Measuring odor concentration with an electronic nose

Adaptive Logic Network (ALN) software, a type of artificial neural network (ANN), was used to train a function to convert the measurements of a commercially available electronic nose into odor concentrations. A data set was developed by evaluating odor samples with both an olfactometer and the electronic nose. The odor concentrations measured with the olfactometer served as observed values, and the responses of a 32-sensor array in the electronic nose, together with the humidity of the odor sample and reference air, served as input variables. By applying a principal component analysis, the number of input variables in the data set was reduced from 34 to 3 which represented 99% of the variance. This data preprocessing procedure is crucial to the success of the ALN. Well-trained ALNs combined with an electronic nose can measure odor concentrations with about 20% mean error.