Sensorized Facemask With Moisture-Sensitive RFID Antenna

Abstract

Due to the ongoing COVID-19 pandemic, the use of filtering facepiece respirators (FFRs) is increasingly widespread. Since the masks’ wetness can reduce its filtering capabilities, the World Health Organization advises to replace the FFRs if they become too damp, but currently, there is no practical way to monitor the masks’ wetness. A low-cost moisture sensor placed inside the FFRs could discriminate a slightly damp mask from a wet one, which must be replaced. In this letter, a radio frequency identification (RFID) tag exploiting an auto-tuning microchip for humidity sensing is designed and tested during an ordinary working day and a physical exercise. The tag returns about 1 unit of the digital metric every 3 mg of water generated by breathing and sweating, and it can identify excessively wet masks from commonly used ones.

I. Introduction

Face protections, like filtering facepiece respirators (FFRs), can significantly reduce the risk of infection caused by respiratory pathogens [1]. Because of the ongoing COVID-19 pandemic, the World Health Organization advises for the universal masking in health facilities, and the FFRs should be replaced as soon as they become damp [1]. However, such widespread use of masks can cause supply shortages, and their use must be optimized. As recommended by the USA National Institute for Occupational Safety and Health [2], the current practice is not to wear the same mask twice and replace it after a maximum of 8 h of continuous use. Indeed, moisture produced by breath causes an increase in the relative humidity (RH) of the interior microclimate and some condensation on the internal surface of the FFR. This was reported to decrease the filtering effect [3], [4]. Moreover, the moisture in the FFR's deadspace, (i.e., the volume between the mask and face) is linked to user discomfort [5].

To discriminate a dry or slightly damp mask from a wet one that needs to be replaced, a low-cost moisture sensor could be placed inside the mask to monitor its effectiveness. The moisture sensing could also be useful to optimize the decontamination cycles of the FFRs [6]. FFRs equipped with radio frequency identification (RFID) tags have been recently proposed for managing the medical supply chains [7]. RFID sensor tags for humidity can exploit the antenna performance variations, such as frequency shift [8] or power collected by the integrated circuit (IC) [9]. The price to pay for sensing is a degradation of the read range as humidity and moisture increase. A different approach considers instead a dedicated external sensor connected to the tag [10]. Although the communication performance is preserved, this solution involves higher complexity and cost, making the application into a cheap FFR rather impractical.

This letter proposes a facemask-integrated moisture sensor that only relies on a thin-wire antenna and an auto-tuning [11] IC so that this advanced sensing device can be low-cost and avoid degradations of the link range. The antenna is fixed onto a hygroscopic and hydrophilic fiber inside an N95 FFR. The tag's electromagnetic performance is optimized through numerical simulations comprising a model of the head and an N95 facemask. Hence, prototypes are manufactured and experimented in both regular and stressful activities.

II. Sensing by Auto-Tuning ICs

The RFID-based moisture sensor exploits the unique features of a new family of RFID ICs with an auto-tuning capability. They can dynamically change their internal impedance to maximize the power collected by the IC and compensate for a possible impedance mismatch with the antenna due to changes in the boundary conditions [11], here the amount of moisture on the facemask. Indeed, the mismatch is related to the absorbed moisture, which changes the substrate's permittivity locally and disturbs the antenna. The analog front end of an auto-tuning IC includes a varactor device that can be modeled as a resistance parallel to a variable capacitance [11]. The capacitance, in turn, can be considered as a switchable ladder of N identical capacitors having a total capacitance . spans from a minimum to a maximum value with incremental steps according to the law where n is the number of connected equivalent capacitors. The auto-tuning feature can be exploited for the indirect sensing of the status of the facemask, which is related to the internal IC capacitance by the following equation [12]:

where f is the frequency, and is the susceptance of the tag's antenna. Equation (1) holds for , whereas outside of this range, there is saturation [13]. Auto-tuning ICs return a digital metric, hereafter denoted as sensor code (SC), which is directly proportional to n by inverting (1) and accounting for the saturation

where nint(x) is the nearest integer to x. User-specific baselines can be removed through calibration with respect to the sensor code returned just after the placement on the face, thus obtaining a differential sensor code [14], . Since is unknown in real use, the relationship must be determined through an experimental calibration curve, as shown in the following.

III. Textile Moisture Sensor Tag

The sensitive substrate to collect the moisture produced by the breath is an inexpensive twill woven fabric (50% cotton, 50% nylon) of size 70 mm 25 mm 1 mm. The textile has hydrophilic and hygroscopic properties [15] and absorbs liquid due to both the RH and condensation. The twill woven substrate hosts a symmetric loop-matched [16] meandered dipole [Fig. 1(a)] that is connected to an Axzon S3 Magnus auto-tuning IC [13] with parameters , , mS, pF, pF, and power threshold dBm. Accordingly, the resulting IC susceptance can retune in the range  mS  mS at 866 MHz, and the corresponding middle-range impedance is . The resulting sensor tag is placed in the core region of the FFR [Fig. 1(b)], which is expected to be the wettest portion of the internal surface [17]. The added surface covers a small part of the mask, and the only expected consequence on the FFR's effectiveness is a slightly improved filtering capability.

Fig. 1.

(a) Layout of the loop-matched meandered dipole inside the N95 mask and exploded view of the tag. The loop is an aluminum 35 μm thick trace, whereas the dipole is made of a thin copper wire (diameter 40 μm). (b) Simplified anthropomorphic numerical phantom of the head used for the electromagnetic simulations, and tag's placement on the internal surface of the N95 mask.

The geometrical parameters of the antenna were optimized through numerical simulations (by CST Microwave Studio 2018) assuming it is attached on an N95 FFR made of polypropylene (ɛ = 2.2, tanδ = 0.002 [18]). Simulations accounted for the human head by an anthropomorphic homogeneous numerical phantom¹ (ɛ = 42.7, σ = 0.99 S/m [19]). The final antenna form factor constrains the input susceptance inside the IC's tuning range [the shadowed region in Fig. 2(a)] for 830 MHz 980 MHz. The achieved realized gain is higher than  dBi along the direction of the nasal septum [Fig. 2(b)]. Thanks to the auto-tune property of the IC, exhibits the typical nearly flat profile over a broad band. By assuming the reader emits not less than 22 dBm (low-cost handheld device) a read distance longer than 40 cm is expected. In this way, an operator with arms’ length longer than 60 cm [20] could monitor a third-person facemask's status while keeping a safety droplet-free distance of at least 1 m [1]. Moreover, as previously demonstrated in [10], this arrangement will be fully compliant with specific absorption rate regulations.

Fig. 2.

Simulated performance of the tag's antenna. The EU unlicensed 865–867 MHz band is highlighted in grey. (a) Admittances of the tag's antenna and the IC. The shadowed region indicates the dynamic range of the susceptance that the auto-tuning IC can match. (b) Realized gain of the sensor-tag along the nasal septum direction.

IV. Prototypes and Experimentation

A. Test in Controlled Conditions

Some tag's prototypes were manufactured by fixing the dipole, made of a thin copper wire, onto the fabric substrate through a semipermeable, transparent polyvinyl chloride film (Tegaderm, thickness 25 μm). The boundaries of the resulting tags were then glued inside N95 FFRs (Fig. 3). Three volunteers wore a prototype each, and the realized gain was measured in the direction of the nasal septum through a Voyantic Tagformance Pro station. The measured data (Fig. 4) compare well with the simulations, and the average difference is just 1 dB. The three prototypes were then placed inside a Binder MKF 56 climatic chamber to perform a test on the SCs when the ambient humidity is first kept at 10%, then raised to 100% and kept for 10 min. The ΔSC was measured by ensuring a fixed value of on-chip received signal strength indicator (set to 15 units) to avoid possible nonlinearities of the self-tuning mechanism [11]. For comparison, RH values were accounted for in the simulations by considering two permittivity values of the tag substrate (ɛ = 2 at RH = 10% and ɛ = 2.4 at RH = 100%) [15]. The measured ΔSC is slightly higher than the simulated one [Fig. 5(a)], probably due to the approximated model of auto-tuning ICs and the effects of condensation of moisture on the tag, not accounted for in the simulations. The sensor's response time is evaluated by stimulating the step response through a sharp exposure of 200 mg of physiologic solution sprayed inside a sensorized dry facemask [Fig. 5(b)]. The SC was continuously measured before, during, and after the event. The discontinuity was nearly instantaneously detected with just  s settling time, corresponding to a single sample.

Fig. 3.

(a) Prototype of a facemask tag glued inside an N95 FFR. (b) Zoomed-in view of one tag prototype.

Fig. 4.

Measurement of the electromagnetic performance of the tagged FFRs. (a) Arrangement in a semianechoic space. (b) Average measured and simulated (RH = 100%) realized gain over the three prototypes each worn by a different volunteer. Grey lines indicate the average measured values plus and minus the standard deviations. The EU unlicensed 865–867 MHz band is highlighted in grey.

Fig. 5.

(a) Measured and simulated ΔSC after 10 min in the climatic chamber. (b) Step response of the tagged facemask when 200 mg of a physiologic solution are sprayed over its interior surface.

B. Experimentation in Everyday Conditions

The volunteers wore the three N95 FFRs for one 8-h working shift. They interacted with computers and performed several nonchallenging physical tasks, like walking to a shop. The FFRs were occasionally removed for less than ten consecutive minutes, and the ΔSC was always checked before and after the removal to be sure that the measurements had not been affected. Three measurements per hour were performed by using a Caen qID mini handheld reader. The resulting ΔSCs (Fig. 6) fluctuated in the (−2, 10) range depending on the environment and the performed tasks. In particular, there is a cyclic profile with a 3-h periodicity in the three cases that is probably due to the balance between cyclic generation and evaporation of the condensed moisture inside the FFRs’ deadspaces. Consequently, the effective variation of the SC in regular use is modest and does not permit discrimination of relevant events. A more significant behavior was instead appreciated in stressful conditions, as described next.

Fig. 6.

Measured ΔSC during 8 h of continuous use of the tagged facemasks during everyday activities, compared with the simulated value.

C. User Under Physical Stress

The volunteers were asked to perform an intense physical exercise while continuously wearing the FFRs. They had to ride a stationary bike for an hour (maximum resistance, speed 16 km/h, indoor temperature 27 °C). The test was repeated with regular masks without the RFID sensors to evaluate the absorption capability of the sensor's substrate. In both tests, the masks were weighted after 30 min and then after 60 min of exercise. The weight increments of the tagless N95s, as measured during the second session, are consistent with values reported in [17], [21], and [22]. In comparison with the tagless masks [Fig. 7(a)], the properties of the tag substrate allowed the sensorized masks to collect an additional quantity of moisture (in average, an extra 27% of water after 1 h of physical exercise), especially during the first 30 min. Therefore, the sensitivity to breath is amplified, and the capability of predictive monitoring of the mask status is improved. Correspondingly, the measured ΔSCs are up to four times higher than in the previous experiment involving a weaker activity. Values look correlated with the weight of absorbed moisture, as shown in Fig. 7(b). There is a detectable difference between the two states which are well identified in all the cases. The ) profile (averaged over the three users) is nearly linear, and the estimated sensitivity is ) ≈ 0.35 units per mg of moisture. Finally, the reduction of realized gain () versus the condensed moisture during the exercise was of the order of just 0.1 dB per 10 mg. The maximum degradation in the extreme condition is 1 dB that converts in a 10% reduction of the read distance, at most. Accordingly, the electromagnetic performance stays almost stable during the increasing humidification of the mask.

Fig. 7.

(a) Measured ΔSC after 30 and 60 min of physical exercise with a stationary bike and the corresponding measured water moisture collected by the tagged and tagless FFRs. (b) Measured and ΔSC versus the moisture collected by the mask.

V. Conclusion

A textile integrated moisture sensor tag based on an auto-tuning RFID IC was designed for application onto an N95 facemask and tested by volunteers in realistic conditions. When the mask is worn, an initial calibration is required to reduce the user variability of the response. Afterward, tags discriminated the use of the FFRs during an ordinary working day (ΔSC ≤ 10) from a more demanding physical activity (ΔSC ≥ 15), which could resemble a caregiver's hectic day in a COVID-19 hospital where the device's ageing accelerates. Hence, the sensor could help to identify excessively wet FFRs to be replaced. With the current performance, qualitative information can be retrieved. However, since the available dynamic range of the SC was only partially used, there is a margin to improve the sensitivity by further optimizing the antenna layout. It is finally worth mentioning that the employed IC also provides temperature measurements that could be merged with the moisture information to achieve more accurate predictions of the mask's ageing and even produce data on the user's respiration, fever rise, and lung status [23]. These potentialities will be investigated in future works.