Maintenance Scheduler System

An AI-powered maintenance planning system that uses Deep Q-Learning to optimize equipment maintenance scheduling.

Table of Contents

1. System Overview
2. Project Structure
3. Architecture
4. Installation & Setup
5. Data Model
6. AI Model
7. Key Features
8. Usage Guide
9. API Reference
10. Deployment Instructions

System Overview

The Maintenance Scheduler is an intelligent system that helps optimize maintenance planning by considering multiple factors including:

• Equipment condition and history
• Current issues and priorities
• Resource availability
• Cost constraints
• Risk factors

The system uses Deep Q-Learning to make optimal maintenance decisions while balancing preventive and corrective maintenance needs.

Project Structure

maintenance-scheduler/
├── data/                    # Data files and templates
│   ├── data_dictionary.md   # Data field definitions
│   ├── maintenance_schedule.xlsx
│   ├── training_history.csv
│   ├── sample_data/        # Sample datasets
│   └── uploads/            # User uploaded files
├── models/                 # ML model implementations
│   ├── dqn_agent.py       # Deep Q-Network agent
│   ├── maintenance_env.py  # Maintenance environment
│   └── saved/             # Saved model weights
├── tests/                 # Test files
├── ui/                    # Frontend application
│   └── webapp/           # UI5 web application
├── utils/                # Utility functions
├── server.py            # Backend Flask server
├── train.py             # Model training script
└── requirements.txt     # Python dependencies

Architecture

Backend (Python/Flask)

The backend provides REST APIs for:

• Issue management
• Schedule generation
• Data export/import
• Model inference

Key endpoints:

• POST /api/upload_issues: Process new maintenance issues
• POST /api/generate_schedule: Create optimized schedules
• GET /api/download_template: Get issue reporting template
• GET /api/download_schedule: Export maintenance schedules

Frontend (OpenUI5)

Web-based interface featuring:

• Issue upload interface
• Schedule visualization
• Template downloads
• Schedule exports

ML Component

Deep Q-Learning implementation with:

• State space (8 dimensions):

  1. Days since maintenance (normalized)
  2. Equipment age
  3. Criticality score
  4. Maintenance cycle completion
  5. Cost ratio
  6. Breakdown risk
  7. Issue priority
  8. Workload factor

• Actions:

  • Schedule maintenance
  • Postpone maintenance

Installation & Setup

1. Clone the repository
2. Install Python dependencies:

pip install -r requirements.txt

3. Install UI dependencies:

cd ui/webapp
npm install

4. Start the backend server:

python server.py

5. Start the UI development server:

cd ui/webapp
ui5 serve

Data Model

Equipment Master Data

• Equipment identification and specifications
• Location and technical details
• Criticality ratings:
  • A: Critical (1.0) - Severe impact
  • B: Medium (0.6) - Moderate impact
  • C: Lower (0.3) - Lesser impact
• Cost and maintenance parameters

Maintenance History

• Past maintenance records
• Maintenance types:
  • PM01: Regular maintenance
  • PM02: Major maintenance
  • PM03: Emergency maintenance
• Performance metrics and costs

Current Issues

• Active maintenance notifications
• Priority levels (1-4)
• Issue types:
  • PERF: Performance
  • ELEC: Electrical
  • HYDR: Hydraulic
  • MECH: Mechanical
• Impact categories:
  • Production Stop
  • Performance Degradation
  • Quality Impact
  • Safety Risk

AI Model

Decision Making Process

The DQN agent makes maintenance decisions by considering:

1. Equipment condition and age
2. Historical maintenance patterns
3. Current issues and priorities
4. Resource availability
5. Cost implications

Risk Assessment

Risk calculation incorporates:

• Equipment age and criticality
• Maintenance history
• Current issues
• Historical performance
• Cost factors

Reward Structure

The model optimizes for:

• Timely maintenance
• Cost efficiency
• Risk minimization
• Resource utilization
• Issue priority handling

Key Features

Predictive Scheduling

• AI-driven maintenance timing
• Risk-based prioritization
• Balanced maintenance approach

Resource Optimization

• Workload distribution
• Budget management
• Priority-based allocation

Cost Management

• Budget tracking
• Cost estimation
• Optimization for efficiency

Usage Guide

Issue Reporting

1. Download the issue template (/api/download_template)
2. Fill in required information:
   • Equipment ID
   • Issue type
   • Priority
   • Description
3. Upload completed template
4. Review generated schedule

Schedule Generation

1. Upload current issues file
2. System processes equipment states
3. AI model generates recommendations
4. Review generated schedule
5. Export schedule as Excel file

Maintenance Types

• PM01 (Regular)

  • Lower priority issues
  • Preventive maintenance
  • 7-30 days scheduling window

• PM02 (Major)

  • Priority 2 issues
  • Mechanical/Electrical issues
  • 2-3 days scheduling window

• PM03 (Emergency)

  • Priority 1 issues
  • Hydraulic issues
  • Next-day maintenance

API Reference

POST /api/upload_issues

Upload current issues for processing

Request:

• Method: POST
• Content-Type: multipart/form-data
• Body: CSV file

Response:

{
  "equipment_id": "string",
  "maintenance_type": "string",
  "suggested_date": "string",
  "priority": "string",
  "confidence": "number",
  "breakdown_risk": "number"
}

GET /api/download_template

Download issue reporting template

Response:

• Content-Type: text/csv
• File: current_issues_template.csv

GET /api/download_schedule

Download generated maintenance schedule

Response:

• Content-Type: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
• File: maintenance_schedule.xlsx

Deployment Instructions

Backend Deployment (PythonAnywhere)

1. Create an account on PythonAnywhere (https://www.pythonanywhere.com)
2. Go to the Web tab and create a new web app
3. Choose Manual Configuration and Python 3.9
4. In the Code section:
   • Upload all project files
   • Set the working directory to your project root
   • Set the WSGI configuration file path to wsgi.py
   • Set up a virtual environment and install requirements:

     mkvirtualenv maintenance-env --python=/usr/bin/python3.9
     pip install -r requirements.txt
     

5. Set up environment variables in the web app configuration:
   • FLASK_ENV=production
   • DATABASE_URL=(your database URL)
   • CORS_ORIGINS=https://[your-github-username].github.io

Frontend Deployment (GitHub Pages)

1. Create a new GitHub repository
2. Push your code to the repository:

   git init
   git add .
   git commit -m "Initial commit"
   git remote add origin https://github.com/[your-username]/maintenance-scheduler.git
   git push -u origin main
   

3. Enable GitHub Pages:
   • Go to repository Settings > Pages
   • Select 'gh-pages' branch and '/root' folder
   • Save the configuration

The frontend will be available at: https://[your-github-username].github.io/maintenance-scheduler/ The backend will be available at: https://[your-pythonanywhere-username].pythonanywhere.com/

Development Setup

1. Clone the repository
2. Install backend dependencies:

   pip install -r requirements.txt
   

3. Install frontend dependencies:

   cd ui/webapp
   npm install
   

4. Start the development servers:
   • Backend: python server.py
   • Frontend: cd ui/webapp && ui5 serve