Using AI/ML and Space Technology to Identify Manganese Reserves and Overcome Production Shortfalls.
“MOIL, India's top manganese producer, struggles to find new reserves and predict production issues. This problem asks teams to build an AI tool that uses satellite data and company records to accurately locate manganese, foresee production shortfalls, and suggest solutions to keep ore flowing.”
Deconstructed Problem Blueprint
WHAT TO BUILD?
An AI/ML-based software solution with a user-friendly dashboard that integrates geological, operational, and space technology data to predict manganese reserves and production shortfalls, offering corrective actions.
WHO IS IT FOR?
MOIL Limited, specifically their planning, operations, and management teams, who need better tools for reserve estimation and production scheduling.
WHY DOES IT EXIST?
Current manual methods for reserve estimation and production planning are slow, prone to errors, and often lead to mismatches between expected and actual ore production, causing losses and supply instability.
INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE
Geological data (MOIL), historical production records (MOIL), equipment performance data (MOIL), drilling results (MOIL), satellite imagery/APIs for rainfall, soil moisture, vegetation index, and land temperature.
Data ingestion and preprocessing (cleaning, feature engineering); AI/ML models for spatial analysis and semantic segmentation to identify and map manganese reserves from surface/sub-surface indicators; Time-series and predictive AI/ML models to forecast production shortfalls based on equipment, weather, and blasting data; A recommendation engine to generate optimal corrective actions (e.g., schedule adjustments, equipment redeployment).
An interactive dashboard displaying predicted manganese reserve maps, real-time production trends, identified risks of production shortfalls (with severity), and actionable, recommended corrective steps.
CONSTRAINTS
- Accuracy of reserve identification (sub-meter level where possible)
- Timeliness of shortfall predictions and recommendations (near real-time)
- Integration capability with MOIL's existing data infrastructure (simulated for hackathon)
- User-friendliness and intuitiveness of the dashboard for non-technical users
- Robustness to varying data quality and missing values
- Scalability to cover large mining areas and multiple sites
- Ethical considerations for data usage and AI recommendations
JURY METRICS
- Precision, Recall, F1-score, and IoU (Intersection over Union) for manganese reserve mapping accuracy
- RMSE (Root Mean Squared Error) or MAE (Mean Absolute Error) for production shortfall prediction accuracy
- Reduction in simulated production shortfalls (percentage)
- Latency of prediction and recommendation generation
- User Experience (UX) score for dashboard usability and clarity
- Cost-effectiveness of the proposed solution (simulated operational savings)
- System uptime and reliability
CORE BOTTLENECK
The single hardest challenge will be integrating highly disparate data sources (static geological maps, dynamic satellite imagery, time-series equipment logs) into a cohesive model that can accurately predict both static reserves and dynamic operational shortfalls, especially with limited access to MOIL's proprietary ground truth data for training and validation.
Why You Should Choose This PS (Pros)
- High real-world impact and direct economic benefit for a PSU (MOIL) and national resource security.
- Clear problem statement with well-defined inputs and expected outputs (dashboard).
- Strong 'wow' factor by combining cutting-edge AI/ML with advanced space technology.
- Addresses a critical and tangible pain point for a major industry.
- Opportunity to showcase interdisciplinary skills (AI, ML, GIS, UI/UX, domain knowledge).
Challenges & Risks to Watch (Cons)
- Significant challenge in acquiring or simulating realistic, high-quality geological and operational data from MOIL for training and validation.
- Requires deep domain knowledge in geology, mining operations, and remote sensing, which most student teams may lack.
- Building robust and accurate AI/ML models for both spatial mapping and time-series prediction is complex.
- High expectations for accuracy and practical utility from the Ministry/Organization.
- Potential for data privacy and security concerns if real MOIL data were to be used (simulated for hackathon).
Recommended Tech Stack & Grand Finale Features
Killer Features to Impress the SIH Jury
View Official SIH 2026 Portal Raw Text▼
MOIL Limited is the largest producer of Manganese Ore in India. To meet future demand, it is important to accurately identify available reserves and avoid production shortfalls. At present, reserve estimation and production planning are mainly based on manual surveys, drilling results, and production records. These methods are time-consuming and sometimes lead to a mismatch between expected and actual ore production. Detailed
The challenge is to develop an AI/ML-based solution that uses geological data,historical production, equipment performance, and satellite/space technology inputs (such as rainfall, soil moisture, vegetation index, and land temperature) to:
The expected solution is a user-friendly dashboard that shows predicted reserves, production trends, possible risks of shortfall, and recommended corrective steps.
This will help MOIL improve planning, reduce losses, and ensure steady ore supply to customers.