Search

An improved probabilistic neural network (IPNN) algorithm for use in chemical sensor array pattern recognition applications is described. The IPNN is based on a modified probabilistic neural network (PNN) with three innovations designed to reduce the computational and memory requirements, to speed training, and to decrease the false alarm rate. The utility of this new approach is illustrated with the use of four data sets extracted from simulated and laboratory-collected surface acoustic wave sensor array data. A competitive learning strategy, based on a learning vector quantization neural network, is shown to reduce the storage and computation requirements. The IPNN hidden layer requires only a fraction of the storage space of a conventional PNN. A simple distance-based calculation is reported to approximate the optimal kernel width of a PNN. This calculation is found to decrease the training time and requires no user input. A general procedure for selecting the optimal rejection threshold for a PNN-based algorithm using Monte Carlo simulations is also demonstrated. This outlier rejection strategy is implemented for an IPNN classifier and found to reject ambiguous patterns, thereby decreasing the potential for false alarms.

Chemiluminescent chemical sensors for oxygen and nitrogen dioxide have been investigated via the immobilization of 3-aminophthalhydrazide (luminol) within a hydrogel or polymeric, sorbent coating that is positioned in front of a photomultiplier tube. Some selectivity is tailored into these devices by careful selection of the polymer type, pH, and metal catalyst incorporated within the film. Oxygen levels as low as 2.4 ppm in nitrogen have been detected using the oligomer fluoropolyol as the support matrix for immobilizing luminol, KOH, and the metal catalyst, [Fe.sub.2]-[(S[O.sub.4]).sub.3]. For the detection of N[O.sub.2](g), the use of sorbent, polymer coatings such as fluoropolyol or polyethylenimine resulted in strong humidity dependencies, exhibiting no chemiluminescence under dry conditions. Incorporation of the chemiluminescent reagents within a hydrogel (e.g., poly(vinyl alcohol) or superabsorbing polymer Waterlock) eliminated this effect. A survey of various hydrogel immobilization matrices and metal catalysts found that, for the determination of N[O.sub.2](g), Cu(II) and the superabsorbing polymer Waterlock provided the best film performance with respect to sensitivity (0.46 ppb) and stability. Discussions center about the complete characterization of these chemiluminescent chemical sensors and an examination of specific methods used to extend the sensitivity and selectivity of these devices.

A smart sensor system for the detection of toxic organophosphorus and toxic organosulfur vapors at trace concentrations has been designed, fabricated, and tested against a wide variety of vapor challenges. The key features of the system are an array of four surface acoustic wave (SAW) vapor sensors, temperature control of the vapor sensors, the use of pattern recognition to analyze the sensor data, and an automated sampling system including thermally desorbed preconcentrator tubes (PCTs). All the electronics necessary to control and operate the various subsystems and to collect and process the data are included in the system. Organophosphorus analytes were detected at concentrations as low as 0.01 mg/[m.sup.3] in 2 min, and the organosulfur analyte was detected at 0.5 mg/[m.sup.3] in 2 min. Pattern recognition algorithms correctly classified these analytes at these concentrations and discriminated them for a variety of other organic vapors.

The outputs from a physics-based modeler of magnetometry data have been successfully used with a probabilistic neural network (PNN) to discriminate unexploded ordnance (UXO) from ordnance-related scrap. Cross-validation predictions were performed on three data sets to determine which modeler parameters were most valuable for UXO classification. The best performing parameter combination consisted of the modeler outputs depth, size, and inclination. The cross-validation results also indicated that good prediction performance could be expected. Model outputs from one location at a site were used to train a PNN model, which could correctly discriminate UXO from scrap at a different location of the same site. In addition, data from one site, the former Buckley Field, Arapahoe County, CO, was used to predict targets detected at an entirely different training range, The Badlands Bombing Range, Bull's Eye 2 (BBR 2), Cuny Table, SD. Through careful selection of the probability threshold cutoff, the UXO detection rate obtained was 95 % with a false alarm rate of only 37%. The ability to distinguish individual UXO types has been demonstrated with correct classifications between 71% and 95%. Index Terms--Classification, magnetometry, ordnance, probabilistic neural network (PNN), unexploded ordnance (UXO).