Thermal Modulation Feature Extraction And VOCs Molecular Recognition Of Oxide Semiconductor Gas Sensor

Metal oxide semiconductor(MOS)gas sensors have been widely used in military,scientific research and various fields of national economy,which has the advantages of small size,low power consumption,high sensitivity and good silicon process compatibility.However,its poor selectivity is the biggest obstacle to its application.In order to solve this problem,the thermal modulation response signals of a single gas sensor to different gas molecules have been analyzed by means of variable temperature modulation.With the rapid development of artificial intelligence algorithms,more features of gas molecules can be extracted,which improves the recognition ability of a single gas sensor.Finally,a prototype device for intelligent gas molecule recognition has been proposed.The research results and innovations are as follows:1.p-type NiO sensor was used to identify volatile organic compounds(VOCs)gas molecules under thermal modulation.p-type NiO nanoparticles were synthesized with the assistance of bacterial cellulose(BC)scaffold.Transient response characteristics of NiO sensor(modulated by a staircase waveform)toward 5 kinds of VOCs were investigated.A signal pretreatment method was proposed to gradually remove the irrelevant electrical signals.Firstly,the conversion of the raw electrical signal into sensitivity or response data(Rgas/Rair under the same temperature profile)allows to get rid of temperature coefficient of resistance(TCR)interferences(arising from the temperature dependent electrical properties of sensing channel,other than molecule/MOS interfacial charge exchange).Then,the gas concentration related information could be further removed via the normalization of the thermal modulation sensitivity response signal.Finally,the influence of other noises could be removed via the discrete wavelet transform(DWT).According to this signal preprocessing method,successful classification and recognition of tested VOCs molecules,including three kinds of benzene series(benzene,toluene and chlorobenzene),were achieved by typically non-selective p-type NiO sensor with a low sensitivity.Our work highlight the importance of irrelevant thermal modulated electrical signals to extend the recognition capability of a single MOS sensor(toward VOCs molecules),and sheds light on the exploring future smart gas molecule recognition chips.2.The recognition of benzene-toluene-xylene(BTX)gas molecules was realized by metal oxide gas sensor under thermal modulation.Instead of limited features extracted from isothermal resistance measurements,transient response signals from a temperature modulated sensor were adopted in this work.Herein,the transient responses of BTX molecules(including three xylene isomers)were investigated by generic,non-selective metal oxide(NiO,WO3 and commercial SnO2 based TGS2602)sensors.Robust prompt(within?5 s)discrimination of BTX molecules could be successfully achieved by either linear discrimination analysis(LDA)of three sensors,or convolution neural network(CNN)analysis of single sensor.Our work highlight that appropriated thermal modulation offers a powerful route in extracting the feature differences of highly similar molecules like BTX including xylene isomers,and the recent advance in deep learning algorithms in the area of artificial intelligence(AI)could substantially empower the recognition capability of non-selective semiconductor sensors.Combination of two strengths could pave the way for innovating smart molecule recognition chips for real applications in the near future.3.A portable intelligent gas molecule detection and recognition system has been developed.Firstly,the signal acquisition system and temperature acquisition system of MOS sensor was developed based on 51 single chip microcomputer.Display the signal voltage(corresponding to the resistance of the sensor layer)in real time.According to the resistance change of MOS sensor layer with the gas concentration under constant heating temperature,real-time gas concentration measurement could be realized.The system includes both threshold alarm and temperature acquisition functions,and the acquisition signal could be transmitted to PC through serial port.Then,an intelligent gas identification system was developed.The circuit was designed to dynamically modulate the voltage of sensor heating end.The data of characteristic change curves were transmitted to the top machine through the serial port.After extracting the features by signal processing method,the responses were fed into BPNN/CNN(constructed by Python-Tensorflow)to realize gas molecule recognition.