We are working to develop open source alternatives to some of the components that frequently fail on oxygen concentrators and make these designs available to the biomedical engineering community.

We have created a short video demonstrating each of the steps required to replace the zeolite in oxygen concentrators sieve beds.  In this video we demonstrate the processes of opening a sieve bed taken from a DeVilbiss 515, filling with new zeolite, compacting the zeolite by vibrating the bed using a concrete poker (vibrator), resealing the sieve bed using a custom cap we manufacture from PVC using our CNC router, and finally pressure testing the finished sieve bed to 40 psi to ensure there are no leaks.

Zeolite replacement in sieve beds

Problem:  Sieve beds used in PSA oxygen concentrators contain a chemical called zeolite, which binds to nitrogen under pressure, resulting in a stream of gas that is almost pure oxygen.  Over time, zeolite looses its ability to bind the nitrogen resulting in decreased purity of oxygen.  Concentrators operated in high humidity environments also loose experience a decrease in performance when the zeolite gets contaminated with moisture.  Importing replacement sieve beds is expensive.

Solution:  Rather than replacing the entire sieve bed we just replace the zeolite.  This is considerably less expensive  for several reasons: 1) we only pay for the chemical and not the canister,  2) the zeolite is purchased in bulk, so there re economies of scale,  3) the shipping costs are cut by two-thirds as shipping is largely charged by weight.  The process is not complicated, and takes about 20 minutes per sieve bed, once you get beyond the learning curve.  However, it should be done in a humidity-controlled room.  This can be achieved using a standard air conditioning unit set to "dry" mode (we achieved 30% relative humidity in our dry room).  Zeolite can be purchased directly from vendors in China and sent by air freight (it is not considered dangerous good), or by surface freight to conserve costs.

Adapting old sieve beds for reuse

ProblemSeveral manufacturers use sieve beds that were never designed to be refilled.  Both ends are sealed with aluminium caps that are pressed on by a machine and cannot be removed.

SolutionWe have developed a process for cutting the aluminium cap off the top of the sieve beds.  The device uses a small lathe chuck to rotate the sieve bed slowly while a cutting disk is used to penetrate the lip of the aluminium cap.  After two or three revolutions the cap is cut free.  Once the zeolite has been replaced we install a PVC cap manufactured using our CNC router.  Cap design varies to suit the model of sieve bed.

Built-in oxygen purity display on concentrator front panel

Solution:  If we are able to display the actual purity then clinicians would be able to make more informed decisions about where or not to use the concentrator and in what situations.  For example, a concentrator that is performing poorly at 5lpm may perform adequately at 1lpm flow, which would be sufficient for most infants.

There are many different ways to measure oxygen purity.  One of the most common sensors uses ultrasound to measure the speed of sound in the gas flow.  Since the speed of sound in oxygen is 326 m/s and in room air is roughly 344 m/s, the relative proportion of oxygen in the gas flow can be calculated.  Many handheld spot check analyzers use this approach.  The actual sensor is relatively inexpensive (~$17).  One of these sensors is pictured here (top).  To prototype our solution we tested this sensor by integrating it with an Arduino micro-controller and an LCD display pictured here (bottom).  We compared the reading to commercially purchased oxygen analyzer and found they were within 1% of each other.  To reduce the cost of the overall solution we are currently integrating the sensor with a PIC32 micro-controller (less expensive and smaller than an Arduino) and a 2 digit LCD display.

Handheld Oxygen Analyzer

Problem:  Over time, the purity of oxygen from a concentrator decreases.  Handheld analyzers are required to do spot checks periodically.  However, almost all hospitals in Malawi report that they do not have a unit.

Solution:  we have manufactured an inexpensive handheld oxygen analyzer for sport checking concentrators.  We plan to distribute these to all Central and District Hospitals in Malawi.  The prototype uses the ultrasonic sensor shown above in conjunction with an Arduino Nano micro-controller unit and a 128 x 64 pixel OLED display.  The unit is powered by a locally available 9 Volt battery.  The push button switch powers the unit on.  The button needs to stay engaged to keep the unit operational.  This design decision was made with the goal of extending the battery life as much as possible.  The next version of the analyzer will replace the Arduino Nano with a PIC32 Micro-controller.  Production quality circuit boards will be ordered from JLCPCB.

Many concentrators already have sensors

Problem:  Most oxygen concentrators  have a warning light that is illuminated when the purity of the oxygen is below a certain threshold (commonly 84%).  Some have audible warnings as well.  We have observed a common practice is that clinical staff continue to use the concentrators when the warning light is on, as they believe that low oxygen is better than no oxygen, and while this is less than ideal, such are the realities of the setting.  One problem with this approach is that the alarm does not differentiate between concentrations of say 83% and just room air (21%) for example, and in reality, depending on the age of the concentrator, the purity could be anywhere within that range.

Looking at the electronic control boards in many oxygen concentrators a sensor of this kind can be seen, as visible in the image to the left.  It appears this is the approach used to control the warning light/alarm.  Since the control board already employs a micro-controller, the only changes required are in software and the addition of a 2-digit display to the control panel.

Developing a generic electronic control board

Problem:  The electronic control board (commonly referred to as logic boards) is a common single point of failure.  Damage is often caused by surges on the national power grid.  It is difficult to procure replacement boards, and for older models of concentrators boards are no longer available.

Solution: To address this problem we have developed a generic replacement control board that can be manufactured locally at a fraction of the cost (Prototype board developed to test proof of concept shown to the left).  We are currently in the process enhancing our design to add a telemetry module that will allow remote monitoring of the concentrators while in use, generating real-time alerts when an error occurs.