after the diffuser outlet, where the piping is the same diameter as the tubing attached to the mask. After any helmholtz cavities may be best as well to prevent any signal loss of the user’s breathing.

    Noise Reduction

    A spiral inlet would reduce noise escaping the compressor for this high pressure, low...

    Read more »

  • Compressor Design Status & Requirements

    Andrew Smart08/28/2017 at 04:05 2 comments

    Project Status

    Primitive analysis can be made in impeller.m, however it does not model losses and efficiency cannot be predicted. In my search for loss models, I discovered an excellent open-source project from Carleton University, Canada, for designing radial open-shroud centrifugal compressors, which uses a meanline approach. Once I have evaluated it completely for suitability I will link to the project, otherwise I'll have to move onto something else, such as one of the open-source turbomachinery projects in NASA's catalog.

    I'm able to use it to significantly accelerate the design. But even in the preliminary design of the impeller I still have to do a lot of "plumbing" work. I still have to design the diffuser, inlet whirl, and helmholtz cavity, but they should all be much easier as the impeller is the most critical piece to optimize.

    I currently have 84.7% efficient parameters for impeller design, but I must do a lot more plumbing such as varying the rpm, r2, blades, and separator blades, as well as making scripts for analyzing the data products to sort for efficiency.

    This software predicted real-world efficiency correctly within +/- 5% for a small set of compressors they built/tested. This could be improved by augmenting the software with additional loss models and other improvements, such as the proposed optimal axial length [L4, p18] (then again, maybe this is in the code somewhere).

    Compressor Complexity & Project Costs

    From [L2]:

    When designing a new pump, a designer has two choices: to design from scratch, or to model from other available (similar designs) with experimental results (e.g. compressor performance maps) and using the affinity laws to scale that design to the appropriate size. “Blank piece of paper” designs are rare, and are developed for special applications and extreme or unusual conditions.

    From a patent describing the field [3, p7]:

    Compressors used in [breathing ventilation apparatuses] must be capable of generating different flow rates depending upon the type of respiratory treatment to be provided as well as the respiratory condition and physiological size of the patient. For example, patients undergoing treatment can range from pre-term infants, neonates and pediatric patients up to full-grown adult patients. As may be appreciated, the pressurized gas requirements of a neonatal patient differ markedly from the pressurized gas requirements of a full grown adult. Flow settings for neonates can be as low as 2 liters per minute (LPM) at pressures as low as 5 cm H20 as compared to the flow settings for a full grown adult [male] patient requiring flow rates of up to 120 LPM and pressure settings of 20 cm H20 and higher.

    As a result of these differing flow requirements, different compressor assemblies are designed for use with a certain range of flow settings. The compressor assemblies are optimized to produce the desired flow requirements at maximum operating efficiency and with minimal power consumption. In this regard, a common practice in the industry is to develop and manufacture a specific compressor assembly which produces optimal flow characteristics for a specific set of patient types and/or flow settings. As may be appreciated, the need to design, test and manufacture completely different configurations of compressor assemblies for different patients having differing flow requirements substantially increases the overall cost of these devices.

    It should be clear that parameterized design of compressor assemblies will be very useful in the field.

    In order to optimize user pressure/flow matching the user’s lung capacity and pressure needs, either a set of unique impeller, diffusers, and motors would need to be developed, or a parameterized system that would take the flow/pressure requirements then output the geometry of the components for 3D printing.

    As I do not have access to an engineering CAD suite capable of 1D/2D/3D flow analysis, I have...

    Read more »

  • Motor Research

    Andrew Smart08/22/2017 at 23:07 0 comments

    Traditionally efficient motors use a mass produced insulated silicone-steel laminate for the stator (solenoid cores/legs). However it isn't cost effective to make these for small batches. This led me to search for a 3D printable ferromagnetic material, with minimal efficiency losses and minimal cost. The following are the only ones commercially available I have discovered at the time of this writing:

    • Proto-Pasta Iron PLA (FDM). Magnetic permeability of 6-10, magnetic saturation of .13T. Inexpensive.
    • Alumide (SLS). Magnetic permeability of 10-13. Material is 20% aluminum powder by weight. Most expensive, but more accurate- would be best for a minimal airgap between the stator and magnets.
    • PLA Aluminum (FDM). Material is 25% aluminum powder. Should have the highest magnetic permeability and magnetic saturation of these choices.

    The reason a high magnetic permeability and magnetic saturation is desired for the motor is it allows the motor to be smaller and more energy dense. The silicone-steel laminate has a magnetic permeability of around 6000 and a magnetic saturation of around 1.5-2T.

    Using purely the Proto-Pasta Iron PLA would require the stator to have a diameter of around 9cm and thickness of 6mm according to my rough calculations, with the rotor that would make a total of around 11cm diameter and 1.5cm thickness for the motor.

    A far higher magnetic permeability and magnetic saturation may be obtained by making our own polymer, at a lower cost. This will allow the motor to be much smaller, cheaper, and more efficient. I've seen explanations that a better ferromagnetic material has more of a "magnifying" effect on the solenoids/magnets, so you can use less powerful solenoids (less current/copper) and cheaper magnets.

    See this study from 1996 funded by the U.S. Office of Naval Research "Relative Magnetic Permeability of Polymeric Composites with Hybrid Particulate Fillers" (unfortunately the last page of figures is missing from this scan).

    Increasing ferromagnetic material added to the polymer increases magnetic permeability quadratically, not linearly. Though it is essentially linear up to around 15%-20% ferromagnetic filler (look at figures 1-2).

    Note especially figure 4, where they have experimental evidence backing their hypothesis that a hybrid of ferromagnetic fillers increases the magnetic permeability substantially higher than their individual sum. Magnetic permeability of >120 is reached with a LDPE polymer, 60% NiZn 60 µm diameter powder, and 20% Metglas 2705M flakes (% by volume).

    Inspired by a hackaday project where glue sticks were melted and re-formed in order to add glitter or color, I realized that this procedure could be used to add ferromagnetic fillers to the polymer. The inexpensive glue stick and ferromagnetic fillers will be melted in a small can, then slowly poured into a 3D printed "cookie cutter" stator. The ferromagnetic filler dust can easily be obtained with a small file and scrap iron/steel/aluminum, such as an aluminum can, steel can, or iron nail (the more types the better).

    The magnetic permeability and saturation of the filler can be characterized using this hackaday project. However such characterization's contribution probably wouldn't be worth the effort given this low repeatability context, but may be nice as an experimental estimate.

    I need to model the losses in the compressor before I'll be able to finish a model for it, and then I'll be able to accurately model the motor. I have more of a perfectionist/analyst personality so I strongly prefer to deeply analyze the problem before spending money on printing prototype models.

    I've since learned that the Shapeways shipping discount isn't too useful given the amount of material for these parts. Either the 10% off materials code or 3Dhubs competitors would be more effective, so I no longer have an end-of-the-month deadline.

  • Status

    Andrew Smart08/15/2017 at 23:06 0 comments

    Formulas to design the centrifugal compressor were obtained, however the losses still need to be modelled. The impeller is modelled in impeller.m, a Matlab/Octave script. The diffuser, inlet whirl, and outlet cavity pipe still need to be modelled. The geometry can then be modelled in Blender.

    The motor is modelled in dcmotor.m. Likely a 10 pole 12 slot motor will be selected due to low cogging torque (less noise). FEMM will be used for analysis. Will likely use Proto-Pasta's iron PLA for the stator core, unless a better alternative can be found. Possibly with a metal film wrapping to improve the magnetic saturation of the volume. Needs to be efficient at 5400rpm and 12.4mNm torque to match the 6.8cm impeller (this may change as losses are characterized).

    Hoping to get the part geometry all modelled by the end of the month in order to take advantage of the free shipping promo on the Shapeways 3D printing service.

View all 5 project logs

Enjoy this project?



Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates