tdossett/aircraft-risk-assessment

Aircraft Risk Assessment Model

This model predicts ICAO/FAA risk assessment codes for aircraft maintenance tasks based on:

• Commodity type (e.g., "ATA 28 FUEL")
• Sub-commodity (e.g., "ATA 28-00 General")
• Aircraft type (e.g., "B737")
• Severity level (e.g., "Immediate production line")

Model Details

• Base Model: DistilBERT (distilbert-base-uncased)
• Task: Multi-class text classification
• Domain: Aircraft maintenance and safety
• Risk Scores: 1-100 scale (1 = lowest risk, 100 = highest risk)

Risk Assessment Codes

The model predicts risk codes like:

• 1A (Score: 1) - Very Low Risk
• 3C (Score: 26) - Medium Risk
• 5D (Score: 88) - High Risk
• 5E (Score: 100) - Critical Risk

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_name = "your-username/aircraft-risk-assessment"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Example prediction
input_text = "Commodity: ATA 28 FUEL | Sub-Commodity: ATA 28-00 General | Aircraft: B737 | Severity: Immediate production line"
inputs = tokenizer(input_text, return_tensors="pt", truncation=True, padding=True)

with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class = predictions.argmax().item()
    confidence = predictions.max().item()

print(f"Predicted class: {predicted_class}, Confidence: {confidence:.2%}")

Training Data

Trained on aircraft maintenance task cards with the following features:

• Commodities: ATA 28 (FUEL), ATA 29 (HYDRAULIC), ATA 30 (ICE PROTECTION), ATA 31 (INDICATING SYSTEMS)
• Aircraft Types: B737, B787, B777, A321, B781
• Severity Levels: Production line stages, preflight, delivery, in-service

Limitations

• Trained specifically for aircraft maintenance scenarios
• Limited to the commodity types and aircraft in training data
• Best performance on inputs similar to training format

Citation

If you use this model, please cite:

Aircraft Risk Assessment Model
Trained on aviation maintenance data
Base model: DistilBERT