Moisture-Scheduled Drip-Irrigation Controller with Flow-Totalised Water Accounting

Moisture-Scheduled Drip-Irrigation Controller with Flow-Totalised Water Accounting answers the question every irrigation project skips: did the automation actually save water, and how much? Our rig runs a three-week paired trial on two identical tomato beds on the college farm. One bed is irrigated by our Arduino Uno controller — two capacitive soil-moisture sensors at root depth trigger a 12 V solenoid valve whenever the mean reading crosses a threshold set from field capacity minus a 30 per cent management-allowed depletion, and the valve closes on a target volume. The second follows a fixed daily timer mimicking the farmer's habit of watering by the clock.

Every litre through either bed passes a YF-S201 hall-effect flow meter whose pulses the controller counts into per-bed totals, so the comparison ends with numbers, not impressions: litres per bed, litres per event, events per week, and the water saved. A rain sensor skips waterings, an SD card records every event, and a Python pipeline turns the logs into water-accounting tables and plots.

The logic is deliberately conservative — minimum interval, maximum runtime guard, valve verification through the flow meter — because a stuck valve floods a bed or starves it. Sensors are calibrated gravimetrically against weighed soil samples — the step most hobby builds omit, and the one that makes our threshold meaningful. The build costs about ₹6,800 and follows the VTU toolset: C firmware, Python analysis and a lab-to-field workflow our agri-eng batch can defend end to end.

Drip irrigation saves water compared with flood irrigation, but on most Indian smallholdings the schedule still runs by habit: the valve or pump opens at a fixed time each day regardless of soil condition, season or crop stage. Over-watering wastes water and electricity, leaches nutrients and encourages root disease; under-watering during heat waves stresses the crop. Sensor-based commercial controllers solve this but are priced beyond small growers, and almost none meter the water actually delivered, so nobody can say whether the automation paid for itself.

There is a need for a low-cost, calibrated, moisture-scheduled drip controller with flow-totalised water accounting — one that irrigates from soil-moisture thresholds set by agronomic logic, meters every litre delivered to each bed, logs events transparently, and demonstrates the saving against a fixed-schedule baseline in a controlled paired-bed trial. The system must also be safe to leave unattended: valve-fault detection, runtime guards and rain-skip logic, because automation that fails silently is worse than the habit it replaces.

Abstract

Drip systems are promoted across India under PMKSY's Per Drop More Crop, yet most installed systems are still scheduled by habit — a fixed hour each morning — while the pump runs regardless of what the soil needs. This project builds and tests an Arduino-based drip-irrigation controller that schedules irrigation from soil-moisture thresholds, meters every litre, and quantifies the water saved against a fixed-schedule baseline in a paired-bed trial on tomato.

Introduction

Karnataka's horticulture runs heavily on drip, and KVK extension advice already says what our controller repeats: irrigate at depletion, not by the clock. The advice needs a cheap instrument to execute it and a measurement chain to prove it worked. We built both on two raised beds with inline drip laterals, so the whole argument fits on a college farm plot and in a viva.

Existing system and gap

• Fixed-schedule operation: the commonest real-world practice — same duration daily, adjusted by feel — over-waters in cool weeks and under-waters in heat.
• Timer-based upgrades: add convenience but still know nothing about the soil.
• Commercial sensor-based controllers: exist but are priced beyond small growers, and rarely meter water flow for accounting.
• Published student builds: usually demonstrate one bed and one sensor, without calibration, comparison or water accounting.
• Gap: a low-cost controller that schedules from calibrated soil-moisture sensing, meters flow, and proves savings in a paired-bed trial.

Proposed system

• Sensing: two capacitive soil-moisture sensors per automated bed, calibrated gravimetrically so thresholds are volumetric, not raw counts; a rain sensor to skip events.
• Scheduling: irrigate when the mean reading falls below field capacity minus a 30 per cent management-allowed depletion; stop at a computed target volume; enforce minimum interval and runtime guard.
• Actuation: 12 V solenoid valves driven through a relay board with flyback protection; a small pump from an overhead tank supplies both beds identically.
• Accounting: a YF-S201 flow meter per bed totalises litres; every event is logged to SD and serial with timestamp, litres and sensor readings.
• Trial: three weeks, two beds — controller-scheduled versus fixed — with identical plants, medium and fertigation, comparing water use and simple plant observations.

Feasibility

• Technical: capacitive sensors, flow meters, relays and Arduino are all standard modules; the scheduling logic is a small state machine; gravimetric calibration uses an oven and balance the college already owns.
• Economic: indicative cost ≈ ₹6,800 (Uno ₹600, four capacitive sensors ₹800, two YF-S201 meters ₹700, two solenoid valves ₹900, relay board ₹200, rain sensor ₹150, SD module ₹120, drip kit and laterals ₹900, raised-bed materials ₹1,400, tank, pump and plumbing ₹1,000). The trial reuses the department's existing plot.
• Operational: the rig runs unattended; the team visits twice daily for readings and photographs; a week of data costs nothing to collect.
• Schedule: bench development and calibration in weeks 1–8 under the VTU Phase-I umbrella, field installation in week 9, three trial weeks, analysis and documentation in weeks 13–14.

In scope

• Two 1.5 m × 1 m raised beds, twelve tomato plants, three trial weeks on the college farm.
• Two solenoid valves, four capacitive sensors, two flow meters, one rain sensor, one controller, SD logging.
• Gravimetric sensor calibration and flow-meter verification against a measuring cylinder.
• Water-accounting analysis: litres per bed, per event, per week; savings percentage; event statistics.
• Simple plant observations (height, flowering, fruit set) recorded weekly as indicative, not statistical, yield evidence.

Out of scope

• Yield statistics and agronomic significance testing — three weeks and twelve plants cannot support that, and the report says so.
• Fertigation automation, EC/pH sensing and pressure regulation beyond the drip kit's built-in regulator.
• Weather-station integration or ET0-based scheduling — FAO-56 methodology is used only as a cross-check reference.
• LoRa/GSM telemetry and app dashboards (future scope); the rig logs locally for viva-day reliability.

We follow a bench-then-field plan mapped to the VTU two-phase structure: Phase-I (BXX685) covers survey, design and bench calibration; Phase-II (BXX786) covers installation, the field trial and the report.

Test plan: (1) bench — scheduler logic exercised with simulated sensor voltages and a lamp in place of each valve; (2) calibration — sensor linearity against gravimetric samples, flow meter within ±5 per cent of cylinder volume; (3) field — valve-closed zero-flow check after every event, runtime guard tripping on a deliberately clamped line, rain-skip behaviour; (4) trial — daily reconciliation of logged litres against a manual tank-dip estimate. Every case is recorded with input, expected, observed and status, and each objective maps to at least one test.

The system is a three-layer loop: sensing, deciding, and acting — with accounting woven through all three.

1. Sensing layer: two capacitive moisture sensors per automated bed feed the Uno's analog inputs through RC filtering; the rain sensor gives a digital wet flag; two YF-S201 meters emit pulses on interrupt pins.
2. Decision layer: the scheduler evaluates each bed every minute: if the rain sensor is wet, skip; if the mean soil moisture is below the depletion threshold and the minimum interval has elapsed, open the valve; close at the target volume, on timeout, or if the flow meter shows zero with the valve commanded open (stuck or air-locked valve alarm).
3. Actuation and accounting layer: a four-channel relay board drives the 12 V solenoid valves with flyback diodes; every valve transition, litre total and sensor reading is written to the SD card and streamed over serial.

flowchart TD
  TANK["Overhead tank + small pump"] --> MAN["Manifold with two YF-S201 flow meters"]
  MAN -->|"Bed A line"| VALVE_A["Solenoid valve A (controller-scheduled bed)"]
  MAN -->|"Bed B line"| VALVE_B["Solenoid valve B (fixed-schedule bed)"]
  VALVE_A --> DRIPA["Drip laterals, Bed A"]
  VALVE_B --> DRIPB["Drip laterals, Bed B"]
  SM["4 x capacitive soil-moisture sensors"] -->|"analog"| UNO["Arduino Uno scheduler"]
  RAIN["Rain sensor"] --> UNO
  MAN -->|"flow pulses"| UNO
  UNO -->|"relay commands"| RELAY["4-channel relay board + flyback diodes"]
  RELAY --> VALVE_A
  RELAY --> VALVE_B
  UNO -->|"event log lines"| SD["SD card + USB serial"]
  SD --> PC["Laptop: Python water-accounting analysis"]

flowchart TD
  A["Evaluate bed every 60 s"] --> B{"Rain sensor wet?"}
  B -->|"Yes"| C["Skip scheduled irrigation today"]
  B -->|"No"| D{"Mean moisture below depletion threshold?"}
  D -->|"No"| A
  D -->|"Yes"| E{"Minimum interval elapsed?"}
  E -->|"No"| A
  E -->|"Yes"| F["Open valve, start counting litres"]
  F --> G{"Target volume reached?"}
  G -->|"Yes"| H["Close valve, log event"]
  G -->|"No"| I{"Runtime guard expired or zero-flow?"}
  I -->|"Yes"| J["Close valve, raise VALVE FAULT, log"]
  I -->|"No"| G
  H --> A
  C --> A
  J --> A

Key design numbers we defend in the viva: depletion threshold set 30 per cent below field capacity from the gravimetric curve; target volume per event computed from bed area, root depth and the moisture deficit between threshold and field capacity; minimum interval six hours; runtime guard 20 minutes; flow-meter pulse rate 450 pulses per litre nominal, verified against a measuring cylinder before deployment.

• ET-based double check: add a small weather station and compute FAO-56 reference evapotranspiration, scheduling against ET rather than moisture alone, with the moisture loop as a safety net.
• Fertigation injection through a third valve and venturi, timed by the same controller with its own accounting.
• LoRa or GSM telemetry so the farmer sees litres, soil status and faults on a phone, with the SD log kept as the offline fallback.
• Multi-valve expansion to four or eight beds with per-valve accounting, closer to a real smallholding's layout.
• Season-length agronomic trial with yield measurement and repetition, upgrading the indicative observations into statistics.
• Solar power with a small panel and charge controller, removing the mains adapter entirely for field autonomy.