Commercial in Confidence

Report on a Study by E-Nose to Understand Action of Biofilter Material on Odour Study commissioned by Bioaction P/L

E-Nose Pty Ltd

145 NIC Building Australian Technology Park Eveleigh, Sydney NSW Australia 2015. Web: www.e-nose.info Makers of E-Nose Mk4 and graffit-e-nose
Corporate Member ASIAL

Ph. +61 2 9209 4083 Fax + 61 2 9209 4081 Mobile: 0416 099 529 E-mail: g.bell@e-nose.info ABN: 14 103 740 567

20/1/12

E-Nose Pty Ltd

Table of Contents Page Aim Method Results Discussion Conclusions Acknowledgements 3 4 5 7 7 8

Cover Figure: Bioaction Pty Ltd biofilters with seven ports for extraction of air from varying depth in the filter. The E-Noses were housed in the adjacent trailer.

Figure 1: The Bioaction vertically structured biofilter with seven ports for monitoring activity in the filter material at different depths.
2

E-Nose Pty Ltd Aim Biofilters form an important part of odour control in industrial operations which involve smell. The problem that currently exists for plant operators is that there is no gauge on the filters to show how close to spent the material in them is, or indeed if they are effective at all. In some cases anaerobic processes in the filter add to the smell that they are supposed to be filtering. Occasionally a spike of odour is generated in the industrial operation that passes directly through the filter and out into the community, causing extreme concern in the community (as evidenced by the recent Orica emissions of ammonia at a plant in the Newcastle area). It is the aim of this study to bring E-Nose technology to address these problems, by continuously monitoring the potentially polluting air entering the filter and alarming or setting off a defensive action (dosing or secondary filtration) in the event of a spike.

Figure 2: Odour from Yates operation piped to the two Bioaction biofilters.

E-Nose Pty Ltd Method Measurement of Total Chemical Load (TCL) by a Mk4 E-Nose was made during operation at different depths and at the inlet and outlet of two Bioaction P/L biofilters at the premises of Yates Fertilizer plant, Wyee, NSW. The dependent variable, TCL, is the sum of the outputs in mV registered by the devices six chemical sensors. Measurements were made directly from odour piped from the biofilters into three Mk 4 E-Noses (ENose Pty Ltd., Sydney NSW) and sampled in nalophan bags at different depths at ports on the two biofilter chambers. Three E-Noses inside steel cabinets sucked air from the biofilters over a period of four weeks.

Figure 3: Odour from the biofilters were drawn into one of three customised steel cabinets inside the trailer parked next to the biofilters. The bagged air was put through three E-Noses and duplicate samples and room air samples were given blind to 9 healthy non-smoker adults (7 females and 2 males all aged between 25 and 50y) to sniff. The odour was delivered in 50ml syringes drawn from the sample bags, to both human and machine. The human subjects were told that the odour in the syringes, if it were present, would be the smell of fertilizer. There were two sets of 9 syringes, 8 with odour and one without). They were asked to mark a questionnaire for each randomly presented sample, firstly if they could smell any odour at all, and if so, how strong it was on a scale of 1 to 10 (from very weak to strongest imaginable). They were also to indicate which bag smelled of fertilizer. 4

E-Nose Pty Ltd Sampling of air from the inlet and outlet of a biofilter was made by the independent supplier of dynamic olfactometry(DO) measurement (The Odour Unit Pty Ltd - TOU). These two points were the only ones in common with the present study, although other DO measurements were made on site. Results The continuous data from the directly ducted odour was mostly beyond the range of the E-Noses, at their lowest sensitivity setting. The three devices did however show a high level of concordance indicating that three devices could be used to monitor points in a filtration system to act as a control for dosing of fugitive odour spikes or other measures. Development needs to continue to systematically dilute the inputs and to synchronise flow of air from several sources to a single device. The bagged air experiment allowed for a small sample of odour to be applied to the devices without overburdening the sensors. From bagged air put into the devices, Figure 3 indicates that the biofilter filter has removed 73.33% of the airborne chemicals entering the filter. What is the relationship to these levels of TCL to human perception? Is the remaining chemical load coming out of the filter smelly? The human sniffing experiment, showed that the remnant TCL leaving the biofilter is below the human threshold of detection of odour from the biofilter, as measured by a panel of 9 adults sniffing bagged odour from the various biofilter ports. This means that 50% of people will not be able to detect any odour leaving the biofilter at the outlet, nor at depth 1 (in this figure, port B on the filter). At depth 3, (Port D on the filter) air at this depth is above the human detection level but below human recognition threshold. This means that 50% of people while being aware of an odour could not say what it was (in terms such as good or bad). The human threshold cut-offs are shown in Fig 2. Supra-threshold level of TCL informs the operator (or a control system) that action is needed, such as activity change, dosing, or refreshing filtration material.
5000 4500 4000 3500 3000 TCL (mV) 2500 2000 1500

Total Chemical Load at Varying Depths in a Biofilter Bed

Human 1000 threshold ? 500

0 Raw Depth 6 Inlet 5 4 3 2 1 0 (outlet)

Figure 4: Data of odour (total chemical load on all E-Nose sensors) at vary6ing depths in the biofilter. Depth 6 is closest to the inlet and Depth 1 in near the outlet or bed surface. 5

E-Nose Pty Ltd

Table 1: Identification of human thresholds for detection (green) and recognition (blue) of odour sampled at different depths of two biofilters
% Subjects rating sample as zero odour I A B C D 44 33 22 22 11 56 33 44 11 22 50 33 33 16.5 16.5 % Subjects rating sample as zero or level 1 odour I A B C D 89 66 22 56 4 66 56 56 22 56 77.5 61 39 39 50

Sample Unit 1 Unit2 Mean Sample Unit 1 Unit2 Mean

E 11 0 5.5 E 33 0 16.5

F 0 0 0 F 11 11 11

G 0 0 0 G 0 0 0

H 0 0 0 H 0 0 0

Most of the group correctly identified that Sample I had no odour. Not until Sample C was detection taking place in the larger majority of subjects. In sample D most subjects could still not recognise the odour. This indicates that the human odour detection threshold is at air loaded with molecules at level B in the biofilter, but they remain difficult to pin down (mentally) as having any particular quality until they reach a load down at level D. This allows the placement of the cut-offs shown in the next figure.
9 8 7 6 5 Perceived Intensity 4 3 2 1 0 I A B C D E Samples F G H

Yates Biofilters Units 1 and 2: Human Intensity Ratings 30/11/11

Thresholds: Detection Recognition

Unit 1 Unit 2 Mean Expon. (Mean) y = 0.742e0.2625x R = 0.9036

Figure 5: Human assessments of the two sets of air samples from Biofilters 1 and 2.

E-Nose Pty Ltd

5000 4500 4000 3500 3000 TCL (mV) 2500 2000 1500 1000

Total Chemical Load at Varying Depths in a Biofilter Bed

Thresholds: Recognition Detection

Human threshold ? 500

0 Raw Depth 6 Inlet 5 4 3 2 1 0 (outlet)

Figure 6: Superimposition of human thresholds found in this study on the data from duplicate samples applied to the E-Nose. After level 4, odour fumes passing through the filter have been reduced to below human recognition threshold and by level 1 they no longer can be detected at all by 50% of normal adults. Dynamic Olfactometry (Odour Units) The Odour Unit Pty. Ltd. (TOU) sampled and measured odour in terms of Odour Units/m3 at the following points: raw Inlet, Stacks, Biofilter unit 1 and Biofilter Unit 2. Each measurement was repeated (two measurements per sampling point). The measurement for Stack sample (Process without Bioactions Biofilters)were not the subject of study here so the data cannot be compared to the E-nose results, nor were the Odour Units measured at different depths in the biofilters. The dynamic olfactometry results are shown on the following Table:

E-Nose Pty Ltd Table 2

E-Nose Pty Ltd Table 3: E-Nose and TOU Results Compared E-nose (mV) Raw Inlet Biofilter Unit 1 (Outlet) Biofilter Unit 2 (Outlet) 4500 1200 Odour Units Repetition 1 10800 892 892 Odour Units Repetion 2 12400 832 892

From the two measurement points made by both methods a linear calibration were derived. Once other points are made (at different depths in the biofilter by DO, the non-linear (better predictive) relationship and calibration equation can be derived. Using the available data, the following linear relationship between E-Nose and DO can be seen in Figure 7.

OU vs E-Nose Calibration (Linear)

14000 12000 O U (TOU Pty Ltd) 10000 8000 6000 4000 2000 0 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 E-Nose Reading (mV)

Figure 7: Plot of data by the two types of measurement of odour from the Biofilter. The equation for this linear calibration is: y=3.2494x-3022.3. This can be used to express E-Nose readings as OU values automatically for future work on the biofilters. Discussion It is worth noting that the E-Nose shows that most airborne chemicals entering the filter bed are removed between levels 6 (nearest filter) and level 4 (about midway into the filter). This predicts that the deeper material is doing most of the work and that deeper material is likely to become exhausted sooner than material closer to the outlet. The E-Nose offers a continuous real-time monitor of what is happening inside the filter and how much volatile chemical material is being removed by the filter.

E-Nose Pty Ltd Using both methods on at the same sampling points, the relationship between them can be expressed as an equation. The equation can be used to convert E-Nose measurements to Odour Units automatically. This is demonstrated in the linear model derived from the input and output sampling points where data was recorded by both methods. Further work aims to show that the E-Nose can control variables which will keep the filter optimally effective, such as material changes or exchanges of shallow for deep material, and minimise the effects of fugitive odours, by also controlling a dosing system, routing odour to a special emergency system or other odour-abatement regime. Conclusions This study has shown the usefulness of an electronic nose in monitoring odour from a biofilter. The data from the E-Nose matched the human data in a predictable way and showed that once calibrated with human perception, it can be used to inform an operator what is happening to odour going into, inside, and coming out of biofilters. The electronic nose is likely to play an important part in the control of odours in this and other chemical engineering contexts, with benefit to plant operators in terms of cost and social responsibility.

Acknowledgements

I would like to thank Larry and Peter Botham (Bioaction) as well as David King (Yates) for commissioning this study and helping with carrying it out. E-Nose assistance was given by Brian Crowley, Martin Kwong, Anthony Kasih and Ruan VanWyk.

G.A. Bell, Ph.D. CEO, E-Nose Pty Ltd. 20/1/12

10