Zvi Boger

Optical flame detection using large-scale artificial neural networks

A model for intelligent hydrocarbon flame detection using artificial neural networks (ANN) with a large number of inputs is presented. Joint time-frequency analysis in the form of short-time Fourier transform was used for extracting the relevant features from infrared sensor signals. After appropriate scaling, this information was provided as an input for the ANN training algorithm based on conjugate-gradient (CG) descent method. A classification scheme with trained ANN connection weights was implemented on a digital signal processor for an industrial hydrocarbon flame detector.

2008 Special Issue: An adaptive method for industrial hydrocarbon flame detection

An adaptive method for an infrared (IR) hydrocarbon flame detection system is presented. The model makes use of joint time-frequency analysis (JTFA) for feature extraction and the artificial neural networks (ANN) for training and classification. Multiple ANNs are trained independently on a computer, using the backpropagation conjugate-gradient (CG) method, with input data collected from various flame and non-flame nuisance signals at four different IR wavelengths. The trained ANN connection weights are programmed into an embedded system as part of the filtering scheme for distinguishing flames from nuisance sources. Signal saturation caused by the excessive intensity of some IR sources is resolved by an adjustable gain control mechanism. The model described herein is employed in an industrial hydrocarbon flame detector.

Infrared Flame Detection System Using Multiple Neural Networks

A model for an infrared (IR) flame detection system using multiple artificial neural networks (ANN) is presented. The present work offers significant improvements over our previous design (Huseynov et al., 2005). Feature extraction only in the relevant frequency band using joint time-frequency analysis yields an input to a series of conjugate-gradient (CG) method-based ANNs. Each ANN is trained to distinguish all hydrocarbon flames from a particular type of environmental nuisance and ambient noise. Signal saturation caused by the increased intensity of IR sources at closer distances is resolved by adjustable gain control.

Enabling MEMS Chemical Microsensor Arrays for Trace Analyte Detection

We describe the development of a conductometric gas microsensor technology that combines, in an optimized manner, nanostructured sensing films, MEMS microhotplate array platforms, and artificial neural networks signal processing. Individually addressable microelements, including varied semiconducting oxides, are temperature modulated to produce analytically rich data streams that allow recognition of low concentration target analytes in background mixtures. This brief report emphasizes: 1) recently developed selection and processing methods for incorporation of high performance sensing materials on the MEMS platforms, 2) special operational modes and data acquisition approaches for ensuring good signal quality while maximizing information content, and 3) signal analysis techniques that include preprocessing routines and advanced recognition algorithms.

Reliable recognition of low level simulated chemical weapons

The detection of chemical warfare simulants (CWSs) in the presence of high concentrations of common interferences was demonstrated with solid-state chemical microsensor arrays. Conductometric SnO2 and TiO2 sensing films deposited on microhotplates were utilized as transducers. The embedded microheaters were used to modulate the temperature of the sensor-array elements, while measuring conductance at each temperature. The resulting temperature-modulated sensing data was normalized to the high-temperature conductance and an artificial neural network was used to perform pattern-recognition on the resultant conductivity data vectors

Artificial neural networks methods applied to conductometric microhotplate data for the identification of the type and relative concentration of chemical warfare agents

Response data from microhotplate (MHP) sensor arrays were measured for various chemical warfare (CW) agents in several concentrations in dry hair. Efficient large-scale artificial neural networks (ANN) modeling has been evaluated as a method for the classification and concentration prediction of the CW agents based on the MHP data. Four MHP sensor elements, two pairs of SnO/sub 2/ and two pairs of TiO/sub 2/ were operated in a pulsed, ramped temperature mode to generate the data used. The CW agents and related compounds tested were tabun (GA), sarin (GB), sulfur mustard (HD), and chloroethyl-ethyl-sulfide (CES), in four concentration levels in dry hair, between several nmole/mole (ppb) to several /spl mu/mole/mole (ppm). Recursive ANN pruning and re-training techniques were used to identify the more relevant inputs, among the original 80 inputs (different sensor elements and temperatures). ANN models with 6-15 inputs produced good classification between the different CW agents. Other ANN models, trained for each agent, gave good prediction values for the concentrations of the CW agents.