Agricultural machinery: settings that do not follow the field

Drivers leave machine settings unchanged as conditions change

What happens. A combine, a drill or a baler is set up in the morning and then left alone. Crop moisture, straw, slope and ground change through the day, but the settings stay where the driver put them. Few drivers have the time or the experience to correct them every hour, so the machine runs at a compromise.

What it costs. Grain left in the straw, damaged grain, uneven work, more fuel than needed. Customers see it as lost yield and blame the machine, not the settings.

Why the usual fix fails. A table of recommended settings in the manual needs a driver who reads it at the right moment. More buttons and more sensors add to the workload instead of removing it. Presets for crop types are too coarse for a field that changes every few hundred metres.

How a fixed-price remote diagnosis checks it. From machine data logs (settings, loads, speeds, losses where measured) I look at how far the real operating point moves from a good one, and which conditions cause it. The fixed-price remote diagnosis tells you whether a computed crop state can adjust the settings automatically, and what data that would need.

Background: Learning-augmented adaptive predictive control

Soil and load change, working depth drifts

What happens. Soil hardness, moisture and slope change within one field. The implement digs deeper in soft ground and rides up in hard ground, while the tractor load jumps with it. Drivers correct by hand and by feel.

What it costs. Uneven work quality, extra fuel when the tool is too deep, wear on tools and hitch, and tired drivers at the end of the day.

Why the usual fix fails. A fixed gain on the hydraulic depth control is tuned for one soil. In another it either reacts too slowly or starts to oscillate. A force sensor helps only with the part it measures, and it does not tell the controller what the ground will do next.

How a fixed-price remote diagnosis checks it. From drive logs (depth, draft force, hydraulic valve commands, speed, slope) I look at how depth error and load relate to ground conditions. The fixed-price remote diagnosis answers whether an estimate of soil resistance can make the depth control adapt, and what it would take on your hardware.

Wider sprayer booms oscillate and height control struggles

What happens. Booms keep getting wider. A wide boom flexes and swings with the sprayer’s movements on uneven ground. The height of the nozzles above the crop then varies along the boom, and the actuators fight the swing instead of damping it.

What it costs. Uneven spray distribution, drift when the boom swings high, overdosing where it dips low, and booms that are built wider than the control can handle.

Why the usual fix fails. Stiffer mechanics add weight. A faster hydraulic loop on each section excites the boom’s own swing. Tuning on the test field works for the test field and not for the farmer’s slope.

How a fixed-price remote diagnosis checks it. From height sensor and valve logs of real drives I identify how the boom moves and where the current control loses it. The fixed-price remote diagnosis tells you whether a model-based height and damping control can keep wider booms steady, and what it needs in sensors and valve speed.

Background: Industrial process predictive control

Grain losses are not measured well enough to steer by

What happens. Loss sensors on a combine count kernels hitting a plate. They react to the crop and to the sensor’s own position, so the figure on the screen is a rough hint. Drivers look back at the field and judge by eye.

What it costs. Yield left in the field, or speed and capacity given away to be safe. Both are hard to see until the harvest is in.

Why the usual fix fails. Recalibrating the loss sensor for each crop is a manual task nobody repeats. More sensors add more disagreeing numbers. Without a reliable loss value no automatic setting can be tuned against it.

How a fixed-price remote diagnosis checks it. From machine logs together with any reference measurements (collected loss, tank samples, speeds, rotor and fan values) I look at whether a computed loss estimate from the existing sensors is closer to the reference than the sensor itself. The fixed-price remote diagnosis tells you what such a computed value would take.

Background: Computed measurements

Engine and transmission are set by the driver, not for fuel

What happens. Two drivers on the same tractor with the same job use different engine speeds and gears. Nobody tells them where the efficient point is for the load they have right now, so habit decides.

What it costs. Fuel, which is the largest running cost of the machine for the customer, and engine wear at points the driver does not notice.

Why the usual fix fails. Eco modes are fixed maps. They do not know the implement, the soil or the working speed the job needs. Training courses help for a season. Drivers paid by area do not trade speed for fuel on a guess.

How a fixed-price remote diagnosis checks it. From drive logs (engine speed and torque, gear, ground speed, fuel rate, implement) I look at how much spread there is between drivers at the same work, and where an economic operating point would lie. The fixed-price remote diagnosis tells you whether a controller that picks speed and gear for lowest fuel per hectare is realistic on your machine.

Background: Economic predictive control

A new exhaust stage needs new control

What happens. A new emission stage brings a different after-treatment system: different sensors, different heating-up behaviour, new regeneration cycles. The engine and exhaust control from the previous stage does not simply carry over.

What it costs. Development time on the test bench and in the field, a tight deadline before the machine may be sold, and the risk of derating or regeneration at the wrong moment in the harvest.

Why the usual fix fails. Calibration by many test runs and big parameter tables takes long and breaks whenever hardware changes. More staff do not help if the experienced calibrators are busy on the last stage.

How a fixed-price remote diagnosis checks it. From test logs (temperatures, dosing, pressures, loads, regeneration events) I look at where the current control leaves the limits and which part of the behaviour a model could cover. The fixed-price remote diagnosis tells you whether a model-based control would cut the calibration effort, and what it takes to try.

Background: Calibration-efficient predictive control

Implements drift on slopes and overlap grows

What happens. On a slope the tractor follows its track, but the implement slides downhill. Satellite guidance steers the tractor, not the tool, so the working line shifts and the driver corrects at the edge.

What it costs. Overlap with the previous pass, or gaps. Overlap means seed, fertiliser and spray applied twice; gaps mean uneven crop.

Why the usual fix fails. A constant offset set by the driver fits one slope. A sensor on the implement only reports the drift after it has happened. Steering harder on the tractor makes it work against the hitch.

How a fixed-price remote diagnosis checks it. From guidance logs (tractor position, heading, implement position or angle, steering commands, slope) I look at how drift builds up and whether combining the existing sensors gives an early enough estimate. The fixed-price remote diagnosis tells you whether predictive steering of tractor and implement together is a realistic next step.

Background: Computing unmeasured states for control

Seed rate does not follow the variation inside a field

What happens. Soil, moisture and earlier yield differ inside one field, but the drill sows an even rate or follows an application map made weeks before. The map and the field at the moment of sowing do not always agree.

What it costs. Seed spent where the soil cannot carry it, thin stands where it could. Either way yield and seed cost are the loss.

Why the usual fix fails. Maps need data collection and processing before the season, and they age. A rate controller that simply follows the map has no feedback on what the drill actually puts in the ground. Manual correction by the driver depends on how much he trusts the map.

How a fixed-price remote diagnosis checks it. From drill logs, application maps and, where you have them, yield or emergence data, I look at how closely the real rate follows the plan and how much field variation the plan misses. The fixed-price remote diagnosis tells you whether adaptive rate control with computed field conditions is worth building.

Fertiliser spread pattern suffers from wind and slope

What happens. A disc spreader throws granules in a fan. Wind blows the fan sideways, slope tilts it, and granule size and humidity change the throw distance. The pattern the test stand certified is not the pattern in the field.

What it costs. Uneven fertilising: too much at the overlap, too little at the edge. That costs yield and fertiliser, and it can mean nutrients lost into ditches.

Why the usual fix fails. Setting tables from the manual assume calm weather and a standard granule. Edge spreading is a coarse compromise. A fixed correction by the driver follows only one wind direction at a time.

How a fixed-price remote diagnosis checks it. From spreader logs (disc speeds, flap positions, flow, ground speed, position, slope, wind where known) and any tray-test results, I look at how the pattern shifts with conditions. The fixed-price remote diagnosis tells you whether a spread model that adjusts disc speed and flow along the pass would even out the result.

Electric tractors run out of battery before the work is done

What happens. An electric tractor has a battery that carries a limited working time. Heavy implements drain it faster than a customer expects, charging takes a break in the middle of the day, and nobody can say in the morning whether the field will be finished.

What it costs. Downtime at the wrong moment, a diesel back-up that defeats the purpose, and buyers who doubt the range.

Why the usual fix fails. Bigger batteries cost weight and money. Simple range displays assume the last hour repeats, not that the next field is heavier. Planning by the driver’s gut feeling stops working with two machines and one charger.

How a fixed-price remote diagnosis checks it. From operating logs (battery state, power per implement, speeds, field sizes, charging points) I look at how well the energy for a given job can be predicted. The fixed-price remote diagnosis tells you whether a work and charging plan, updated while the work goes on, can raise the area done per charge.

Background: Trajectory optimisation and setpoint generation

Field robots lose their way on rough ground

What happens. An autonomous field robot follows a planned track. On uneven or soft ground it slips and drifts, satellite signals drop near trees, and the wheels do not do what the steering commanded. When the position is uncertain, it stops and waits for the farmer.

What it costs. Stops during work, a farmer who has to come out at any hour, and a machine that is trusted less every time it halts.

Why the usual fix fails. Stricter stopping rules make it safer and still stop more often. A better satellite receiver does not help under trees or on slopes where the real problem is slip. Rewriting the track planner does not cure a wrong state estimate.

How a fixed-price remote diagnosis checks it. From logs of real runs (position, heading, wheel speeds, inertial sensors, steering commands, error events) I look at where the position estimate goes wrong and why the robot stops. The fixed-price remote diagnosis tells you whether a better estimate with predictive control can keep it moving, and where recovery from faults is missing.

Background: Autonomous mobile navigation

Harvest trailers wait while others stand idle

What happens. At harvest the combines unload on the move into trailers. Some trailers wait at the field, others queue at the dryer or the silo. Dispatch is done by phone and by experience, and the weather changes it every hour.

What it costs. Fuel for idling tractors, driver hours, combines that stop because no trailer arrives, and grain waiting for the dryer.

Why the usual fix fails. More trailers cover the problem with more equipment. A fixed rotation does not see that one field is further away than the next, or that a breakdown changed everything. Fleet screens show where machines are, not what the best next step is.

How a fixed-price remote diagnosis checks it. From machine positions and event logs of a past harvest, plus distances and unload times, I look at where the waiting arises. The fixed-price remote diagnosis tells you whether an optimised dispatch that updates during the day would remove a worthwhile part of it.

Background: Trajectory optimisation and setpoint generation

Irrigation runs by schedule, not by soil moisture

What happens. Irrigation starts at fixed times and runs for fixed durations. The soil does not follow the schedule: rain, wind and crop growth change how much water the plants need.

What it costs. Water and pumping energy spent when the soil is already wet, and stress for the crop when a dry spell comes before the next slot.

Why the usual fix fails. A few moisture probes show a few spots. Rain forecasts alone do not say how much the soil took up. Manual changes depend on who is on the farm that week.

How a fixed-price remote diagnosis checks it. From irrigation logs, weather data and any probe values I look at how well soil moisture can be computed between the probe spots. The fixed-price remote diagnosis tells you whether a planned irrigation, driven by a computed moisture value and pumping cost, is realistic with the sensors you already have.

Background: Computed measurements

Field tests only run in the season

What happens. The conditions a harvester or sprayer has to be tested in occur for a few weeks a year. A control fault found in September waits until next summer for its next test.

What it costs. Development time lost to waiting, rushed tests in the first days of the season, and series machines that leave the factory with settings no one has seen on the real crop.

Why the usual fix fails. More test machines in the field cost money and still wait for the same weather. Winter tests on a bench cannot reproduce the crop. A simulation that was never matched to field data is not trusted.

How a fixed-price remote diagnosis checks it. From your field logs of past seasons and the models you already have, I look at how well simulation reproduces what the machine did. The fixed-price remote diagnosis tells you whether a simulation can take over the first rounds of testing, and what data would make it credible.

Many machine variants mean endless parameter work

What happens. The same control software runs on many variants: widths, engines, tyres, implements. Each one has its own set of parameters, found by testing, and each software change reopens the table.

What it costs. Engineering time that grows with the number of variants, errors where one variant gets the parameters of another, and slow release of new variants.

Why the usual fix fails. Copying parameters from the nearest variant works until it does not. Hiring more calibrators makes the table grow faster than the knowledge. Nobody can say afterwards why a value was chosen.

How a fixed-price remote diagnosis checks it. From your parameter sets, change history and test results, I look at which values follow from physical quantities of the variant and which really need testing. The fixed-price remote diagnosis tells you whether a model can generate the parameters for a new variant, and how much testing would stay.

Background: Calibration-efficient predictive control

Breakdowns in the harvest cost days nobody has

What happens. A harvester stops in the middle of the season. A fault code shows on the display, the driver calls the workshop, and the technician arrives without knowing what happened in the minutes before.

What it costs. Harvest days lost on weather-dependent work, rushed repairs, and customers who count machine availability as the first quality of a brand.

Why the usual fix fails. Fault codes tell what tripped, not why. Service intervals ignore how hard the machine was used. Remote data transfer from every machine produces a lot of data and little advice.

How a fixed-price remote diagnosis checks it. From fault logs and the measured values around past breakdowns (temperatures, pressures, speeds, loads) I look at whether the failures show early signs and which values carry them. The fixed-price remote diagnosis tells you whether on-board diagnosis can warn the driver before the stop, and what it would need.

Slurry spreading needs the nutrient content known and certified

What happens. Before slurry is spread, its nutrient content has to be known. Near-infrared sensors on the tanker measure it on the move, and the spreading rate is then set through pump speed or a valve. Authorities in some regions accept the values only when the sensor is independently certified.

What it costs. Accuracy limited by the laboratory calibration the sensor was built on, and uncertainty about acceptance, which makes buyers hesitate.

Why the usual fix fails. A sensor calibrated at the factory meets slurry from different animals, feed and dilution. Recalibrating against laboratory samples is slow. A rate controller that trusts the sensor passes its error on to the field.

How a fixed-price remote diagnosis checks it. From sensor spectra, laboratory results and application logs I look at how the calibration holds across slurry types and where the rate control adds its own error. The fixed-price remote diagnosis tells you whether a better calibration model and a controlled link from sensor to pump would be worth building.

Background: Computed measurements