Data-Driven Scheduling for Efficient Cast House Operations
How a leading aluminium manufacturer eliminated planning blind spots, reduced fuel oil consumption by 3.16 L/ton, and replaced shift-by-shift guesswork with a data-driven scheduling engine across its Remelt–Transfer–Cast production line.
Business Objective
The Cast House at a large aluminium plant runs a continuous multi-stage process: scrap is recycled and remelted, hot metal is alloyed in holding furnaces, then cast into rolling ingots across four casting stations.
If casting was scheduled for 8:00 AM, the remelt furnace had to begin melting by 3:00 or 4:00 in the morning. But there was no tool to plan this – no system that tied casting readiness back to melt start times, transfer windows, and furnace availability. Shift managers relied entirely on personal experience and verbal coordination across stations.
Three problems driving cost and unpredictability:
No Standardized Process Planning
No formal Standard Operating Procedures(SOP) existed for Remelt or Recycle operations. Each shift planned differently, entirely on individual experience. No shared baseline meant no path to improvement.
Four Stations, Zero Coordination
Recycle, Remelt, Holding, and Casting operated in silos within the same factory. The casting team had no visibility into remelt status. Delays at one station cascaded silently – by the time the problem surfaced, the hot metal had already been held too long. Physical tracking tags were missing for critical process parameters.
Fuel Consumption Running Above Target
Fuel oil usage was at 48 litres per ton against a target of 45 litres per ton. The driver: avoidable holding time in the remelt furnace. When casting preparation overran, the furnace kept burning fuel with no productive output.
The Solution
A scheduling engine built around this plant’s reality.
A constraint-based scheduling engine, built specifically around this plant’s operating reality – its shift structures, meal break timings, metallurgical readiness requirements, and grade-level production parameters.
Data Pipeline Live PLC(Programmable Logic Controller) data from both the Recycle and Remelt furnaces flows via Kepware OPC UA(Unified Architecture) across the OT–IT(Operational Technology / Information Technology) network boundary into an Azure PostgreSQL database. Process data refreshes every minute. User inputs (production requirement and alloy grade per batch) are captured through the application UI.
Scheduling Model The optimization model treats the production sequence as a multi-stage flow-shop scheduling problem: Recycling → Melting → Transfer → Casting-1 → Casting-2
A sequential greedy heuristic schedules batches one by one under global plant constraints – shift time windows, meal break offsets, metallurgical readiness per grade and mould, and resource precedence rules. Decision variables define start and end times for each stage per batch. The objective: minimize total cycle time while respecting all constraints with zero violations.
The Application
A Streamlit-based interface deployed, with role-based access for operators, shift in-charges, and administrators.
Live Production Tracking
Real-time batch status, stage-by-stage progress, and fuel Key Performance Indicators (KPIs) across all active batches
Optimized Plan Generation
Input grade, mould size, and equipment availability; get a full batch-wise schedule with precise start and end times for every stage
Production Plan Analytics
7-day batch trends, daily average batch times, historical schedule review for continuous improvement
Daily Email Reports
Automated performance summaries to all Heads of Department (HODs) and stakeholders every 24 hours
Business Impact & Value Delivered
Digitalized production planning has transformed shift operations from being memory-driven to system-driven. Instead of relying on individual experience or manual coordination, the shift in-charge now inputs grade and mould requirements into the system, which generates an optimized production schedule. This ensures consistency across shifts, improves transparency and traceability, and creates a structured process that can be continuously analyzed and improved over time.