Table 2. The cost of delivering the ML-derived preventative maintenance model compared to other models.
| Category | Sub-category | Nominal baseline maintenance (no sensor, as needed) | Circuit (scheduled) maintenance (no sensor, preventative) | Ambulance maintenance (sensor-enabled, as-needed) | Machine learning maintenance (sensor-enabled, preventative) |
|---|---|---|---|---|---|
| Capital Exp. | Pump | $10,000 | $10,000 | $10,000 | $10,000 |
| Sensor | $0 | $0 | $360 | $360 | |
| Annual Operational Exp. | Sensor | $0 | $0 | $410 | $410 |
| Kenya-Based Pump | $120 | $230 | $300 | $300 | |
| USA-Based Admin. | $730 | $730 | $820 | $820 | |
| NPV (5% cost of money) | CapEx | $10,000 | $10,000 | $10,360 | $10,360 |
| OpEx | $6,563 | $7,413 | $11,814 | $11,814 | |
| Total | $16,563 | $17,413 | $22,174 | $22,174 | |
| Cost of Service Delivered | Uptime | 67.5% | 72.9% | 96.5% | 99.0% |
| USD per Working Year | $2,453 | $2,387 | $2,298 | $2,240 |
Note: the uptimes for the nominal baseline and circuit models were not observed in this study; uptime data for the nominal baseline and circuit models are shown for comparison from Nagel et al., 2015. Net present value calculations assume a 5% annual cost of money.