Standardization of metal oxide sensor array using artificial neural networks through experimental design

The shift in sensor signal measured by identical gas sensor array system (commonly called an electronic nose) makes the analysis of merged measurement data difficult. This would grossly affect the gas quantification accuracy of such electronic nose (E-nose) instruments. Thus, a real-time calibration transfer based on reference alcohol projection transfer model (RAPT) was designed in this paper which aims to project onto the hazardous gas and set up a “bridge” to transfer from instrument to instrument through three artificial neural networks (ANN), and attempt to solve the problem of signal shift between E-nose instruments of identical sensor array. Besides, principal component analysis (PCA) is also used for validation of different models in component space. For comparison, previous four models including univariate direct standardization (UDS), partial least square (PLS), neural, and global affine transformation based on robust weighted least square (GAT-RWLS) are also presented. Qualitative and quantitative results demonstrate that the proposed RAPT model is competitive in E-nose signal shift standardization.