1. The problem our project solves

Our team has decided to tackle the problem of the shortage of FFP2 or N95 class masks, which are the best means today for healthcare staff to carry out their jobs (i) without exposing themselves to hazards, and (ii) without endangering patients.

2. The solutions we bring to the table

We propose a disinfection solution for FFP2 or N95 masks using the re-adaptation of basin washers - present in all healthcare facilities.
We also propose using a physics-guided machine learning approach to predict how a given mask's protective power degrades over time while incorporating uncertainties. It would enable informed decision-making regarding the optimal wear time for each model of mask in a way that guarantees a good level of protection and a maximization of the product’s lifetime.
You could consider these two solutions independently. To get more in-depth knowledge about the subject and the model we used, you can check our GitHub repository !

2.A. Re-adaptation of basins washer for mask decontamination

First of all, it must be noted that the basins washer fulfills a set of specifications that make it particularly tolerant to ozone treatment being :

  • Hermetically sealed and does not leave any leaks of vapours (especially water vapours whose diameter is smaller than that of the ozone) produced in the chamber.
  • Not sensitive to corrosion (because capable of withstanding hydrogen peroxide treatments).
  • With adaptable compartments to fit masks.
  • Finally, programmed for drying cycles that are nothing more than fluid evacuation and dry heat heating cycles.

alt text

The basins washer are equipped with a clean water inlet and a waste water outlet. So we thought of :

  • Connecting an ozone generator (available in numbers on the net for a price of c. 400€ for systems used today in industry notably for disinfection and odour destruction) to the inlet initially planned for clean water. Let us recall here that an ozone generator only requires an electrical energy input, it also uses ambient air. Thus, electricity is the only consumable that is used!
  • Connect an ozone destructor pot (also available in numbers on the net for a price of c. 100€ for industrial quality systems that guarantee an output concentration of less than 0.01 ppm far above the safety standards and below the olfactory limit). Carry out a 20 min cycle at a concentration of 25 ppm in the chamber according to the results of our study above. This concentration is easily reached thanks to the generators available for the industry and the tightness of the enclosure.

alt text

  • Once the disinfection cycle has been completed, the laboratory protocols stipulate purges of 5 times the volume of the chamber in which the ozone has been diffused, which is easily achievable with the machine's waste water extraction system. Thus, given the capacity of the ozone destruction pots (c. 100 L/min) this action can be carried out in c. 10 min. We also propose to catalyse the destruction of ozone that would also be retained in the mask fibers to use a cycle initially designed for drying with a temperature of c.70°C (dry heat) demonstrated to have low effects on the integrity of the masks.

alt text

2.B. Predicting mask wear and an optimal time for taking off a mask

We propose here a complementary approach to assess optimal re-use of FFP2 or N95 masks while minimizing risks. To our knowledge, masks are already being re-used despite official guidelines due to operational constraints and limited available equipment. We aim to find an optimal total mask wear time as well as optimal scheduling of when to take off a mask for dehumidification, cleaning, and rest.In Bergman et al. , the performance of a mask is measured according to the penetration of nanoparticles of size from 10 nm to 600 nm that are not filtered and captured by the fabric. The mask is deemed to fail when penetration exceeds a given threshold such as 5%. In practice, it is not possible to test a mask’s performance in a live hospital setting during operations. Thus, it is necessary to build a predictive model for how a mask would respond if it were to be tested for penetration.Modern machine learning approaches often have trouble with lack of data, or overconfidence in predictions: this is not suitable for safety-minded applications with not a lot of data which is the case here. We propose a physics-guided machine learning approach,with the following design principles: rely on a physical model and incorporate predictive uncertainty.

alt text

As the wear model is set up, it could be supplied with data from manufacturers such as fibre charge density and diameter, layer thickness to compute the optimal wearing time of the mask so as to extend its lifetime without exposing the holder.

3. What we have done during the weekend

During this weekend, we conducted a large study in the healthcare facilities to find out (i) the practices of the staff faced with the lack of masks, (ii) the equipment available in all the heathcare facilities that could be used for mask decontamination and (iii) to have a knowledge of the processes in these structures, particularly in terms of decontamination/sterilization.
This work was carried out in parallel with a broad review of the scientific literature on the subject of coronavirus, masks (models, statistics on practices, etc.) and decontamination methods.
Subsequently, the team tried to build the model presented above based on robust physics hypotheses and data extrapolated from articles reporting, for example, the effectiveness of masks after different wearing times. Once the model project had been launched, we thought about the adaptation protocol described above and how to integrate it into existing hospital processes.
Our weekend was punctuated by calls to experts from our schools, or from the medical profession.

4. The solution’s impact to the crisis

Both of our solutions can have a large impact during this crisis.
The predicting mask wear and an optimal time for taking off a mask model, under the condition that a mask manufacturer provides the data requested by the model, could drive how to behave towards mask re-use. We know that the re-use of FFP2 or N95 masks is a common practice both in the medical world and also in civil society, the model could make this practice safer and solve serious sanitary issues.
The re-adaptation methodology presented here would provide a safe and inexpensive way to re-use FFP2 and N95 masks. As these are rare resources in healthcare facilities, re-use of this resource represents (i) a substantial saving of money, (ii) avoids exposure to mask shortages, and (iii) offers a mean to decontaminate other resources that are sensitive to current treatments such as 3D printed visors or over-jackets. So implementing the proposed solution in healthcare facilities would have a huge impact on public health. Indeed, it would enable healthcare workers to change masks as frequently as advised by safety guidelines without any risk of shortage, since used masks would be safely decontaminated and available for reuse. Above mentioned high risks of infection and adverse health effects associated with extended wear of face masks would then be avoided.
It would also reduce the amount of waste that is produced in healthcare facilities and hence have an ecological impact

5. The necessities in order to continue the project

To take our project to the next step, a week-long experiment for the validation of the efficiency of the ozone treatment on contaminated masks would have to be conducted in an accredited laboratory, we could suggest the GEEPS laboratory of CentraleSupelec in which we are used to work and possessing all the equipment and accreditations needed. We would then be able to build a prototype of our device by ordering all required equipment (basin washer, ozone generator, catalytic destructor pot) and by assembling components in a controlled environment. We could test its efficiency within a few days before starting trials in healthcare facilities.
In parallel, we would request factory data and products specifications from manufacturers to train our machine learning algorithm. We would also conduct physical data on filtration performance under varying humidity conditions in order to improve the fit between our model and experimental data.

6. The value of our solutions after the crisis

It would appear that bringing the use of the FFP2 and N95 mask into a new paradigm by making it reusable may have many effects after the crisis.
First of all, it is clear that the ecological impact of the crisis being reduced by the implementation of our solutions, the post-crisis period will be positively impacted.
Furthermore, post-crisis healthcare facilities will have at their disposition a means and a process for decontaminating material sensitive to the treatments proposed previously by these infrastructures.
Finally, as our model can be extended to surgical masks, it would allow for the optimal wearing of the largest classes of masks, making them more available to face a new outbreak of the epidemic.

Share this project:

Updates