{ "benchmark_id": "001_synthetic_mine_throughput", "scenarios": { "baseline": { "scenario_id": "baseline", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 12546.666667, "ci95_low": 12491.42646, "ci95_high": 12601.906874, "std": 147.935991, "n": 30 }, "tonnes_per_hour": { "mean": 1568.333333, "ci95_low": 1561.428307, "ci95_high": 1575.238359, "std": 18.491999, "n": 30 }, "average_truck_cycle_time_min": { "mean": 29.660399, "ci95_low": 29.527727, "ci95_high": 29.793072, "std": 0.355303, "n": 30 }, "average_truck_utilisation": { "mean": 0.994422, "ci95_low": 0.99384, "ci95_high": 0.995003, "std": 0.001556, "n": 30 }, "crusher_utilisation": { "mean": 0.911607, "ci95_low": 0.907362, "ci95_high": 0.915852, "std": 0.011368, "n": 30 }, "average_loader_queue_time_min": { "mean": 2.50709, "ci95_low": 2.405992, "ci95_high": 2.608189, "std": 0.270747, "n": 30 }, "average_crusher_queue_time_min": { "mean": 3.279038, "ci95_low": 3.105964, "ci95_high": 3.452112, "std": 0.4635, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.601581, "ci95_low": 0.595071, "ci95_high": 0.608091, "std": 0.017435, "n": 30 }, "mean_queue_wait_min": { "mean": 2.618506, "ci95_low": 2.495675, "ci95_high": 2.741336, "std": 0.328946, "n": 30 }, "services_completed": { "mean": 44.3, "ci95_low": 43.789045, "ci95_high": 44.810955, "std": 1.368362, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.803143, "ci95_low": 0.79463, "ci95_high": 0.811655, "std": 0.022797, "n": 30 }, "mean_queue_wait_min": { "mean": 2.449391, "ci95_low": 2.322927, "ci95_high": 2.575854, "std": 0.338675, "n": 30 }, "services_completed": { "mean": 85.333333, "ci95_low": 84.767176, "ci95_high": 85.89949, "std": 1.516196, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.911607, "ci95_low": 0.907362, "ci95_high": 0.915852, "std": 0.011368, "n": 30 }, "mean_queue_wait_min": { "mean": 3.279038, "ci95_low": 3.105964, "ci95_high": 3.452112, "std": 0.4635, "n": 30 }, "services_completed": { "mean": 125.466667, "ci95_low": 124.914265, "ci95_high": 126.019069, "std": 1.47936, "n": 30 } }, "edges": { "E03_DOWN": { "edge_id": "E03_DOWN", "utilisation": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "traversal_count": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 } }, "E03_UP": { "edge_id": "E03_UP", "utilisation": { "mean": 0.052746, "ci95_low": 0.052132, "ci95_high": 0.053359, "std": 0.001643, "n": 30 }, "mean_queue_wait_min": { "mean": 10.890609, "ci95_low": 10.727612, "ci95_high": 11.053605, "std": 0.436512, "n": 30 }, "mean_traversal_time_min": { "mean": 3.164733, "ci95_low": 3.127922, "ci95_high": 3.201544, "std": 0.098581, "n": 30 }, "traversal_count": { "mean": 8.0, "ci95_low": 8.0, "ci95_high": 8.0, "std": 0.0, "n": 30 } }, "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.351171, "ci95_low": 0.349337, "ci95_high": 0.353004, "std": 0.004911, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000893, "ci95_low": 0.000217, "ci95_high": 0.001569, "std": 0.001811, "n": 30 }, "mean_traversal_time_min": { "mean": 1.34707, "ci95_low": 1.342926, "ci95_high": 1.351213, "std": 0.011096, "n": 30 }, "traversal_count": { "mean": 125.133333, "ci95_low": 124.572295, "ci95_high": 125.694372, "std": 1.502488, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.420842, "ci95_low": 0.418574, "ci95_high": 0.42311, "std": 0.006074, "n": 30 }, "mean_queue_wait_min": { "mean": 0.150605, "ci95_low": 0.140048, "ci95_high": 0.161162, "std": 0.028273, "n": 30 }, "mean_traversal_time_min": { "mean": 1.586026, "ci95_low": 1.581245, "ci95_high": 1.590806, "std": 0.012802, "n": 30 }, "traversal_count": { "mean": 127.366667, "ci95_low": 126.750491, "ci95_high": 127.982842, "std": 1.650148, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.219214, "ci95_low": 0.216676, "ci95_high": 0.221752, "std": 0.006797, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.389809, "ci95_low": 2.375487, "ci95_high": 2.404131, "std": 0.038354, "n": 30 }, "traversal_count": { "mean": 44.033333, "ci95_low": 43.54812, "ci95_high": 44.518547, "std": 1.299425, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.191156, "ci95_low": 0.189069, "ci95_high": 0.193244, "std": 0.005589, "n": 30 }, "mean_queue_wait_min": { "mean": 0.014375, "ci95_low": 0.010832, "ci95_high": 0.017918, "std": 0.009489, "n": 30 }, "mean_traversal_time_min": { "mean": 2.030075, "ci95_low": 2.018737, "ci95_high": 2.041414, "std": 0.030364, "n": 30 }, "traversal_count": { "mean": 45.2, "ci95_low": 44.746367, "ci95_high": 45.653633, "std": 1.214851, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.416421, "ci95_low": 0.412653, "ci95_high": 0.420188, "std": 0.01009, "n": 30 }, "mean_queue_wait_min": { "mean": 0.004326, "ci95_low": 0.001752, "ci95_high": 0.0069, "std": 0.006894, "n": 30 }, "mean_traversal_time_min": { "mean": 2.355096, "ci95_low": 2.343477, "ci95_high": 2.366714, "std": 0.031114, "n": 30 }, "traversal_count": { "mean": 84.866667, "ci95_low": 84.331956, "ci95_high": 85.401378, "std": 1.431983, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.361034, "ci95_low": 0.358756, "ci95_high": 0.363312, "std": 0.0061, "n": 30 }, "mean_queue_wait_min": { "mean": 0.015961, "ci95_low": 0.01208, "ci95_high": 0.019841, "std": 0.010393, "n": 30 }, "mean_traversal_time_min": { "mean": 2.001987, "ci95_low": 1.994417, "ci95_high": 2.009556, "std": 0.020272, "n": 30 }, "traversal_count": { "mean": 86.566667, "ci95_low": 86.0147, "ci95_high": 87.118633, "std": 1.478194, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.911607, "mean_queue_wait_min": 3.279038, "composite_score": 2.989194 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.803143, "mean_queue_wait_min": 2.449391, "composite_score": 1.96721 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.601581, "mean_queue_wait_min": 2.618506, "composite_score": 1.575243 }, { "resource_id": "E03_UP", "resource_kind": "edge", "utilisation_mean": 0.052746, "mean_queue_wait_min": 10.890609, "composite_score": 0.574431 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.420842, "mean_queue_wait_min": 0.150605, "composite_score": 0.063381 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "trucks_4": { "scenario_id": "trucks_4", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 7650.0, "ci95_low": 7611.083146, "ci95_high": 7688.916854, "std": 104.22125, "n": 30 }, "tonnes_per_hour": { "mean": 956.25, "ci95_low": 951.385393, "ci95_high": 961.114607, "std": 13.027656, "n": 30 }, "average_truck_cycle_time_min": { "mean": 24.418577, "ci95_low": 24.299544, "ci95_high": 24.53761, "std": 0.318776, "n": 30 }, "average_truck_utilisation": { "mean": 0.996168, "ci95_low": 0.995441, "ci95_high": 0.996895, "std": 0.001947, "n": 30 }, "crusher_utilisation": { "mean": 0.556981, "ci95_low": 0.55255, "ci95_high": 0.561412, "std": 0.011866, "n": 30 }, "average_loader_queue_time_min": { "mean": 0.690894, "ci95_low": 0.647821, "ci95_high": 0.733966, "std": 0.11535, "n": 30 }, "average_crusher_queue_time_min": { "mean": 0.69713, "ci95_low": 0.64953, "ci95_high": 0.744729, "std": 0.127473, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.322521, "ci95_low": 0.314497, "ci95_high": 0.330545, "std": 0.021488, "n": 30 }, "mean_queue_wait_min": { "mean": 0.640979, "ci95_low": 0.533094, "ci95_high": 0.748865, "std": 0.288922, "n": 30 }, "services_completed": { "mean": 23.566667, "ci95_low": 23.0147, "ci95_high": 24.118633, "std": 1.478194, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.516733, "ci95_low": 0.509515, "ci95_high": 0.523951, "std": 0.01933, "n": 30 }, "mean_queue_wait_min": { "mean": 0.710502, "ci95_low": 0.652839, "ci95_high": 0.768166, "std": 0.154425, "n": 30 }, "services_completed": { "mean": 55.1, "ci95_low": 54.307279, "ci95_high": 55.892721, "std": 2.122945, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.556981, "ci95_low": 0.55255, "ci95_high": 0.561412, "std": 0.011866, "n": 30 }, "mean_queue_wait_min": { "mean": 0.69713, "ci95_low": 0.64953, "ci95_high": 0.744729, "std": 0.127473, "n": 30 }, "services_completed": { "mean": 76.5, "ci95_low": 76.110831, "ci95_high": 76.889169, "std": 1.042213, "n": 30 } }, "edges": { "E03_DOWN": { "edge_id": "E03_DOWN", "utilisation": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "traversal_count": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 } }, "E03_UP": { "edge_id": "E03_UP", "utilisation": { "mean": 0.026139, "ci95_low": 0.025669, "ci95_high": 0.02661, "std": 0.00126, "n": 30 }, "mean_queue_wait_min": { "mean": 4.467327, "ci95_low": 4.349859, "ci95_high": 4.584796, "std": 0.314586, "n": 30 }, "mean_traversal_time_min": { "mean": 3.136727, "ci95_low": 3.080274, "ci95_high": 3.193181, "std": 0.151186, "n": 30 }, "traversal_count": { "mean": 4.0, "ci95_low": 4.0, "ci95_high": 4.0, "std": 0.0, "n": 30 } }, "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.214677, "ci95_low": 0.21319, "ci95_high": 0.216165, "std": 0.003984, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000572, "ci95_low": -8.5e-05, "ci95_high": 0.001228, "std": 0.001758, "n": 30 }, "mean_traversal_time_min": { "mean": 1.350516, "ci95_low": 1.344308, "ci95_high": 1.356724, "std": 0.016626, "n": 30 }, "traversal_count": { "mean": 76.3, "ci95_low": 75.918316, "ci95_high": 76.681684, "std": 1.022168, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.255324, "ci95_low": 0.253708, "ci95_high": 0.25694, "std": 0.004328, "n": 30 }, "mean_queue_wait_min": { "mean": 0.102811, "ci95_low": 0.091574, "ci95_high": 0.114049, "std": 0.030094, "n": 30 }, "mean_traversal_time_min": { "mean": 1.584795, "ci95_low": 1.578713, "ci95_high": 1.590876, "std": 0.016286, "n": 30 }, "traversal_count": { "mean": 77.333333, "ci95_low": 76.91346, "ci95_high": 77.753207, "std": 1.124441, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.11507, "ci95_low": 0.111845, "ci95_high": 0.118294, "std": 0.008635, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.366428, "ci95_low": 2.34542, "ci95_high": 2.387435, "std": 0.056259, "n": 30 }, "traversal_count": { "mean": 23.333333, "ci95_low": 22.750439, "ci95_high": 23.916228, "std": 1.561019, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.100536, "ci95_low": 0.098338, "ci95_high": 0.102734, "std": 0.005887, "n": 30 }, "mean_queue_wait_min": { "mean": 0.004926, "ci95_low": 0.000147, "ci95_high": 0.009704, "std": 0.012797, "n": 30 }, "mean_traversal_time_min": { "mean": 2.022172, "ci95_low": 2.005531, "ci95_high": 2.038812, "std": 0.044565, "n": 30 }, "traversal_count": { "mean": 23.866667, "ci95_low": 23.359648, "ci95_high": 24.373685, "std": 1.357821, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.268335, "ci95_low": 0.264534, "ci95_high": 0.272135, "std": 0.010178, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000994, "ci95_low": -1.6e-05, "ci95_high": 0.002004, "std": 0.002706, "n": 30 }, "mean_traversal_time_min": { "mean": 2.346368, "ci95_low": 2.334254, "ci95_high": 2.358483, "std": 0.032444, "n": 30 }, "traversal_count": { "mean": 54.9, "ci95_low": 54.107279, "ci95_high": 55.692721, "std": 2.122945, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.231461, "ci95_low": 0.228477, "ci95_high": 0.234445, "std": 0.007991, "n": 30 }, "mean_queue_wait_min": { "mean": 0.006332, "ci95_low": 0.003812, "ci95_high": 0.008853, "std": 0.00675, "n": 30 }, "mean_traversal_time_min": { "mean": 1.994958, "ci95_low": 1.984515, "ci95_high": 2.005401, "std": 0.027967, "n": 30 }, "traversal_count": { "mean": 55.7, "ci95_low": 54.939472, "ci95_high": 56.460528, "std": 2.036732, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.556981, "mean_queue_wait_min": 0.69713, "composite_score": 0.388288 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.516733, "mean_queue_wait_min": 0.710502, "composite_score": 0.36714 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.322521, "mean_queue_wait_min": 0.640979, "composite_score": 0.206729 }, { "resource_id": "E03_UP", "resource_kind": "edge", "utilisation_mean": 0.026139, "mean_queue_wait_min": 4.467327, "composite_score": 0.116773 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.255324, "mean_queue_wait_min": 0.102811, "composite_score": 0.02625 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "trucks_12": { "scenario_id": "trucks_12", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 12906.666667, "ci95_low": 12826.440013, "ci95_high": 12986.89332, "std": 214.850924, "n": 30 }, "tonnes_per_hour": { "mean": 1613.333333, "ci95_low": 1603.305002, "ci95_high": 1623.361665, "std": 26.856365, "n": 30 }, "average_truck_cycle_time_min": { "mean": 42.683402, "ci95_low": 42.408501, "ci95_high": 42.958303, "std": 0.736198, "n": 30 }, "average_truck_utilisation": { "mean": 0.98791, "ci95_low": 0.987156, "ci95_high": 0.988665, "std": 0.00202, "n": 30 }, "crusher_utilisation": { "mean": 0.937067, "ci95_low": 0.934207, "ci95_high": 0.939926, "std": 0.007659, "n": 30 }, "average_loader_queue_time_min": { "mean": 3.473894, "ci95_low": 3.336878, "ci95_high": 3.61091, "std": 0.366936, "n": 30 }, "average_crusher_queue_time_min": { "mean": 14.234655, "ci95_low": 13.888087, "ci95_high": 14.581224, "std": 0.928127, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.640663, "ci95_low": 0.632903, "ci95_high": 0.648422, "std": 0.02078, "n": 30 }, "mean_queue_wait_min": { "mean": 3.030277, "ci95_low": 2.869517, "ci95_high": 3.191037, "std": 0.430523, "n": 30 }, "services_completed": { "mean": 47.366667, "ci95_low": 46.816445, "ci95_high": 47.916888, "std": 1.473521, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.844934, "ci95_low": 0.835144, "ci95_high": 0.854725, "std": 0.02622, "n": 30 }, "mean_queue_wait_min": { "mean": 3.711354, "ci95_low": 3.540127, "ci95_high": 3.882582, "std": 0.458556, "n": 30 }, "services_completed": { "mean": 89.633333, "ci95_low": 88.909515, "ci95_high": 90.357151, "std": 1.93842, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.937067, "ci95_low": 0.934207, "ci95_high": 0.939926, "std": 0.007659, "n": 30 }, "mean_queue_wait_min": { "mean": 14.234655, "ci95_low": 13.888087, "ci95_high": 14.581224, "std": 0.928127, "n": 30 }, "services_completed": { "mean": 129.066667, "ci95_low": 128.2644, "ci95_high": 129.868933, "std": 2.148509, "n": 30 } }, "edges": { "E03_DOWN": { "edge_id": "E03_DOWN", "utilisation": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "traversal_count": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 } }, "E03_UP": { "edge_id": "E03_UP", "utilisation": { "mean": 0.078432, "ci95_low": 0.077577, "ci95_high": 0.079287, "std": 0.00229, "n": 30 }, "mean_queue_wait_min": { "mean": 16.896822, "ci95_low": 16.642584, "ci95_high": 17.151061, "std": 0.680864, "n": 30 }, "mean_traversal_time_min": { "mean": 3.137268, "ci95_low": 3.103068, "ci95_high": 3.171468, "std": 0.091589, "n": 30 }, "traversal_count": { "mean": 12.0, "ci95_low": 12.0, "ci95_high": 12.0, "std": 0.0, "n": 30 } }, "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.361146, "ci95_low": 0.358559, "ci95_high": 0.363732, "std": 0.006927, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000812, "ci95_low": 0.000111, "ci95_high": 0.001513, "std": 0.001877, "n": 30 }, "mean_traversal_time_min": { "mean": 1.34727, "ci95_low": 1.343453, "ci95_high": 1.351087, "std": 0.010222, "n": 30 }, "traversal_count": { "mean": 128.666667, "ci95_low": 127.83845, "ci95_high": 129.494883, "std": 2.218004, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.445851, "ci95_low": 0.442826, "ci95_high": 0.448877, "std": 0.008103, "n": 30 }, "mean_queue_wait_min": { "mean": 0.164807, "ci95_low": 0.154506, "ci95_high": 0.175107, "std": 0.027586, "n": 30 }, "mean_traversal_time_min": { "mean": 1.589553, "ci95_low": 1.585082, "ci95_high": 1.594025, "std": 0.011974, "n": 30 }, "traversal_count": { "mean": 134.633333, "ci95_low": 133.827777, "ci95_high": 135.438889, "std": 2.157318, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.234915, "ci95_low": 0.231856, "ci95_high": 0.237974, "std": 0.008193, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.388816, "ci95_low": 2.376884, "ci95_high": 2.400747, "std": 0.031953, "n": 30 }, "traversal_count": { "mean": 47.2, "ci95_low": 46.668295, "ci95_high": 47.731705, "std": 1.423933, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.201635, "ci95_low": 0.198756, "ci95_high": 0.204514, "std": 0.00771, "n": 30 }, "mean_queue_wait_min": { "mean": 0.014353, "ci95_low": 0.009456, "ci95_high": 0.019249, "std": 0.013114, "n": 30 }, "mean_traversal_time_min": { "mean": 2.007913, "ci95_low": 1.997481, "ci95_high": 2.018344, "std": 0.027936, "n": 30 }, "traversal_count": { "mean": 48.2, "ci95_low": 47.576713, "ci95_high": 48.823287, "std": 1.669193, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.437121, "ci95_low": 0.432988, "ci95_high": 0.441254, "std": 0.011069, "n": 30 }, "mean_queue_wait_min": { "mean": 0.005595, "ci95_low": 0.002894, "ci95_high": 0.008297, "std": 0.007234, "n": 30 }, "mean_traversal_time_min": { "mean": 2.354825, "ci95_low": 2.34639, "ci95_high": 2.36326, "std": 0.022591, "n": 30 }, "traversal_count": { "mean": 89.1, "ci95_low": 88.344546, "ci95_high": 89.855454, "std": 2.023142, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.378439, "ci95_low": 0.374909, "ci95_high": 0.381969, "std": 0.009454, "n": 30 }, "mean_queue_wait_min": { "mean": 0.021421, "ci95_low": 0.011985, "ci95_high": 0.030857, "std": 0.02527, "n": 30 }, "mean_traversal_time_min": { "mean": 1.999103, "ci95_low": 1.992233, "ci95_high": 2.005973, "std": 0.018399, "n": 30 }, "traversal_count": { "mean": 90.866667, "ci95_low": 90.071624, "ci95_high": 91.661709, "std": 2.129163, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.937067, "mean_queue_wait_min": 14.234655, "composite_score": 13.338819 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.844934, "mean_queue_wait_min": 3.711354, "composite_score": 3.135851 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.640663, "mean_queue_wait_min": 3.030277, "composite_score": 1.941386 }, { "resource_id": "E03_UP", "resource_kind": "edge", "utilisation_mean": 0.078432, "mean_queue_wait_min": 16.896822, "composite_score": 1.325246 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.445851, "mean_queue_wait_min": 0.164807, "composite_score": 0.073479 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "ramp_upgrade": { "scenario_id": "ramp_upgrade", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 12606.666667, "ci95_low": 12545.479796, "ci95_high": 12667.853538, "std": 163.86145, "n": 30 }, "tonnes_per_hour": { "mean": 1575.833333, "ci95_low": 1568.184974, "ci95_high": 1583.481692, "std": 20.482681, "n": 30 }, "average_truck_cycle_time_min": { "mean": 29.546462, "ci95_low": 29.404954, "ci95_high": 29.68797, "std": 0.378965, "n": 30 }, "average_truck_utilisation": { "mean": 0.994723, "ci95_low": 0.994066, "ci95_high": 0.99538, "std": 0.001759, "n": 30 }, "crusher_utilisation": { "mean": 0.915834, "ci95_low": 0.911683, "ci95_high": 0.919985, "std": 0.011116, "n": 30 }, "average_loader_queue_time_min": { "mean": 2.721649, "ci95_low": 2.631236, "ci95_high": 2.812062, "std": 0.24213, "n": 30 }, "average_crusher_queue_time_min": { "mean": 3.297463, "ci95_low": 3.138513, "ci95_high": 3.456414, "std": 0.425677, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.603468, "ci95_low": 0.597362, "ci95_high": 0.609574, "std": 0.016353, "n": 30 }, "mean_queue_wait_min": { "mean": 2.54961, "ci95_low": 2.428445, "ci95_high": 2.670775, "std": 0.324485, "n": 30 }, "services_completed": { "mean": 44.466667, "ci95_low": 43.969221, "ci95_high": 44.964112, "std": 1.332183, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.807299, "ci95_low": 0.799527, "ci95_high": 0.815071, "std": 0.020813, "n": 30 }, "mean_queue_wait_min": { "mean": 2.810961, "ci95_low": 2.703243, "ci95_high": 2.918679, "std": 0.288474, "n": 30 }, "services_completed": { "mean": 85.766667, "ci95_low": 85.189159, "ci95_high": 86.344174, "std": 1.546594, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.915834, "ci95_low": 0.911683, "ci95_high": 0.919985, "std": 0.011116, "n": 30 }, "mean_queue_wait_min": { "mean": 3.297463, "ci95_low": 3.138513, "ci95_high": 3.456414, "std": 0.425677, "n": 30 }, "services_completed": { "mean": 126.066667, "ci95_low": 125.454798, "ci95_high": 126.678535, "std": 1.638614, "n": 30 } }, "edges": { "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.352854, "ci95_low": 0.351201, "ci95_high": 0.354506, "std": 0.004425, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000788, "ci95_low": 0.000149, "ci95_high": 0.001426, "std": 0.001711, "n": 30 }, "mean_traversal_time_min": { "mean": 1.34925, "ci95_low": 1.344887, "ci95_high": 1.353613, "std": 0.011683, "n": 30 }, "traversal_count": { "mean": 125.533333, "ci95_low": 124.947149, "ci95_high": 126.119518, "std": 1.569831, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.421397, "ci95_low": 0.41897, "ci95_high": 0.423824, "std": 0.0065, "n": 30 }, "mean_queue_wait_min": { "mean": 0.149839, "ci95_low": 0.13845, "ci95_high": 0.161227, "std": 0.030499, "n": 30 }, "mean_traversal_time_min": { "mean": 1.583956, "ci95_low": 1.578668, "ci95_high": 1.589244, "std": 0.014161, "n": 30 }, "traversal_count": { "mean": 127.7, "ci95_low": 127.094318, "ci95_high": 128.305682, "std": 1.622046, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.218484, "ci95_low": 0.215915, "ci95_high": 0.221053, "std": 0.006879, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.379869, "ci95_low": 2.366139, "ci95_high": 2.393598, "std": 0.036769, "n": 30 }, "traversal_count": { "mean": 44.066667, "ci95_low": 43.607416, "ci95_high": 44.525917, "std": 1.229896, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.19136, "ci95_low": 0.189377, "ci95_high": 0.193342, "std": 0.005309, "n": 30 }, "mean_queue_wait_min": { "mean": 0.015895, "ci95_low": 0.010474, "ci95_high": 0.021317, "std": 0.014519, "n": 30 }, "mean_traversal_time_min": { "mean": 2.03227, "ci95_low": 2.023017, "ci95_high": 2.041522, "std": 0.024779, "n": 30 }, "traversal_count": { "mean": 45.2, "ci95_low": 44.735889, "ci95_high": 45.664111, "std": 1.242911, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.419517, "ci95_low": 0.416033, "ci95_high": 0.423, "std": 0.009329, "n": 30 }, "mean_queue_wait_min": { "mean": 0.004167, "ci95_low": 0.001867, "ci95_high": 0.006467, "std": 0.006159, "n": 30 }, "mean_traversal_time_min": { "mean": 2.356959, "ci95_low": 2.347856, "ci95_high": 2.366061, "std": 0.024377, "n": 30 }, "traversal_count": { "mean": 85.433333, "ci95_low": 84.847559, "ci95_high": 86.019107, "std": 1.568732, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.364576, "ci95_low": 0.362019, "ci95_high": 0.367132, "std": 0.006848, "n": 30 }, "mean_queue_wait_min": { "mean": 0.61218, "ci95_low": 0.601274, "ci95_high": 0.623085, "std": 0.029206, "n": 30 }, "mean_traversal_time_min": { "mean": 2.006166, "ci95_low": 2.000152, "ci95_high": 2.01218, "std": 0.016106, "n": 30 }, "traversal_count": { "mean": 87.233333, "ci95_low": 86.592675, "ci95_high": 87.873992, "std": 1.715715, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.915834, "mean_queue_wait_min": 3.297463, "composite_score": 3.019929 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.807299, "mean_queue_wait_min": 2.810961, "composite_score": 2.269286 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.603468, "mean_queue_wait_min": 2.54961, "composite_score": 1.538608 }, { "resource_id": "E09_TO_LOAD_S", "resource_kind": "edge", "utilisation_mean": 0.364576, "mean_queue_wait_min": 0.61218, "composite_score": 0.223186 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.421397, "mean_queue_wait_min": 0.149839, "composite_score": 0.063142 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "crusher_slowdown": { "scenario_id": "crusher_slowdown", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 6513.333333, "ci95_low": 6456.37896, "ci95_high": 6570.287707, "std": 152.526613, "n": 30 }, "tonnes_per_hour": { "mean": 814.166667, "ci95_low": 807.04737, "ci95_high": 821.285963, "std": 19.065827, "n": 30 }, "average_truck_cycle_time_min": { "mean": 55.488254, "ci95_low": 55.020586, "ci95_high": 55.955922, "std": 1.252437, "n": 30 }, "average_truck_utilisation": { "mean": 0.973473, "ci95_low": 0.971487, "ci95_high": 0.97546, "std": 0.005321, "n": 30 }, "crusher_utilisation": { "mean": 0.947676, "ci95_low": 0.945675, "ci95_high": 0.949677, "std": 0.005359, "n": 30 }, "average_loader_queue_time_min": { "mean": 0.641623, "ci95_low": 0.569792, "ci95_high": 0.713455, "std": 0.192368, "n": 30 }, "average_crusher_queue_time_min": { "mean": 26.566543, "ci95_low": 26.195048, "ci95_high": 26.938037, "std": 0.994881, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.329364, "ci95_low": 0.320814, "ci95_high": 0.337914, "std": 0.022898, "n": 30 }, "mean_queue_wait_min": { "mean": 0.080543, "ci95_low": 0.043791, "ci95_high": 0.117295, "std": 0.098424, "n": 30 }, "services_completed": { "mean": 24.3, "ci95_low": 23.735402, "ci95_high": 24.864598, "std": 1.512021, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.445261, "ci95_low": 0.435628, "ci95_high": 0.454895, "std": 0.0258, "n": 30 }, "mean_queue_wait_min": { "mean": 0.931278, "ci95_low": 0.822841, "ci95_high": 1.039716, "std": 0.290401, "n": 30 }, "services_completed": { "mean": 47.233333, "ci95_low": 46.408899, "ci95_high": 48.057767, "std": 2.207875, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.947676, "ci95_low": 0.945675, "ci95_high": 0.949677, "std": 0.005359, "n": 30 }, "mean_queue_wait_min": { "mean": 26.566543, "ci95_low": 26.195048, "ci95_high": 26.938037, "std": 0.994881, "n": 30 }, "services_completed": { "mean": 65.133333, "ci95_low": 64.56379, "ci95_high": 65.702877, "std": 1.525266, "n": 30 } }, "edges": { "E03_DOWN": { "edge_id": "E03_DOWN", "utilisation": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "traversal_count": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 } }, "E03_UP": { "edge_id": "E03_UP", "utilisation": { "mean": 0.052746, "ci95_low": 0.052132, "ci95_high": 0.053359, "std": 0.001643, "n": 30 }, "mean_queue_wait_min": { "mean": 10.890609, "ci95_low": 10.727612, "ci95_high": 11.053605, "std": 0.436512, "n": 30 }, "mean_traversal_time_min": { "mean": 3.164733, "ci95_low": 3.127922, "ci95_high": 3.201544, "std": 0.098581, "n": 30 }, "traversal_count": { "mean": 8.0, "ci95_low": 8.0, "ci95_high": 8.0, "std": 0.0, "n": 30 } }, "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.18296, "ci95_low": 0.181387, "ci95_high": 0.184534, "std": 0.004214, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 1.351832, "ci95_low": 1.347235, "ci95_high": 1.356429, "std": 0.012311, "n": 30 }, "traversal_count": { "mean": 64.966667, "ci95_low": 64.416445, "ci95_high": 65.516888, "std": 1.473521, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.231771, "ci95_low": 0.229314, "ci95_high": 0.234229, "std": 0.006581, "n": 30 }, "mean_queue_wait_min": { "mean": 0.139459, "ci95_low": 0.126914, "ci95_high": 0.152004, "std": 0.033595, "n": 30 }, "mean_traversal_time_min": { "mean": 1.583957, "ci95_low": 1.577304, "ci95_high": 1.59061, "std": 0.017817, "n": 30 }, "traversal_count": { "mean": 70.233333, "ci95_low": 69.585213, "ci95_high": 70.881453, "std": 1.735697, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.119784, "ci95_low": 0.116928, "ci95_high": 0.122639, "std": 0.007647, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.382314, "ci95_low": 2.366207, "ci95_high": 2.398421, "std": 0.043135, "n": 30 }, "traversal_count": { "mean": 24.133333, "ci95_low": 23.589705, "ci95_high": 24.676962, "std": 1.455864, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.102933, "ci95_low": 0.100538, "ci95_high": 0.105328, "std": 0.006414, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.013834, "ci95_low": 1.999504, "ci95_high": 2.028163, "std": 0.038375, "n": 30 }, "traversal_count": { "mean": 24.533333, "ci95_low": 23.998622, "ci95_high": 25.068044, "std": 1.431983, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.230057, "ci95_low": 0.225674, "ci95_high": 0.23444, "std": 0.011738, "n": 30 }, "mean_queue_wait_min": { "mean": 0.00076, "ci95_low": -0.000392, "ci95_high": 0.001913, "std": 0.003087, "n": 30 }, "mean_traversal_time_min": { "mean": 2.347582, "ci95_low": 2.332881, "ci95_high": 2.362283, "std": 0.03937, "n": 30 }, "traversal_count": { "mean": 47.033333, "ci95_low": 46.227777, "ci95_high": 47.838889, "std": 2.157318, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.1979, "ci95_low": 0.194452, "ci95_high": 0.201349, "std": 0.009236, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000145, "ci95_low": -7.3e-05, "ci95_high": 0.000362, "std": 0.000583, "n": 30 }, "mean_traversal_time_min": { "mean": 1.992913, "ci95_low": 1.983173, "ci95_high": 2.002653, "std": 0.026085, "n": 30 }, "traversal_count": { "mean": 47.666667, "ci95_low": 46.856053, "ci95_high": 48.47728, "std": 2.170862, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.947676, "mean_queue_wait_min": 26.566543, "composite_score": 25.176473 }, { "resource_id": "E03_UP", "resource_kind": "edge", "utilisation_mean": 0.052746, "mean_queue_wait_min": 10.890609, "composite_score": 0.574431 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.445261, "mean_queue_wait_min": 0.931278, "composite_score": 0.414662 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.231771, "mean_queue_wait_min": 0.139459, "composite_score": 0.032323 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.329364, "mean_queue_wait_min": 0.080543, "composite_score": 0.026528 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "ramp_closed": { "scenario_id": "ramp_closed", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 12363.333333, "ci95_low": 12297.930568, "ci95_high": 12428.736099, "std": 175.151822, "n": 30 }, "tonnes_per_hour": { "mean": 1545.416667, "ci95_low": 1537.241321, "ci95_high": 1553.592012, "std": 21.893978, "n": 30 }, "average_truck_cycle_time_min": { "mean": 30.108704, "ci95_low": 29.970093, "ci95_high": 30.247316, "std": 0.371209, "n": 30 }, "average_truck_utilisation": { "mean": 0.994325, "ci95_low": 0.993811, "ci95_high": 0.994839, "std": 0.001376, "n": 30 }, "crusher_utilisation": { "mean": 0.898253, "ci95_low": 0.893775, "ci95_high": 0.90273, "std": 0.01199, "n": 30 }, "average_loader_queue_time_min": { "mean": 3.176232, "ci95_low": 3.069593, "ci95_high": 3.28287, "std": 0.285583, "n": 30 }, "average_crusher_queue_time_min": { "mean": 3.207245, "ci95_low": 3.060322, "ci95_high": 3.354169, "std": 0.393468, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.658173, "ci95_low": 0.653388, "ci95_high": 0.662959, "std": 0.012816, "n": 30 }, "mean_queue_wait_min": { "mean": 4.879233, "ci95_low": 4.698972, "ci95_high": 5.059495, "std": 0.482749, "n": 30 }, "services_completed": { "mean": 48.633333, "ci95_low": 48.189523, "ci95_high": 49.077144, "std": 1.188547, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.744467, "ci95_low": 0.734805, "ci95_high": 0.754129, "std": 0.025875, "n": 30 }, "mean_queue_wait_min": { "mean": 2.125733, "ci95_low": 1.987419, "ci95_high": 2.264046, "std": 0.37041, "n": 30 }, "services_completed": { "mean": 78.833333, "ci95_low": 78.286618, "ci95_high": 79.380049, "std": 1.464131, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.898253, "ci95_low": 0.893775, "ci95_high": 0.90273, "std": 0.01199, "n": 30 }, "mean_queue_wait_min": { "mean": 3.207245, "ci95_low": 3.060322, "ci95_high": 3.354169, "std": 0.393468, "n": 30 }, "services_completed": { "mean": 123.633333, "ci95_low": 122.979306, "ci95_high": 124.287361, "std": 1.751518, "n": 30 } }, "edges": { "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.346279, "ci95_low": 0.344371, "ci95_high": 0.348187, "std": 0.00511, "n": 30 }, "mean_queue_wait_min": { "mean": 0.001119, "ci95_low": 0.000253, "ci95_high": 0.001985, "std": 0.002319, "n": 30 }, "mean_traversal_time_min": { "mean": 1.348433, "ci95_low": 1.344613, "ci95_high": 1.352254, "std": 0.010232, "n": 30 }, "traversal_count": { "mean": 123.266667, "ci95_low": 122.624135, "ci95_high": 123.909199, "std": 1.720732, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.415261, "ci95_low": 0.413028, "ci95_high": 0.417494, "std": 0.00598, "n": 30 }, "mean_queue_wait_min": { "mean": 0.161427, "ci95_low": 0.150463, "ci95_high": 0.172391, "std": 0.029363, "n": 30 }, "mean_traversal_time_min": { "mean": 1.588258, "ci95_low": 1.583576, "ci95_high": 1.592939, "std": 0.012538, "n": 30 }, "traversal_count": { "mean": 125.5, "ci95_low": 124.913679, "ci95_high": 126.086321, "std": 1.570197, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.240989, "ci95_low": 0.238625, "ci95_high": 0.243353, "std": 0.00633, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.388488, "ci95_low": 2.375553, "ci95_high": 2.401423, "std": 0.034641, "n": 30 }, "traversal_count": { "mean": 48.433333, "ci95_low": 47.987361, "ci95_high": 48.879305, "std": 1.194335, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.210064, "ci95_low": 0.207725, "ci95_high": 0.212404, "std": 0.006265, "n": 30 }, "mean_queue_wait_min": { "mean": 1.073342, "ci95_low": 1.055472, "ci95_high": 1.091213, "std": 0.047857, "n": 30 }, "mean_traversal_time_min": { "mean": 2.024729, "ci95_low": 2.015057, "ci95_high": 2.034401, "std": 0.025903, "n": 30 }, "traversal_count": { "mean": 49.8, "ci95_low": 49.296152, "ci95_high": 50.303848, "std": 1.349329, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.385338, "ci95_low": 0.381949, "ci95_high": 0.388727, "std": 0.009076, "n": 30 }, "mean_queue_wait_min": { "mean": 0.004377, "ci95_low": 0.002427, "ci95_high": 0.006328, "std": 0.005223, "n": 30 }, "mean_traversal_time_min": { "mean": 2.357145, "ci95_low": 2.348653, "ci95_high": 2.365637, "std": 0.022741, "n": 30 }, "traversal_count": { "mean": 78.466667, "ci95_low": 77.872337, "ci95_high": 79.060997, "std": 1.591645, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.334398, "ci95_low": 0.33198, "ci95_high": 0.336816, "std": 0.006476, "n": 30 }, "mean_queue_wait_min": { "mean": 0.014559, "ci95_low": 0.010692, "ci95_high": 0.018427, "std": 0.010358, "n": 30 }, "mean_traversal_time_min": { "mean": 2.007319, "ci95_low": 1.997806, "ci95_high": 2.016832, "std": 0.025476, "n": 30 }, "traversal_count": { "mean": 79.966667, "ci95_low": 79.443316, "ci95_high": 80.490017, "std": 1.401559, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.658173, "mean_queue_wait_min": 4.879233, "composite_score": 3.21138 }, { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.898253, "mean_queue_wait_min": 3.207245, "composite_score": 2.880916 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.744467, "mean_queue_wait_min": 2.125733, "composite_score": 1.582538 }, { "resource_id": "E07_TO_LOAD_N", "resource_kind": "edge", "utilisation_mean": 0.210064, "mean_queue_wait_min": 1.073342, "composite_score": 0.225471 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.415261, "mean_queue_wait_min": 0.161427, "composite_score": 0.067034 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }, "trucks_12_ramp_upgrade": { "scenario_id": "trucks_12_ramp_upgrade", "replications": 30, "shift_length_hours": 8.0, "total_tonnes_delivered": { "mean": 12953.333333, "ci95_low": 12869.702929, "ci95_high": 13036.963738, "std": 223.966336, "n": 30 }, "tonnes_per_hour": { "mean": 1619.166667, "ci95_low": 1608.712866, "ci95_high": 1629.620467, "std": 27.995792, "n": 30 }, "average_truck_cycle_time_min": { "mean": 42.541378, "ci95_low": 42.273665, "ci95_high": 42.809091, "std": 0.716949, "n": 30 }, "average_truck_utilisation": { "mean": 0.988183, "ci95_low": 0.987532, "ci95_high": 0.988833, "std": 0.001742, "n": 30 }, "crusher_utilisation": { "mean": 0.940564, "ci95_low": 0.937666, "ci95_high": 0.943462, "std": 0.007761, "n": 30 }, "average_loader_queue_time_min": { "mean": 3.963752, "ci95_low": 3.805748, "ci95_high": 4.121755, "std": 0.423141, "n": 30 }, "average_crusher_queue_time_min": { "mean": 14.296978, "ci95_low": 13.965957, "ci95_high": 14.628, "std": 0.886491, "n": 30 }, "loaders": { "L_N": { "loader_id": "L_N", "utilisation": { "mean": 0.640686, "ci95_low": 0.633425, "ci95_high": 0.647947, "std": 0.019445, "n": 30 }, "mean_queue_wait_min": { "mean": 3.125669, "ci95_low": 2.952307, "ci95_high": 3.299031, "std": 0.464272, "n": 30 }, "services_completed": { "mean": 47.2, "ci95_low": 46.677415, "ci95_high": 47.722585, "std": 1.399507, "n": 30 } }, "L_S": { "loader_id": "L_S", "utilisation": { "mean": 0.849761, "ci95_low": 0.838617, "ci95_high": 0.860905, "std": 0.029844, "n": 30 }, "mean_queue_wait_min": { "mean": 4.401933, "ci95_low": 4.199838, "ci95_high": 4.604028, "std": 0.541219, "n": 30 }, "services_completed": { "mean": 90.366667, "ci95_low": 89.53758, "ci95_high": 91.195753, "std": 2.220334, "n": 30 } } }, "crusher": { "dump_id": "D_CRUSH", "utilisation": { "mean": 0.940564, "ci95_low": 0.937666, "ci95_high": 0.943462, "std": 0.007761, "n": 30 }, "mean_queue_wait_min": { "mean": 14.296978, "ci95_low": 13.965957, "ci95_high": 14.628, "std": 0.886491, "n": 30 }, "services_completed": { "mean": 129.533333, "ci95_low": 128.697029, "ci95_high": 130.369637, "std": 2.239663, "n": 30 } }, "edges": { "E05_FROM_CRUSH": { "edge_id": "E05_FROM_CRUSH", "utilisation": { "mean": 0.362475, "ci95_low": 0.360074, "ci95_high": 0.364876, "std": 0.006429, "n": 30 }, "mean_queue_wait_min": { "mean": 0.000809, "ci95_low": 0.000126, "ci95_high": 0.001492, "std": 0.001829, "n": 30 }, "mean_traversal_time_min": { "mean": 1.347725, "ci95_low": 1.343515, "ci95_high": 1.351935, "std": 0.011276, "n": 30 }, "traversal_count": { "mean": 129.1, "ci95_low": 128.289288, "ci95_high": 129.910712, "std": 2.171127, "n": 30 } }, "E05_TO_CRUSH": { "edge_id": "E05_TO_CRUSH", "utilisation": { "mean": 0.446615, "ci95_low": 0.443862, "ci95_high": 0.449368, "std": 0.007372, "n": 30 }, "mean_queue_wait_min": { "mean": 0.162179, "ci95_low": 0.153482, "ci95_high": 0.170877, "std": 0.023293, "n": 30 }, "mean_traversal_time_min": { "mean": 1.586467, "ci95_low": 1.581068, "ci95_high": 1.591867, "std": 0.01446, "n": 30 }, "traversal_count": { "mean": 135.133333, "ci95_low": 134.2913, "ci95_high": 135.975367, "std": 2.255007, "n": 30 } }, "E07_FROM_LOAD_N": { "edge_id": "E07_FROM_LOAD_N", "utilisation": { "mean": 0.232757, "ci95_low": 0.229915, "ci95_high": 0.235599, "std": 0.007611, "n": 30 }, "mean_queue_wait_min": { "mean": 0.0, "ci95_low": 0.0, "ci95_high": 0.0, "std": 0.0, "n": 30 }, "mean_traversal_time_min": { "mean": 2.383735, "ci95_low": 2.367863, "ci95_high": 2.399606, "std": 0.042505, "n": 30 }, "traversal_count": { "mean": 46.866667, "ci95_low": 46.420156, "ci95_high": 47.313178, "std": 1.195778, "n": 30 } }, "E07_TO_LOAD_N": { "edge_id": "E07_TO_LOAD_N", "utilisation": { "mean": 0.201252, "ci95_low": 0.198576, "ci95_high": 0.203928, "std": 0.007166, "n": 30 }, "mean_queue_wait_min": { "mean": 0.013991, "ci95_low": 0.009343, "ci95_high": 0.018639, "std": 0.012447, "n": 30 }, "mean_traversal_time_min": { "mean": 2.017927, "ci95_low": 2.008034, "ci95_high": 2.027821, "std": 0.026495, "n": 30 }, "traversal_count": { "mean": 47.866667, "ci95_low": 47.331956, "ci95_high": 48.401378, "std": 1.431983, "n": 30 } }, "E09_FROM_LOAD_S": { "edge_id": "E09_FROM_LOAD_S", "utilisation": { "mean": 0.440749, "ci95_low": 0.436165, "ci95_high": 0.445333, "std": 0.012276, "n": 30 }, "mean_queue_wait_min": { "mean": 0.004975, "ci95_low": 0.002417, "ci95_high": 0.007532, "std": 0.00685, "n": 30 }, "mean_traversal_time_min": { "mean": 2.353135, "ci95_low": 2.344126, "ci95_high": 2.362144, "std": 0.024127, "n": 30 }, "traversal_count": { "mean": 89.9, "ci95_low": 89.113368, "ci95_high": 90.686632, "std": 2.10664, "n": 30 } }, "E09_TO_LOAD_S": { "edge_id": "E09_TO_LOAD_S", "utilisation": { "mean": 0.381888, "ci95_low": 0.377851, "ci95_high": 0.385925, "std": 0.010811, "n": 30 }, "mean_queue_wait_min": { "mean": 1.381388, "ci95_low": 1.362496, "ci95_high": 1.400279, "std": 0.050592, "n": 30 }, "mean_traversal_time_min": { "mean": 1.998937, "ci95_low": 1.991415, "ci95_high": 2.006459, "std": 0.020143, "n": 30 }, "traversal_count": { "mean": 91.7, "ci95_low": 90.822092, "ci95_high": 92.577908, "std": 2.351082, "n": 30 } } }, "top_bottlenecks": [ { "resource_id": "D_CRUSH", "resource_kind": "crusher", "utilisation_mean": 0.940564, "mean_queue_wait_min": 14.296978, "composite_score": 13.447226 }, { "resource_id": "L_S", "resource_kind": "loader", "utilisation_mean": 0.849761, "mean_queue_wait_min": 4.401933, "composite_score": 3.740592 }, { "resource_id": "L_N", "resource_kind": "loader", "utilisation_mean": 0.640686, "mean_queue_wait_min": 3.125669, "composite_score": 2.002573 }, { "resource_id": "E09_TO_LOAD_S", "resource_kind": "edge", "utilisation_mean": 0.381888, "mean_queue_wait_min": 1.381388, "composite_score": 0.527536 }, { "resource_id": "E05_TO_CRUSH", "resource_kind": "edge", "utilisation_mean": 0.446615, "mean_queue_wait_min": 0.162179, "composite_score": 0.072432 } ], "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] } }, "key_assumptions": [ "Hard cut at t=480 minutes: only end_dump events with time_min < 480 contribute tonnes; in-flight loads/dumps at the cut are discarded (operator-facing 'tonnes closed at shift end').", "Static shortest-time routing per scenario via Dijkstra on free-flow edge times (distance / max_speed_kph), recomputed only when a scenario closes or upgrades edges.", "Capacity-1 directed edges modelled as independent SimPy Resources, mirroring edges.csv literally (E03_UP and E03_DOWN are decoupled even on the shared physical ramp).", "Travel-time noise: per-edge-traversal lognormal multiplier with mean 1 and cv=0.10; loaded trucks use loaded_speed_factor, empty trucks use empty_speed_factor.", "Loading and dumping sampled as normal_truncated with per-resource (mean, sd), floored at max(0.1, sample) so durations stay strictly positive without biasing the mean.", "Dispatch policy: each empty truck is assigned to argmin(travel_to_loader + current_queue_len * mean_load_time + own_load_time); current_queue_len includes the truck currently being served. Ties broken by lower loader_id.", "Initial dispatch: all trucks released simultaneously at t=0 from PARK; no warmup, no staged ramp-up, no shift-handover modelling.", "Crusher tonnes are credited at end_dump (not start_dump or arrive_crusher); each completed dump credits exactly payload_tonnes (100 t) per truck.", "WASTE and MAINT are out of scope: their edges remain in the graph but are never traversed.", "Per-replication seed = base_random_seed + replication_index, so each replication is independently reproducible while the full run is deterministic.", "95% confidence intervals computed via Student-t with n-1=29 degrees of freedom over 30 replications." ], "model_limitations": [ "No truck breakdowns, refuelling, or maintenance windows: truck availability is treated as 1.00 across the entire 480-minute shift.", "No operator-level decisions: shift handover, lunch breaks, and manual dispatcher overrides are not represented.", "Static routing: trucks commit to their shortest-time path at dispatch and do not re-plan if a capacity-1 edge develops a long queue. A real dispatcher might divert through the bypass.", "Independent edge directions: E03_UP and E03_DOWN are two separate single-lane SimPy Resources. If the physical ramp is genuinely one shared lane, real congestion will be worse than modelled.", "Crusher always available: D_CRUSH never blocks. There is no downstream stockpile back-pressure, no full-bin signal, and no scheduled crusher maintenance.", "Single homogeneous payload: every dump is exactly 100 t. Heterogeneous payloads, ore blending, and grade-dependent processing time are not modelled.", "Free-flow edges have effectively infinite capacity (capacity = 999): trucks do not interact on multi-lane road segments. Real headway and following-distance effects are ignored.", "No warmup trimming: the shift starts empty (all trucks at PARK, all queues empty). The empty-system bias is small because trucks reach steady-state within a few cycles, but is not zero.", "Edge max-speeds and per-truck speed factors are deterministic. Only the per-edge lognormal multiplier (cv=0.10) introduces travel variability \u2014 weather, dust, visibility are not modelled.", "Dispatch decisions use queue_length at the moment of decision; in-flight trucks do not feed back into subsequent dispatch decisions until they physically arrive.", "Node coordinates from nodes.csv are used for visualisation only; haul-road bends and grade are not propagated into travel time." ], "additional_scenarios_proposed": [ "trucks_12_ramp_upgrade (included as the 7th scenario): combines fleet expansion (12 trucks) with the ramp upgrade. Tests whether the two investments are complementary, substitutive, or independent. Empirically super-additive \u2014 throughput rises from ~1568 tph baseline to ~1619 tph, materially above either intervention alone.", "Crusher reliability scenario: inject random short outages on D_CRUSH (e.g. 5 min every 60 min) to size the surge-pile buffer requirement and quantify the operational cost of unplanned downtime.", "Operator break schedule: stagger loader downtime around lunch/break windows to estimate the throughput penalty of synchronised vs staggered crew breaks.", "Dynamic re-routing: re-plan when a queue at a capacity-1 edge exceeds a threshold (e.g. 2 waiting trucks). Quantifies the upper bound on lift available from a smarter dispatcher vs the static baseline.", "Heterogeneous fleet mix: replace a subset of 100 t trucks with 150 t trucks to trade more cycles for higher per-cycle payload.", "Crusher service-time upgrade: reduce mean dump time from 3.5 min to 2.5 min and re-run trucks_12 to test whether crusher service time becomes the binding constraint at higher fleet sizes.", "Mid-shift loader outage: take L_N or L_S offline for 30 minutes mid-shift to size the single-loader fall-back tonnes and inform redundancy planning." ] }