Study On The Specificity Of Graphene Gas Sensor Based On Pattern Recognition

In the study of traditional biochemical sensors based on two-dimensional materials,surface modification,which modified one or more kinds of molecules on the surface of two-dimensional materials,was a common method to achieve the selectivity of sensor.Sensors,fabricated by surface modification,were not easy to obtain and only able to identify one kind of molecule rather than several molecules.According to situations above,the method of biological chemical sensor based on pattern recognition was presented in this paper.Compared to the conventional biochemical sensors,which only identified a kind of target molecule in accordance with the biochemical reaction,biochemical sensors based on the pattern recognition algorithms,combining the technology of artificial intelligence with biological chemistry,could able to identify a variety of organic molecules.The basic principle of the biochemical sensors were shown as follows: the first step was using the custom circuit transformed the human eyes unseen signals of reactions of biochemistry into the electrical signals.Then,using an analog-to-data module,the electrical signals were converted into data signals which could be recognized by computer.Eventually,the method of the artificial intelligence technology of computer was used for making pattern recognition for data signals.According to the idea of pattern recognition algorithms,there are two chemical sensor system was designed in this paper: unmodified graphene foam chemical sensor system based on support vector machine and unmodified graphene foam chemical sensor system based on the principal component analysis(PCA)and back propagation neural network(BPNN).These two systems not only could identify several kinds of molecules,but also not need any molecule in the sensor surface modification.That was good for the fabrication and cost of sensor.What's more,it had good service life.The sensor system consists of electrical resistance real time domain detection system(ERTDS),unmodified graphene foam as sensingelement and a pattern recognition module.And the biggest difference of two systems was on the pattern recognition module.During the work of whole sensor system,we first to obtain the electrical resistance time domain response curve which different organic molecule reacting with the graphene foam.Then,to extract the characteristic of each response curve were based on the methods of data preprocessing(DPP).Eventually,the matrix which consists of extracting characteristics was inputted into pattern recognition module to identify the types of molecules.The biggest difference of two systems was on the pattern recognition module.The first system was directly input the extracted characteristic into SVM making pattern recognition,while the latter system was input the characteristics which were dealt with the principal component analysis(PCA)into back propagation neural network(BPNN)making pattern recognition.However,experiencing many times' validation experiments,the identification accuracy of these two system were above 90%.Except for high identification accuracy and being able to identify a variety of molecules,this class of sensor also had the advantage of measuring speed,fabrication process and cost,common temperature measurement and long service life.Besides,the biological chemical sensor based on pattern recognition algorithms provided a new valuable strategy for the study and application of biochemical sensor in the artificial intelligence technology filed.