🛠️ Predictive Maintenance Strategy System (Industrial AI)

ROI-Focused Machine Learning | $88.2M Simulated Annual Savings

📌 Executive Summary

This project demonstrates how a Data Strategist uses AI to impact the bottom line. By predicting machinery failure before it happens, this system moves a company from reactive to proactive maintenance, drastically reducing downtime and emergency repair costs.

🚀 Key Business Results

• Financial Impact: Reduced simulated maintenance costs from $100M to $11.8M (88% Savings).
• Model Accuracy: Achieved an R² score of 0.81 in predicting Remaining Useful Life (RUL).
• Risk Mitigation: Successfully prevented 98% of catastrophic engine failures.

🛠️ Tech Stack & Workflow

• Database: MySQL (Window Functions for Time-Series Feature Engineering).
• Analysis: Python (Pandas, Seaborn) for Exploratory Data Analysis.
• Machine Learning: Random Forest Regressor with RUL Clipping for optimized performance.
• UI/Deployment: Streamlit Dashboard for real-time monitoring.

📈 Feature Engineering (The "Strategist" Flex)

Instead of using raw, noisy sensor data, I used SQL Window Functions to calculate 10-day rolling averages. This smoothed out sensor "jitter" and provided the model with a clear signal of degradation, boosting accuracy from 0.66 to 0.81.

📊 Visualizations

1. Business ROI Analysis

2. Live Strategy Dashboard

3. Rolling Averages

📂 Project Structure

• sql_scripts/: Database schema and engineering views.
• python_scripts/: Ingestion, EDA, Training, and Cost Analysis.
• src/app.py: Streamlit dashboard code.