Automation principles — two-site smart farm¶
Purpose¶
State what to automate early, late, or never for a 5 ac control hub + 120 ac production operation—architecture and decisions, not a shopping list. Complements East Tennessee two-site farm business plan — smart-tech strategy and Instrumentation priority matrix.
Constraints: Travel time between sites; sensor and comms unreliability; maintenance capacity of a small team; aggressive automation as both leverage and recurring cost (money + attention).
Principles (ranked)¶
- Manual competence before remote actuation — If you cannot run the system by hand, automation hides risk.
- Observe before optimize — Measurement and logging before closed-loop control on welfare-critical paths.
- Fewer moving parts — One well-maintained water stack beats five flaky gadgets.
- Fail known — Prefer documented safe defaults over clever recovery nobody can repeat at 02:00.
- Automation budget is not only CAPEX — Subscriptions, patching, triage hours, spares—
Instrumentation ROI model.
Automate early (typical)¶
| Class | Rationale | Caveat |
|---|---|---|
| Visibility (tank, pump run, power at critical loads) | Trip ROI across COMMUTE_ONE_WAY |
False positives → triage tax |
| Structured logging to durable store | Supports SoR for instrument history | Not a substitute for books |
| Alerting with dedupe | Attention routing | Alert fatigue = disabled alerts = worse than none |
Automate late (after gates)¶
| Class | Gate examples |
|---|---|
| Cameras everywhere | Policy, privacy, storage, bandwidth |
| Remote actuation (gates, valves, feeders) | Manual SOP; redundant sensing where welfare-linked |
| Full irrigation automation | Crop commitment; walk lines first |
| Fleet of LPWAN nodes | Pilot ROI; spares process |
Automate never (or never without human in loop)¶
| Item | Why |
|---|---|
| Animal welfare decisions on one sensor | Use redundancy or schedule |
| Loading / handling for market | Safety + quality |
| Response to stock on public road | Liability |
| “Set and forget” without patch cadence | Security debt |
Distance-specific rule¶
If MANUAL_INTERVENTION_TIME > acceptable risk window for a system, downscope remote dependence until coverage improves—Weekly coverage matrix, Family labor coverage.