The goal of this project is to reduce firefighter deaths and injuries due to flashovers and to enhance firefighting safety and situational awareness in commercial building environments.
Flashover is an extreme fire event. When it occurs, there is a near-simultaneous ignition of most of the directly exposed combustible materials within a compartment. Due to the large heat release rate, gas temperatures increase rapidly and may exceed 800 °C. Rapid fire progression, such as flashover, is the number two cause of firefighter deaths and injuries. Over the past 10 years, approximately 700 firefighters were killed and more than 200,000 were injured. Unfortunately, there are still no tools that firefighters can use to detect flashovers, so they rely on their past experience using so-called flashover indicators that are difficult to recognize. For these reasons, researchers at NIST have been developing data-driven models that can be used to help firefighters predict the potential of flashovers.
Existing modeling approaches cannot be used in real-time firefighting due to two major problems. The first problem is that the existing models are numerically inefficient for real-time applications. Even when high-performance computing is being used, a single calculation takes more than 5 minutes. The second problem is that the fire scenarios being considered by these models are oversimplified. Sensors are assumed to work at extremely high temperatures and the fire locations and vent opening conditions are assumed to be well known. In real-life situations, however, sensors will fail and the inside conditions are never known. NIST has established a smart firefighting project to enhance firefighting safety and situational awareness by enabling real-time prediction of flashover conditions in commercial building environments. To reach this goal, the relationships of fire data, such as temperature, smoke, and species concentrations, and the effect of data quality, must be understood to use machine learning for effective real-time predictions.
Key Responsibilities will include but are not limited to:
Enhance the existing machine learning-based flashover prediction model
Play an active role in developing a functional platform to process data streams in real-time
Work with NIST research staff for machine learning model deployment in full-scale testing
Desired Qualifications:
Bachelors, masters, PhD., or equivalent experience in computer science, engineering, or related fields
Experience in Python
Knowledge of other programming languages, such as Javascript and MATLAB, is a plus
Basic working knowledge of Tensorflow and/or Pytorch
Ability to identify, manage, and overcome technical hurdles for machine learning model deployment
Other Details:
Full-time: the participant is expected to work 40 hours a week, or
Part-time: the participant is expected to work 16-20 hours a week
Location: the participant will work at the NIST Gaithersburg Campus.
Duration: this is expected to be a six-month to one-year position. Extensions are sometimes granted depending on the availability of funds.
Employment type
Full-time
Referrals increase your chances of interviewing at Morgan State University by 2x