MODELING THE PROPERTIES OF GAS SENSOR MATERIALS BASED ON COBALT-CONTAINING POLYACRYLONITRILE USING REGRESSION ANALYSIS AND NEURAL NETWORKS

A modeling approach has been developed for materials based on organic semiconductorsand their physicochemical and gas-sensitive properties. For modeling, such methods as multiplelinear non-linear regression, neural networks were used. As an input vector theproperties of metal-containing polyacrylonitrile are the parameters technological process offorming materials: mass fraction alloying component (cobalt) in film-forming solution,technological modes IR annealing: temperature, time first second stages. Outputvector - functional characteristics physical chemical properties (resistivity,gas sensitivity coefficient, stability selectivity). Abstract—Metal–carbon systems with Co metalparticles have synthesized by pyrolysis. The resistance valueswere measured medium detected gas (chlorine). Modeling characteristicsand was carried out basis data obtainedfrom study 200 samples cobalt/polyacrylonitrile films. Multiple linear regression proved to be effective predicting resistivity values. Neural used predict gassensitivity selectivity, cobalt-containing films.An artificial network form a multilayer perceptron built coefficient sensor elements processes obtainingmaterial (mass stages). Complianceof model checked: experimental data: correlation coefficientR=0.82, root-mean-square error st=0.017. models satisfactorily describe collecteddata within error, which makes it possible optimize compositionand heat treatment conditions.