A Gas Sensor Array Model Calibration Method Based on Conditional Generative Adversarial Network for Data Augmentation

Abstract

The invention discloses a gas sensor array model calibration method for data augmentation based on a conditional generation confrontation network, comprising the following steps: step 1, collecting response signal data sets of the gas sensor array; extracting gases of different concentrations in a standard gas environment The characteristic data of the corresponding response signal is used as an original data sample; Step 2, preprocessing the original data sample to obtain a normalized value of the original data sample; Step 3, using the normalized value of the original data sample to generate an adversarial The network model is trained to obtain a sample generator model; and the characteristic data of the response signals corresponding to the gases of different concentrations are generated by the sample generator model as a generated data sample; step 4, the generated data sample is combined with the original The data samples are mixed to obtain the expanded data samples; Step 5, using the expanded data samples to calibrate the sensor array model.

Description

technical field [0001] The invention belongs to the technical field of gas sensor array signal processing, in particular to a gas sensor array model calibration method for data augmentation based on a conditional generation confrontation network. Background technique [0002] Model calibration of gas sensor arrays is an expensive but necessary process to establish the functional relationship between measured and analyzed quantities. The traditional calibration method is to first select the functional form of the calculation model, then estimate the corresponding model parameters and errors based on the training data set, and finally perform model verification. The resulting computational models are then used to make new measurements and predict concentrations or classes of gases. However, after a period of time, the performance of the model degrades due to changes in the characteristics of the sensing elements, requiring recalibration of the system. [0003] In recent year...

AI Technical Summary

Problems solved by technology

Second, if there are reasons such as partial sensor failure during the data collection process, the data measurement will be incomplete, which will bring great difficulties to the model calibration of the sensor; for example, when analyzing the drift characteristics of the sensor, it will take a long time Data monitoring, if there is data loss, then the sample data corresponding to the time cannot be reacquired

Third, if the gas concentration experimental sample used in model calibration is not completely consistent with the gas sample concentration distribution used in the actual measurement, it is necessary to re-collect and calibrate the sample. However, it is sometimes difficult to obtain low-concentration gas samples during the experiment. will lead to a decrease in the accuracy of the model calibration

Methods to reduce the frequency and associated cost of sensor array recalibration have not been investigated from a data generation standpoint

Images