Chemical Biological Radiological Nuclear and Explosive (CBRNE) sensing systems in the field provide alarms in the form of simple graphical representations, lights, vibrations, and alarm sounds to maximize the reaction time of the user in the event of a hazardous situation. Artificial Intelligence (AI) can be used to reduce the false alarms of chemical detectors, allowing users to react with confidence when an alarm does occurs. However, the Department of Defense’s AI ethics standards states that technologies incorporating AI systems be traceable, reliable, and governable. Given the complex nature of AI and the difficulties of interpreting results, testing and evaluating AI systems poses a challenge for CBRNE sensing systems. To properly interpret and evaluate AI systems it is imperative graphical user interfaces (GUI) are designed to be simple interfaces that provide easy to interpret results. Presented here is an interpretable alarm GUI for orthogonal networked sensors (IAGOnet). IAGOnet provides real-time status of connected sensors utilizing a familiar replication of their onboard results, along with simple to understand graphical representations of confidence metrics from machine learning (ML) predictions. IAGOnet allows a user to compare the detector’s original alarm state to current and previous predictions of classification algorithms, thereby reducing the false alarms. Our work demonstrates the practical nature of IAGOnet by utilizing data from an ion mobility spectrometry (IMS) based detector and a multi-gas detector.
The development of alarm algorithms in ion mobility spectrometry (IMS) based chemical vapor detection is challenged by the presence of overlapping chemical peaks. IMS technology identifies a chemical through hard-coded alarm windows. Alarm windows are designed as range of reduced mobility values, and act as an IF-THEN statement. Where if a peak forms in the region it then assigns a preset alarm label. A majority of IMS alarm algorithm design has relied on setting boundary conditions based on a statistical variance in product ion peak positions. To develop these alarm windows for IMS detectors the variance in peak position had to be captured through extensive laboratory testing. These windows are determined through time consuming and rigorous laboratory testing across multiple detectors under multiple conditions. Machine learning (ML) is a field of science that intersects with computer science and mathematics to “teach” a computer using large amounts of data. The development of traditional alarm algorithms IMS has left a plethora of data available to be explored by ML techniques. Presented here is a random forest (RF) classification model along with a long short-term memory (LSTM) based neural network model to label the spectra of IMS data with high accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.