COMITÉ DE SEGURIDAD Y SALUD OCUPACIONAL

The previous sections concentrated in the simulation of smell transport and the esti- mation of physically accurate concentration values for di↵erent discretisation levels of the computational domain. This section completes the olfactory pipeline by pre- senting the methodology for delivering realistic smell impulses to the VE users. This methodology is based on displaying smell cues using an olfactory display that deliv- ers physically accurate smell impulses based on the concentration values computed in the simulation stage. The full olfactory pipeline is utilised for the implementation of the experimental studiesE2 and E3 whose methodology framework is given in chapter 4.

An olfactory display was used to deliver the smell impulses to the participants of the experimental studies. This device follows the Lorig olfactory display design (described in section 2.4.3) and is composed of several components that can be configured to be close or away from each other and/or the user depending on the scope and needs of the application. For the purpose of conducting experiments, the majority of the olfactory display parts were hidden from the participants and only the smell outlets (tubes) were visible to the subjects so as to avoid any potential bias.

For the experimental studies of this thesis, an olfactory display with two channels/outlets was used. The flow rate of each channel is adjusted using a Digital Mass Flow Controller (DMFC) as a percentage of the maximum flow rate provided which is Fs = 1000 ml/min. One of the channels was used to release smell stimuli (citral) and the other as a control channel for releasing clean air and removing residual smells from previous trials during the experimental procedure. Figure 8.11 shows a set-up of the olfactory display during a testing session for measuring its

output concentration and flow rate. Testing sessions were used to calibrate the device and ensuring that physically accurate smell stimuli are to be delivered during the experimental sessions. During the experimental trials, the device was in compact form under the desk and only the two outlet tubes were visible and located in places near the user’s nose (see Figure 10.2).

As can be seen from figure 8.11, an air compressor is used to supply a con- tinuous air stream of pressure 2 bars to the olfactory display. The stream is filtered through an in-build air filter and directed to the two DMFCs. Each of the DMFCs contains a computer controlled diaphragm that allows any percentage of the incom- ing stream to pass. The modulated air is aromatised by passing through a reservoir where citral in liquid form is stored and released at the edge of the outlet tube. Lastly, a PID detector (see section 2.4.4) is used to record output concentration values. These values are directed to a Raspberry Pi single board computer which outputs them in real time to a computer monitor for inspection and processing. The set-up and calibration of this device was work in common with two other members of the visualisation laboratory.

Based on the maximum flow rate of a DMFC (Fs = 1000) ml/min, The output volume of a smell burst can be calculated in ml using the formula:

V =FsSpt, (8.27)

where Sp is the set point and t is the time of the smell burst required in minutes. For example, for anSp = 0.5Fs and burst time t= ₆₀1 min the output bursts have volume of 8.3 ml. After a specific volume of air is released by the DMFC, it passes through a sealed reservoir where it is enriched with the smell which is stored in liquid form. Finally, the burst is delivered to the user using flexible solenoids (of approximately 1 m length each). Pressure measurements were conducted at the joints of the device during the test sessions to ensure that the pressure level remains invariant throughout the device use.

The operating range for guaranteed concentration accuracy for the citral VOC isI = [C0, Cmax] = [1.2,11.2] ppm for a burst time of 1 sec. These values were

obtained by collecting concentration data with a PID sensor at di↵erent output vol- umes of the air-citral mixture (see formula 8.27 and figure 8.11). The collected data were fitted with the linear prediction modelC = 0.613V + 1 which gives the above range of concentration values for the volume range [V0, Vmax] = [0.3262,16.659] ml

of the output smell bursts. This model has coefficient of determinationR2 = 0.9919 which indicates almost perfect fit to the collected data.

Doctorate Student Annual Progress Review - Amar Dhokia 0937418 Details of Papers in Progress

Below describes the research questions being asked along with details of experiments that are to be put into papers. The focus of the research is on the delivery of smells to a user and after exploring this arrived at these questions, 2 experiments were designed, which are stated below:

Experiment 1 – Smell delivery calibration:

While many have experimented with smells in psycho-physical experiments, there is a general trend in the lack of accurate measurement of the smell being presented to participants that are not in controlled lab conditions. This lack of quantification doesn’t allow for any controlled standardised way in which to conduct experiments and in addition: These lab conditions require very controlled conditions and expensive sensitive equipment and, for the most part, are impractical for general or specific VE experiments. Odour intensity and smell perception are inherently linked and controlling the former proportionally affects the latter. Without an effective way to calibrate smell displays, one can’t tell if participants are receiving the same stimulus between experimental groups potentially rendering some data/conclusions controversial. This brings us to our first question:

• What is the best possible way of presenting smells in an accurate, precise and reproducible manner for use in virtual environments?

The human perception in difficult to accurately quantify due to the high variance in subjectivity. As such, the approach employed here involves controlling any physical aspects as best we can and then

examining the human response to these physical parameters. In this particular case, this involves using an olfactometer as a smell display to generate smells with a variable range of concentration. We then use a sensor to map physical changes to changes in concentration. We have thus built an olfactometer using parts and from the current olfactometer design, we plan to use an airflow through the headspace of a reservoir filled with liquid (or solid) odours and map this to a concentration at the receiving end. The sensor to be used is a photo ionisation detector. These sensors excel at measuring aggregate mixtures of gases and is able to detect compounds down to the parts per billions level (odour tends to be in parts per million/billion level). The sensor however detects everything at once and has a different relative response to different molecules, thus we found a single compound that not only has a recognisable smell but is easily detected by the sensor: Citral. Citral has a sweet lemony odour and has the same response

Figure 1- Diagram of the olfactometer with calibration setup

Figure 8.11: The olfactory display used in this thesis. The blue arrows show the air flow direction. Starting from the air compressor to the mass flow controllers and after that through the smell reservoir the air stream is aromatised and delivered to a PID for inspection (during testing) or the user (during an experimental trial).

The onset and rise times of the olfactory display were also measured using the PID sensor and found 240 ms and 73 ms respectively. These values show that the delivery of any other sensory stimulus needs to be delayed with at least 240 ms delay if synchronous delivery of all the sensory stimuli is required in the application. The onset and rise times of the olfactory display were used for the implementation of the experimental study E3 were synchronous delivery of visual, auditory and olfactory cues is required.