VOC sensor

Overview

In our project, we are exploring a new way to sense the healthy condition of the crops which will further give this information(healthy or not) to bothhuman and T.atroviride, in order to take actions on time and achieve the goal of prevention in advance.Meanwhile,this method must can be conveniently used in varieties of plants.We finally built up the e-nose(electronic nose) system to sense the VOC(volatile organic compounds) emitted by the plants or related microorganisms.Afterwards,by machine learning method, we achieved a high accuracy rate of sensing the tobaccos that infected by P.nicotianae, namely, our device successfully smelled whether the plants were ill.

Background

Plants will release varieties of volatile organic compounds (VOC) to resist infected phytopathogens, which lead to a change of VOC around the plants^[1]. Firstly, we verified the VOC differences between healthy tobaccos and P.nicotianae infected tobaccos by GC-MS.

GC-MS柱状图

Fig1: +title(Thanks for Hunan tobaccos bureau for providing this data)

We decided to use ten highly sensitive CMOS(Complementary Metal Oxide Semiconductor) gas detector to catch the VOC information(Adding an extra one afterwards), and every detector is sensitive to one kind or type of gas. It is important to note that e-nose can’t sense a specific VOC like mass spectrometers did, instead, it catch the overall characteristics of VOC as a "fingerprint".

气体腔体照片

Fig.2

Experimental Design

1.e-nose build-up and gas path design

Number	Sensor Type	Performance characteristics	Minimum detection limit of related gas
a	IST-8000	Highly sensitive to all types of VOC	1 ppm
b	TGS2600	Sensitive to cigarette smoke and cooking odors	1 ppm
c	TGS2610	Sensitive to alkanes such as liquid gas propane butane, low sensitivity to alcohol	10 ppm
d	TGS2603	Sensitive to ammonia and sulfide gas	1 ppm
e	MS1100	Highly Sensitive to aldehydes, toluene and organic solvents	1 ppm
f	TGS2611	Sensitive to methane	10 ppm
g	TGS2602	Highly sensitive to all types of VOC	1 ppm
h	MQ-7	Sensitive to carbon monoxide and other gases	10 ppm
i	MQ-135	Sensitive to ammonia,sulfide and benzene vapor, or harmful smokes	10 ppm
j	TGS822	Sensitive to alcohol and organic solvents	50 ppm
k(substitute b later)	iAQ-core	Extremely high sensitivity to all types of VOC and can output the equivalent concentration directly	125 ppb

Table1: Highly sensitive CMOS gas detectors we used(All of these detectors have long-term stability)

图：长期稳定性及经典响应曲线

Fig3

See more details about the device in the Hardware page.

Full automatic headspace sampling and standardized testing

We employed full automatic headspace sampling to detect the VOC:Taking six basins of tobaccos infected by P.nicotianae for 5 days and another six basins of healthy tobaccos as a control, put them into a box and sealed it. Started to measure the VOC after 30 minutes.In addition, the device is controlled by Arduino Single Chip Microcomputer(SCM), in order to achieve full automatic measure. We standardized the testing steps for later comparison between every group data.Click here to see more information about the device.

图：测量流程

Fig4: The standardized measuring steps

There is median filter algorithm in SCM to remove the extremums. We firstly preprocessed the data: identified and removed the base line values, after that,read the response value on 1min, 2 min, and the average value and maximum value of the response curve. We have measured 17 groups data for healthy tobaccos and 18 groups for infected ones.Click here to see our raw data and preprocessed data.

图：响应曲线

Fig5

After collecting the data from every detector, we applied algorithm to do data classification. We employed Decision Tree, Multi-Layer Perception algorithm, and Leaner Model, to achieve more than 85% accuracy rate of sensing the intruding P.nicotianae on tobaccos. Moreover based on the result of the modeling, four CMOS detectors were enough to make a judgement for tobaccos healthy condition, which means we can further reduce the cost of our device.Click here to see the details of our modeling process.

Conclusions

Our method achieved more than 85% accuracy rate of sensing the infected tobaccos, and only four detectors were necessary for our devices making such judgments. Moreover, the VOC device can link to the main engine which will send the data to PC port. We also built the corresponding web app to receive these data to achieve the real time monitoring, in the future we plan to make the web app possible for showing the healthy condition of plants real time.

On the other hand, the sensitivity of our device is still limited in real application, it can’t fit different field conditions. Therefore, we came up with a plan to solve this problem, click here to this improvement.

Reference

[1]Dicke M, Loon J J A V, Soler R. Chemical complexity of volatiles from plants induced by multiple attack[J]. Nature Chemical Biology, 2009, 5(5):317-324.