Pattern recognition algorithm for identifying and quantifying single and mixed contaminants in air with an array of nanomaterial-based gas sensors

Abstract

A method is described for identifying and quantifying single and mixed contaminants in air by reading nanohybrid gas sensors multivariate output and processing it inside the algorithm. The algorithm analyzes sensor signal in real time and outputs estimated values for concentrations of target gases.

What is claimed is: 
     
         1 . A method for the selective detection of a target gas and measuring the concentration values comprising:
 taking resistance values of 8, 16, 32, 64, or 128 channels of nanohybrid gas sensors sampled every 80, 120, 160, or 200 milliseconds;   filtering out the high frequency noise using the exponential average low pass filter;   computing the rate of sensor response change; and   evaluating sensor response with respect to other sensor channels including the temperature sensor.   
     
     
         2 . The method of  claim 1 , further comprising predicting settled sensor resistance values to estimate algorithm input values when sensor output values are in transition following the change in gas concentration values. 
     
     
         3 . The method of  claim 2 , further comprising using a gas model that relates change in resistance of material segments to target gas concentration via model coefficients, wherein the relation between sensor response and change in target gas concentration described by equation:
     C   i =Σ j α j   i ( R   j   −R   j   0 )/ R   j   0   +C   i   0 ;
   Wherein R j   0  is defined as the channel resistance for material j right before the exposure, R j  is defined as the resistance right after the exposure, and wherein the sum is taken over all channels of various materials j contributing to the algorithm input; and   C i   0  is defined as the target gas i concentration right before the exposure, C i  is defined as the target gas i concentration right after exposure, wherein for every target gas i each material j channel contains certain material-gas coefficient value α j   i .   
     
     
         4 . The method of  claim 1 , wherein preprocessed signals from nanohybrid gas sensor channels are grouped into segments each representing a specific material deposited on sensor channel, and wherein multiple segments are used in engaging a single target gas model. 
     
     
         5 . The method of  claim 4 , wherein multiple models are concurrently executed in the algorithm predicting concentration values for gases, including at least one of NO2, SO2, CO, CO2, O3, CH2O, CH4, NH3, N20, organic compounds such as Acetone and Ethanol, and various Hydrocarbons. 
     
     
         6 . The method of  claim 1 , wherein a response of a sensor is a result of exposure to multiple gas constituents in the atmosphere as well as the reaction of the sensor to various environmental factors such as humidity, temperature, pressure and air flow, and further comprising resolving the cross-sensitivity complexity via an over-constrained system of modeling equations. 
     
     
         7 . The method of  claim 6 , wherein the compensation coefficients to account for environmental factors are a combination of: humidity compensation coefficient, temperature compensation coefficient, and pressure and air flow compensation coefficient. 
     
     
         8 . A method for tracking null reference baseline using multiple-channel time series signal from a hybrid nanostructure gas sensor, comprising:
 taking resistance values of multiple channels of nanohybrid gas sensors;   comparing them against the reference resistance values benchmarked in ambient atmosphere with known concentrations of contributing gases; and   adjusting the starting values for target gas concentrations using the deviations from benchmarked values for at least some of temperature, humidity and multiple channels of nanohybrid gas sensors.