Crusher Liner Wear Management System

A full-stack platform for tracking and forecasting the wear of crusher liners in minerals processing operations, built from raw 3D laser scan data through to production deployment.

The challenge

Gyratory and cone crushers are among the largest and most critical machines in a minerals processing plant. Their concave and mantle liners degrade continuously under extreme loads, and knowing when to schedule a liner change is a costly decision: too early wastes material, too late risks catastrophic failure. Traditionally, wear is estimated with manual caliper measurements taken during maintenance windows – a slow, imprecise, and sometimes dangerous process. 3D laser scanning offers a far richer picture, but the raw point clouds (millions of points in DXF or PTS format) require significant processing before they become actionable.

System architecture

1. Point cloud ingestion and parsing: Raw scan files (DXF, PTS) are parsed and transformed into cylindrical coordinate representations aligned to the crusher axis. Aggregation routines collapse the dense point cloud into interpretable radial-axial wear profiles.
2. Campaign and survey management: Each crusher liner installation defines a campaign. Multiple surveys (scans) are registered within a campaign, enabling wear progression tracking over time. The system manages metadata, alignment corrections, and reference geometries.
3. Wear trend modeling and forecasting: Wear rates are computed per profile zone, and regression models project remaining useful life. The system generates recommended change dates with confidence bounds, supporting maintenance planning.
4. Dual deployment architecture: The platform operates in two modes. A desktop application (Streamlit/Dash packaged with PyInstaller) serves offline mine sites with no internet connectivity. A web platform (Next.js frontend, FastAPI backend, PostgreSQL database, Redis caching) provides centralized management for multi-site operations.
5. Infrastructure: Docker Swarm orchestration with Traefik as reverse proxy and load balancer. Ansible playbooks automate provisioning and deployment across environments.

Technology stack

• Backend: FastAPI (Python), PostgreSQL, Redis, Celery for async tasks
• Frontend: Next.js (React/TypeScript)
• Desktop: Streamlit and Dash applications, packaged via PyInstaller
• 3D Processing: NumPy, SciPy, Open3D for point cloud manipulation
• Infrastructure: Docker Swarm, Traefik, Ansible, Nginx
• ML/Forecasting: scikit-learn, custom regression models for wear projection

Note: This description reflects the general type and architecture of systems I have built as a consultant for a mining equipment services company. Specific client details, operational data, and proprietary methodologies are omitted.