LLM-Powered Automated FMEA Generator

An intelligent system that automatically generates Failure Mode and Effects Analysis (FMEA) from both structured and unstructured data using Large Language Models

---

📋 Table of Contents

• Overview
• Features
• Architecture
• Installation
• Quick Start
• Usage
• Configuration
• Project Structure
• Examples
• API Reference
• Contributing
• License

---

🎯 Overview

Traditional FMEA is manual, time-consuming, and expert-dependent. This system revolutionizes the process by:

• Automating extraction of failure information from customer reviews, complaints, and reports
• Processing structured data from Excel/CSV files
• Using LLMs for intelligent semantic understanding
• Computing risk scores (Severity, Occurrence, Detection)
• Generating actionable insights with recommended actions

Problem Statement

Organizations receive failure information in multiple formats:

• Unstructured: Customer reviews, complaint text, incident reports
• Structured: Excel spreadsheets, CSV files with failure data

This system provides a unified, intelligent solution to convert all these inputs into a standardized FMEA.

---

✨ Features

Core Capabilities

• ✅ Dual Input Support: Process both structured and unstructured data
• 🤖 LLM-Powered Extraction: Uses Mistral/LLaMA/GPT models for intelligent entity extraction
• 📊 Automated Risk Scoring: Calculates S, O, D scores and RPN automatically
• 🎯 Action Priority Classification: Categorizes risks as Critical, High, Medium, Low
• 📈 Visual Analytics: Interactive dashboards with charts and risk matrices
• 💾 Multiple Export Formats: Excel, CSV, JSON
• 🔄 Hybrid Processing: Combine multiple data sources seamlessly
• 🚀 Production-Ready: Modular, extensible, well-documented code

Technical Features

• NLP Processing: Sentiment analysis, keyword extraction, text cleaning
• Rule-Based Fallback: Works even without LLM for faster processing
• Batch Processing: Handle large datasets efficiently
• Deduplication: Intelligent removal of similar failure modes
• Configurable: YAML-based configuration for easy customization

---

🏗️ Architecture

User Input (Text/CSV/Excel)
         ↓
┌────────────────────┐
│  Data Preprocessing │ ← Text cleaning, validation, sentiment analysis
└────────────────────┘
         ↓
┌────────────────────┐
│  LLM Extraction     │ ← Extract: Failure Mode, Effect, Cause, Component
└────────────────────┘
         ↓
┌────────────────────┐
│  Risk Scoring       │ ← Calculate: Severity, Occurrence, Detection
└────────────────────┘
         ↓
┌────────────────────┐
│  FMEA Generator     │ ← Compute RPN, prioritize, recommend actions
└────────────────────┘
         ↓
┌────────────────────┐
│  Output & Export    │ ← Dashboard, Excel, CSV, JSON
└────────────────────┘

---

🚀 Installation

Prerequisites

• Python 3.9 or higher
• 8GB RAM minimum (16GB recommended for LLM)
• GPU (optional, for faster LLM inference)

Step 1: Clone Repository

git clone <repository-url>
cd Symboisis

Step 2: Create Virtual Environment

# Windows
python -m venv venv
venv\Scripts\activate

# Linux/Mac
python3 -m venv venv
source venv/bin/activate

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Download NLTK Data

python -c "import nltk; nltk.download('punkt'); nltk.download('stopwords'); nltk.download('averaged_perceptron_tagger')"

Step 5: (Optional) Install spaCy Model

python -m spacy download en_core_web_sm

Step 6: Configure Environment

# Copy example environment file
copy .env.example .env

# Edit .env with your settings (optional)

---

⚡ Quick Start

🎯 Working with YOUR Data (FMEA.csv + Car Reviews)

FASTEST WAY - Process your actual datasets:

python process_my_data.py

This will automatically process:

• ✅ Your FMEA.csv (161 industrial failure modes)
• ✅ Car reviews from archive (3) folder (Ford, Toyota, Honda)
• ✅ Create hybrid analysis combining both
• ✅ Export all results to output/ folder

📖 See YOUR_DATA_GUIDE.md for detailed instructions on working with your datasets!

---

Option 1: Web Dashboard (Recommended)

streamlit run app.py

Navigate to http://localhost:8501 in your browser.

Option 2: Command Line

# From unstructured text
python cli.py --text reviews.csv --output fmea_output.xlsx

# From structured data
python cli.py --structured failures.csv --output fmea_output.xlsx

# Hybrid mode
python cli.py --text reviews.csv --structured failures.csv --output fmea_output.xlsx

Option 3: Run Examples

python examples.py

This will run 3 demonstration examples and generate sample FMEAs.

---

📖 Usage

1. Using the Web Dashboard

1. Start the dashboard: streamlit run app.py
2. Select input type: Unstructured, Structured, or Hybrid
3. Upload files or paste text
4. Click "Generate FMEA"
5. View results: Metrics, tables, charts
6. Export: Download as Excel or CSV

2. Using Python API

from fmea_generator import FMEAGenerator
import yaml

# Load configuration
with open('config/config.yaml', 'r') as f:
    config = yaml.safe_load(f)

# Initialize generator
generator = FMEAGenerator(config)

# Generate from text
reviews = ["Brake failure on highway...", "Engine overheated..."]
fmea_df = generator.generate_from_text(reviews, is_file=False)

# Generate from structured file
fmea_df = generator.generate_from_structured('data.csv')

# Export
generator.export_fmea(fmea_df, 'output/fmea.xlsx', format='excel')

3. Using CLI

# Basic usage
python cli.py --text input.csv --output result.xlsx

# With summary report
python cli.py --text input.csv --output result.xlsx --summary

# Faster rule-based mode (no LLM)
python cli.py --text input.csv --output result.xlsx --no-model

# Custom configuration
python cli.py --text input.csv --config custom_config.yaml --output result.xlsx

---

⚙️ Configuration

Edit config/config.yaml to customize:

Model Settings

model:
  name: "mistralai/Mistral-7B-Instruct-v0.2"  # LLM model
  max_length: 512
  temperature: 0.3
  device: "auto"  # auto, cuda, cpu
  quantization: true  # Use 4-bit quantization

Risk Scoring Parameters

risk_scoring:
  severity:
    high_keywords: ["critical", "catastrophic", "severe"]
    medium_keywords: ["moderate", "significant"]
    low_keywords: ["minor", "slight"]
    default: 5

Text Processing

text_processing:
  min_review_length: 10
  negative_threshold: 0.3  # Sentiment threshold
  max_reviews_per_batch: 100
  enable_sentiment_filter: true

---

📁 Project Structure

Symboisis/
├── src/
│   ├── preprocessing.py       # Data preprocessing module
│   ├── llm_extractor.py       # LLM-based extraction
│   ├── risk_scoring.py        # Risk scoring engine
│   ├── fmea_generator.py      # Main FMEA generator
│   └── utils.py               # Utility functions
├── config/
│   └── config.yaml            # Configuration file
├── output/                    # Generated FMEAs
├── archive (3)/               # Sample car review data
├── app.py                     # Streamlit dashboard
├── cli.py                     # Command-line interface
├── examples.py                # Usage examples
├── requirements.txt           # Python dependencies
├── .env.example              # Environment variables template
└── README.md                 # This file

---

📊 Examples

Example 1: Customer Reviews

reviews = [
    "Brake failure during heavy rain, very dangerous!",
    "Engine overheated and seized, no warning lights."
]

fmea_df = generator.generate_from_text(reviews, is_file=False)

Output:

Failure Mode	Effect	Severity	Occurrence	Detection	RPN	Priority
Brake failure	Unable to stop	10	7	8	560	Critical
Engine seized	Vehicle breakdown	9	6	7	378	High

Example 2: Structured Data

Input CSV:

failure_mode,effect,cause,component
Brake system failure,Cannot stop vehicle,Worn brake pads,Brake System
Engine overheating,Engine damage,Coolant leak,Cooling System

fmea_df = generator.generate_from_structured('failures.csv')

Example 3: Real Car Reviews

# Process actual car review data
fmea_df = generator.generate_from_text('archive (3)/Scraped_Car_Review_ford.csv', is_file=True)

# Generate summary
from utils import generate_summary_report
print(generate_summary_report(fmea_df))

---

🔌 API Reference

FMEAGenerator

Main class for FMEA generation.

Methods

generate_from_text(text_input, is_file=False)

• Generate FMEA from unstructured text
• Args: text_input (str or list), is_file (bool)
• Returns: DataFrame

generate_from_structured(file_path)

• Generate FMEA from structured CSV/Excel
• Args: file_path (str)
• Returns: DataFrame

generate_hybrid(structured_file, text_input)

• Generate FMEA from both sources
• Args: structured_file (str), text_input (str or list)
• Returns: DataFrame

export_fmea(fmea_df, output_path, format='excel')

• Export FMEA to file
• Args: fmea_df (DataFrame), output_path (str), format (str)

DataPreprocessor

Handles data cleaning and preprocessing.

LLMExtractor

Extracts failure information using LLMs.

RiskScoringEngine

Calculates risk scores and RPN.

---

🧪 Testing

Run the examples to test the system:

python examples.py

This will:

1. Generate FMEA from sample reviews
2. Process structured data
3. Analyze real car reviews (if available)

---

🎯 Use Cases

Manufacturing

• Analyze equipment failure reports
• Process quality control data
• Generate preventive maintenance schedules

Automotive

• Process customer complaints
• Analyze warranty claims
• Identify safety issues

Healthcare

• Analyze adverse event reports
• Process medical device failures
• Improve patient safety

Software

• Analyze bug reports
• Process incident tickets
• Identify system vulnerabilities

---

🔬 Research Applications

This system is suitable for:

• Academic research papers
• Case studies
• Benchmarking studies
• Tool comparisons
• Industry reports

Key Advantages:

• Reproducible results
• Configurable parameters
• Comprehensive logging
• Export capabilities

---

🛠️ Troubleshooting

Issue: Model loading fails

# Use rule-based mode instead
python cli.py --text input.csv --output result.xlsx --no-model

Issue: Out of memory

• Enable quantization in config.yaml
• Use smaller batch sizes
• Use rule-based mode

Issue: Slow processing

• Use GPU if available
• Enable quantization
• Reduce batch size
• Use rule-based mode for faster results

---

📈 Performance

Processing Speed

Mode	Speed	Accuracy
LLM (GPU)	~2 reviews/sec	High
LLM (CPU)	~0.3 reviews/sec	High
Rule-based	~50 reviews/sec	Medium

Resource Requirements

Component	Minimum	Recommended
RAM	8GB	16GB
GPU	None	8GB VRAM
Disk	2GB	10GB

---

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests
5. Submit a pull request

---

📄 License

This project is licensed under the MIT License.

---

👥 Authors

Developed as a production-grade academic and industry project for automated FMEA generation.

---

🙏 Acknowledgments

• HuggingFace for transformer models
• Streamlit for dashboard framework
• Open-source community

---

📞 Support

For issues, questions, or suggestions:

• Open an issue on GitHub
• Check the documentation
• Run examples for guidance

---

🚀 Future Enhancements

• Fine-tuned domain-specific models
• Fuzzy FMEA support
• Real-time monitoring
• Multi-language support
• Integration with PLM systems
• Advanced analytics
• Mobile app

---

⚠️ LLM-Powered FMEA Generator

Transforming failure analysis with AI