Using Artificial Intelligence and Machine Learning for Standardized Fire Test Evaluation

Details

• Creators:
  Sidhom, Ahmed Ben ; Carannante, Giuseppina ; Bouaynaya, Nidhal Carla ; Walters, Richard N.
• Corporate Creators:
  Rowan University. Department of Electrical and Computer Engineering ; United States. Department of Transportation. Federal Aviation Administration. Technical Center
• Corporate Contributors:
  William J. Hughes Technical Center (U.S.). Fire Safety Branch
• Subject/TRT Terms:
  • Fire detection systems
  • Deep learning
  • Neural networks
  • Flammability tests
  • Aviation
  • Artificial intelligence
• Publication/ Report Number:
  DOT/FAA/TC-26/17
• DOI:
  https://doi.org/10.21949/mqv8-h177
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Corporate Publisher:
  United States. Department of Transportation. Federal Aviation Administration. William J. Hughes Technical Center
• Description:
  Fire safety remains a paramount concern for the Federal Aviation Administration (FAA), particularly in the context of passenger aircraft materials and cabins. The FAA utilizes standardized bench-scale fire tests, including the Vertical Flame Propagation (VFP) apparatus, to assess material flammability, flame spread, and burning behavior under controlled laboratory conditions. These tests provide critical data on flame propagation, burn-through, and post-flame characteristics essential for certification and rulemaking to ensure passenger safety by minimizing fire hazards. This study advances FAA's Fire Safety Branch objectives by integrating deep learning and computer vision, specifically, object detection algorithms such as Faster Region- Convolution Neural Network (Faster R-CNN), to analyze video records of FAA regulatory flame tests. These algorithms automate the detection and quantification of fire features from video frames, including ignition timing, burn length, flame intensity, and char area, eliminating subjectivity inherent in traditional operator observations. The application of image preprocessing and data augmentation enhances detection accuracy, while transfer learning and deeper network architectures improve model robustness.
  
  Performance evaluations on videos from the FAA's VFP tests demonstrate that deep neural networks can reliably identify fire presence, quantify burning characteristics, and extract precise temporal and spatial fire metrics. This automated and quantitative approach has the potential to standardize material flammability assessments, improve repeatability, and facilitate regulatory compliance, ultimately contributing to enhanced aircraft fire safety.
• Format:
  PDF
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/93558/dot_93558_DS1.pdf
• File Type:
  [PDF - 1.71 MB]
• Collection(s):
  FAA Fire Safety ; FAA William J. Hughes Technical Center for Advanced Aerospace
• Main Document Checksum:
  urn:sha-512:266575077cfee43ef2a8b2322d8f1355706dc7ac5517145ff2c3b9704ec8abfe7fe0ae7db153874be8e3adc6a47578003f80f73c685167a89a9a5f5765760c98