Optimization of electronic nose drift correction applied to tomato volatile profiling

Abstract

30 E-noses can be routinely used to evaluate the volatile profile of tomato samples once the sensor drift 31 and standardization issues are adequately solved. Short-term drift can be corrected using a strategy 32 based on a multiplicative drift correction procedure coupled with a PLS adaptation of the Component Correction. It must be performed specifically for each sequence, using all sequence signals data. With 34 this procedure, a drastic reduction of sensor signal %RSD can be obtained, ranging between 91.5% 35 and 99.7%for long sequences and 75.7% and 98.8% for short sequences. On the other hand, long-36 term drift can be fixed up using a synthetic reference standard mix (with a representation of main 37 aroma volatiles of the species) to be included in each sequence that would enable sequence 38 standardization. With this integral strategy, a high number of samples can be analyzed in different 39 sequences, with a 94.4% success in the aggrupation of the same materials in PLS-DA two-40 dimensional graphical representations. Using this graphical interface e-noses can be used to 41 developed expandable maps of volatile profile similitudes, which will be useful to select the materials 42 that most resemble breeding objectives or to analyze which preharvest and postharvest procedures 43 have a lower impact on the volatile profile, avoiding the costs and sample limitations of gas 44 chromatography.

Review–Modern Data Analysis in Gas Sensors

“…Another study adopting PLS with domain adaptation showed reduced relative standard deviation of the sensor signals ranging between 91.5% and 99.7% depending on long or, short sequences. 141 PLS was also used in another study 142 for recognizing volatile organic gases. Zhang et al 143 proposed a target domain free approach which provided promising results without incorporating the data from the unknown target domain (Fig.…”

Gas Discrimination by Simultaneous Sound Velocity and Attenuation Measurements using Uncoated Capacitive Micromachined Ultrasonic Transducers

“…The gases from industrial grade bottles (1) are mixed by a system of flowmeters (2) controlled by a computer (3) at the desired concentrations. The mixture passes through an oven (4) where the desired temperature is set and verified by a thermocouple (5). The gas enters the measurement cell (6) where a continuous ultrasonic sine wave is sent from an emitter CMUT array to a receiver CMUT array after travelling a distance d. Thus, the wave arrives attenuated exponentially with d. Additionally its phase is shifted with respect to the emission by an amount proportional to the time of flight, τ = d/v, required to go through the measuring cell.…”

“…For applications dealing with hazardous gases such as hydrogen, H 2 , or methane, CH 4 , this becomes dangerous for the operators. Alternatively, it is possible to compensate this drift digitally, for instance, with artificial intelligence 4 or sophisticated models 5 . Nevertheless, this naturally increases both the manufacturing and the development cost.…”

Gas discrimination by simultaneous sound velocity and attenuation measurements using uncoated capacitive micromachined ultrasonic transducers

“…For applications dealing with hazardous gases such as hydrogen, , or methane, , this becomes dangerous for the operators. Alternatively, it is possible to compensate this drift digitally, for instance, with artificial intelligence 4 or sophisticated models 5 . Nevertheless, this naturally increases both the manufacturing and the development cost.…”