Conceptual model

Conceptual Model

System Boundary

The simulation models the haulage system of a synthetic mine. It includes trucks, loaders (North and South pits), the primary crusher, and the road network (represented as a directed graph). The simulation tracks operations over an 8-hour shift, beginning with trucks at the parking area. Waste haulage, maintenance events, and loader/truck breakdowns (availability < 1.0) are excluded or considered out of scope unless explicitly requested, focusing solely on ore throughput.

Entities

• Trucks: The active entities moving through the system. Trucks have properties like payload capacity (100 tonnes), empty speed factor, and loaded speed factor.

Resources

• Loaders: L_N (capacity 1) and L_S (capacity 1). These constrain the loading process. Trucks queue if the loader is busy.
• Crusher: D_CRUSH (capacity 1). Constrains the dumping process.
• Constrained Road Segments: Edges with capacity constraints (e.g., E03_UP with capacity 1). These are modeled as shared resources. A truck must acquire the edge resource before it can commence travel on that segment and releases it upon arrival at the next node. For simplicity, opposite direction edges (e.g., E03_UP and E03_DOWN) are treated as independent resources as per the dataset metadata.

Events

• Simulation Start: Trucks are instantiated and dispatched from the PARK node.
• Travel to Loader: Truck requests constrained edges sequentially based on the shortest path.
• Join Loader Queue: Truck arrives at LOAD_N or LOAD_S and requests the loader resource.
• Loading: Truck captures the loader. Loading takes a stochastic amount of time.
• Travel to Crusher: Truck leaves the loader, payload is set to 100 tonnes, and it routes to the CRUSH node, acquiring constrained edges along the route.
• Join Crusher Queue: Truck arrives at CRUSH and requests the crusher resource.
• Dumping: Truck captures the crusher. Dumping takes a stochastic amount of time. Ore throughput is recorded.
• Return Travel: Truck becomes empty and returns to a loader.

State Variables

• Total tonnes delivered.
• Truck state (current location, loaded/empty).
• Queues at loaders, crusher, and constrained edges.
• Resource utilizations (busy time for loaders and crusher).
• Cycle times (from leaving the crusher/park to completing the next dump).

Assumptions

• Routing: Trucks route dynamically using Dijkstra’s shortest path based on expected travel time. Travel time is calculated as $Distance / (SpeedLimit \times SpeedFactor)$. If an edge is closed, its weight is set to infinity.
• Dispatching: When a truck is empty (either at start or after dumping), it selects the loader that provides the shortest expected time to load. The expected time includes travel time to the loader + (current loader queue length $\times$ mean load time) + mean load time.
• Independence of directional capacities: As stated in the edge metadata, narrow roads are simplified as separate directional edges. We assume they operate as independent queues, so a truck going up does not block a truck going down.
• No shift transitions: The simulation runs exactly for the shift length (8 hours) and abruptly stops or only counts completed dumps within the time window.

Performance Measures

• Expected Ore Throughput: Total tonnes dumped at the crusher over the shift.
• Tonnes per Hour: Total tonnes / Shift hours.
• Cycle Time: Average time to complete a full load-and-dump cycle.
• Utilizations: Percentage of shift time the crusher and loaders were actively processing trucks.
• Wait Times: Average time spent queuing at loaders and the crusher.

README

Synthetic Mine Throughput Simulation

This repository contains a discrete-event simulation model written in Python using SimPy. The simulation estimates ore throughput to a primary crusher over an 8-hour shift for a synthetic mine.

How to Install Dependencies

1. Create a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

2. Install the requirements:

   pip install -r requirements.txt

How to Run the Simulation

To execute the simulation and generate outputs:

source venv/bin/activate
python simulate.py

This will run all six predefined scenarios (30 replications each) and output:

• results.csv: Scenario and replication level metrics.
• summary.json: Aggregated statistics with confidence intervals.
• event_log.csv: Trace of simulation events.

Conceptual Model & Assumptions

The model tracks trucks as active entities navigating a directed graph of the mine topology. Loaders (L_N, L_S), the Crusher (D_CRUSH), and constrained road segments are modeled as shared resources with specific capacities.

Routing and Dispatching Logic

• Routing: Shortest-time path using Dijkstra’s algorithm. Edge weights are dynamically calculated as Distance / (SpeedLimit * SpeedFactor). If an edge is closed, it is ignored by the graph builder.
• Dispatching: When a truck becomes empty, it evaluates both loaders. It selects the loader offering the minimum expected time, calculated as: Travel_Time + (Queue_Length * Mean_Load_Time) + Mean_Load_Time.
• Road Resources: Constrained directional single-lane segments (like E03_UP) require a truck to acquire a capacity token before entering and release it upon leaving. Separate edges (e.g., E03_UP and E03_DOWN) are treated as independent queues.

For more details, see conceptual_model.md.

Operational Decision Questions

Based on the 30 replications per scenario with random seed control, here are the answers to the operational questions:

1. What is the expected ore throughput to the crusher during the baseline 8-hour shift?

The expected throughput is 12,603 tonnes (95% CI: 12,548 - 12,658 tonnes), which equates to ~1,575 tonnes per hour.

2. What are the likely bottlenecks in the haulage system?

The primary crusher is the major bottleneck. In the baseline scenario, its utilization reaches 92.8%. Trucks spend on average ~3.9 minutes queuing at the crusher. The loaders and road segments operate comfortably below their maximum capacities.

3. Does adding more trucks materially improve throughput, or does the system saturate?

The system is deeply saturated by the crusher. Adding more trucks (from 8 to 12) only marginally increases throughput to 12,823 tonnes (a ~1.7% increase), while crusher queue wait times explode to over 15 minutes, and overall truck utilization plummets from 77.5% to 54.6%.

4. Would improving the narrow ramp materially improve throughput?

No. The ramp_upgrade scenario yields 12,623 tonnes, statistically indistinguishable from the baseline. Because the crusher is the actual system constraint, widening the ramp merely delivers trucks to the crusher queue faster, where they end up waiting.

5. How sensitive is throughput to crusher service time?

Highly sensitive. The crusher_slowdown scenario (increasing mean dump time from 3.5 to 7.0 minutes) slashes throughput by nearly half to 6,483 tonnes. Crusher utilization remains pinned at 95.5%, and average queue times skyrocket to ~28.8 minutes.

6. What is the operational impact of losing the main ramp route?

Minimal. In the ramp_closed scenario, trucks reroute via the longer bypass network. The throughput drops slightly to 12,416 tonnes (~1.5% decrease). Since the crusher is the limiting factor, the extra travel time mostly cuts into the time trucks would have spent idling in the crusher queue anyway, buffering the impact of the longer route.

Limitations of the Model

• Traffic independence: Aside from specific constrained single-lane segments, trucks do not slow each other down on normal haul roads.
• Breakdowns: Unplanned maintenance and breakdowns (truck/loader availability) are excluded.
• Shift changes: The simulation assumes 100% operational efficiency right up to the exact 8-hour mark, without hot-seat changeover times or breaks.