Abstract: Wheat grain of five storage ages were detected by an electronic nose(PEN2) from Airsense Company in Germany. A few of unremarkable sensors were removed by multivariate variance analysis and Loading analysis. Finally, response signals of sensor 1, 2, 8, 9, 10 were chose for pattern recognition. Principal component analysis(PCA) was applied to the signal of optimized sensor array, the five different storage ages of wheat grain were discriminated well and each group had strong convergence. The results obtained by network 1(BP neural network for signals of optimized sensor array) presented higher percentage of correct classifications than that by network 2(BP neural network for signals of original sensor array). So, removing unremarkable sensor signals by optimizing sensor array can improve the recognition performance of electronic nose.