INDUSTRIES
Your industry. Problems we have already solved.
Six asset-heavy industries. In each, the problems we solve best — with the result we delivered, by engineers who have spent a decade inside plants, fleets and fields like yours.
- ~€6M p.a.margin added by orchestrating production across sites and energy markets
- −18%flush volume on lubricant changeovers, quality maintained
- Over 90%of reliquefaction-system efficiency within target bands
- +28%pump-fleet MTBF
- USD 12–16 milliona year in value from compressor early warning
01 · CHEMICALS & PROCESSING
Your margin is set by the plan, the power price and every changeover.
Energy is a process plant's biggest controllable cost. The yield of a cracker moves with every feed change, and every changeover and fouled exchanger costs product. We solve the decisions that set the margin — what to run, where, when, and how hard — with the physics and the data the plant already has.
Olefins and crackers · chlor-alkali and vinyls · specialty chemicals and polymers · specialty hydrocarbons and solvents · lubricant blending · refining and crude units
-
PLANNING & ENERGY
Electricity bought at the wrong hour.
Electrolysis runs whether power is cheap or not. The Energy Optimizer plans load against forecast spot prices.
3–5% energy cost savings from optimised scheduling.
Energy Optimizer, in depth -
SUPPLY CHAIN & PLANNING
One plan per site, none for the network.
Each plant optimises alone while prices differ by market and by country. The Production Planner orchestrates production across sites and energy markets.
~€6M p.a. margin added by orchestrating production across sites and energy markets, chlor-vinyl chain.
Solutions -
PROCESS & FURNACE ENGINEERING
Cracker yield left on the table.
Feed, severity and decoke timing are set from yesterday's runs. Real-time steam-cracker optimisation with SPYRO® runs the simulator live against plant data.
+2–3% yield; €29M a year additional yield, 53× five-year ROI.
Solutions -
PROCESS ENGINEERING
A column that cycles.
Steam control lags the column, so temperature swings and product goes to rework. A predictive model sets the steam for the temperature you need next.
An estimated USD 250,000 reduction in annual costs in VCM production.
Read the story -
BLENDING & PACKAGING
Flush matrices that over-flush by design.
Static product-family rules flush more than the quality spec needs; industry-wide, 20–30% of flush volume is avoidable. The Flush Optimizer solves the blending physics for the actual previous and next product.
−18% flush volume on lubricant changeovers, quality maintained; $20M a year value added, 27× five-year ROI.
Flush Optimizer, in depth -
RELIABILITY
Exchangers cleaned on the calendar, at every site.
A chemicals group maintained its heat exchangers on a fixed schedule at tens of thousands of dollars each a year, and an unplanned outage cost a fifth more than a planned one. Differential pressure over the hot side flags fouling in each operating mode and tells the operator when to clean.
A business case on more than 10% lower maintenance cost, built with the group in a study.
Read the story
See chemicals stories →Talk to someone who knows process plants →
02 · MARITIME & TRANSPORTATION
Your fleet is measured voyage by voyage.
Fuel, boil-off, reliquefaction, water, hull — every voyage is a performance and a contract. Sea-trial baselines drift out of relevance as vessels age. We benchmark each vessel against how comparable vessels actually performed, and tell the crew what to change.
LNG carriers · tankers and dry bulk · cargo fleets · cruise · offshore service vessels
-
FLEET TECHNICAL
Boil-off gas burned when it should be reliquefied — or the reverse.
The burn-vs-reliquefy decision is made by habit. The Burn Recommender computes the optimal trade-off for the boil-off gas system.
Over 90% of reliquefaction-system efficiency within target bands.
Solutions -
VESSEL PERFORMANCE
Benchmarks from the shipyard, not the sea.
Design and sea-trial baselines stop describing the vessel after its first years. Every voyage is benchmarked against comparable vessels, cargo and conditions.
360 measurements computed continuously across the fleet.
Read the LNG performance whitepaper -
FLEET OPERATIONS
Water carried as insurance.
Ships sail with more potable water than the voyage needs, at the cost of fuel and cargo space. The plan is set per voyage from crew, route and onboard capacity, and corrected live.
Up to $5M annual savings; on-vessel water reduced 15% across the fleet.
Read the story -
FLEET TECHNICAL
A combustion unit that cracks at sea.
An LNG carrier cannot sail without a working gas combustion unit, so a failure at the wrong moment means off-hire. A model watches the sensor data for the signature that precedes cracking in the cooling jacket and alerts the crew.
USD 200–300 thousand a year in estimated repair and maintenance savings, and lower off-hire risk.
Read the story
03 · OIL & GAS
Uptime and integrity, asset by asset.
Upstream value is lost to equipment that fails early and to maintenance done too often or too late. We predict which asset needs attention, and which does not, from the data your equipment already produces.
Offshore production · drilling · pipelines and integrity · oilfield services and pressure pumping · onshore and subsurface
-
PRODUCTION
A compressor failure nobody saw coming.
Export compressors fail rarely, and a single failure can cost tens of millions in lost production. A model trained on years of compressor data raised an alert two weeks before a dry gas seal failed, and named the sensors to look at.
300–370 kboe of production loss avoided on the single identified failure; USD 12–16 million a year in value.
Read the story -
FLEET MAINTENANCE
Pumps that fail between jobs.
A pressure-pumping fleet loses revenue every time a pump fails on location. Reliability models time the maintenance to the pump, not the calendar.
+28% pump-fleet MTBF; $9M a year lower OPEX, 14× five-year ROI.
Read the story -
PLATFORM OPERATIONS
Power lost on a platform nobody is on.
Generator failures on an unmanned export platform cost production every year, and the sensor patterns were too complex for rules. A model alerts operators when the generators drift from normal, weeks before most failures.
USD 1–2 million a year less production lost to generator failures.
Read the story -
ASSET INTEGRITY
Rehabilitation scoped by hand.
Pipeline coating scopes balance capital spend against integrity risk, one inspection run at a time. Machine learning aligns the inspections and classifies the scopes.
At least 5× return on investment from the automation alone.
Read the story -
MAINTENANCE
Site visits to check healthy pumps.
Seawater and freshwater pumps were serviced on schedule, whatever their condition. Anomaly detection across the fleet names the pumps that need a visit.
50% fewer unnecessary maintenance visits.
Read the story -
ENGINEERING SERVICES
Fatigue analysis too slow to run live.
Finite-element models of flexible risers are accurate and far too slow for real-time monitoring. A machine-learning model trained on the FEA results predicts stress and fatigue from vessel motion.
1000× faster fatigue insights than traditional FEA models, at over 98% accuracy.
Read the story -
RELIABILITY
Exchangers cleaned on a guess.
Fouling is measured on a few exchangers and guessed on the rest. A model trained on the well-instrumented unit predicts every exchanger's performance every 10 minutes.
5–7 million USD reduction in annual production losses from heat exchanger cleaning and failures.
Read the story
See oil & gas stories →Talk to someone who knows oil & gas →
04 · METALS & MINING
Uptime in the pit and the mill, power in the furnace.
Haul trucks and mills lose output to failures nobody saw coming. Melting, atomising and electrolysis turn power into product at a price that moves by the hour. We predict the failure early enough to plan the repair, and plan the load against the power price — within the limits the equipment and the metallurgy allow.
Mining and haulage · cement and minerals · metal powders and metallurgy · aluminium and silicon · electric-arc steel
-
FLEET MAINTENANCE
Haul trucks serviced on the calendar, failing between services.
An alumina miner's trucks raised more than a hundred types of alarm, and the frequent ones said little about a truck's condition. Six years of work orders set against the alarm history showed the pattern that comes in the month before an engine failure or a major overhaul — and models built on it tell the workshop which truck to bring in.
15% reduction in unplanned engine-related downtime, in a pilot on the haul-truck fleet.
Read the story -
CONDITION MONITORING
A haul-truck engine that fails in the pit.
Condition-monitoring agents watched the telemetry of mass haul trucks at mine sites, but could only report a fault once it showed. A survival model on the engine sensors gives the probability of a severe engine fault over the days ahead, and anomaly models tell the team which trucks to look at first.
Engine failures anticipated across the monitored fleet, triaged every day in the monitoring team's own workflow.
Read the story -
MAINTENANCE
Mill liners rotated on the calendar, a gearbox that failed twice.
A cement plant's vertical roller mill lost over a week each time its roller liners were rotated, and its gearbox had failed critically twice. A wear forecast names the day the liners reach their limit; a gearbox health model reproduces the run-up to the last failure.
A forecast of the liner rotation date, and a gearbox health metric that can warn if the known failure returns — from a feasibility study.
Read the story -
ENERGY & PRODUCTION PLANNING
A furnace plan that ignores the power market.
Load is planned for output, not for price. The Energy Optimizer plans power-intensive melting and electrolysis against forecast spot prices.
(result to come)
Energy Optimizer, in depth -
MELTING & ATOMISING
Heats that run long, and an atomiser waiting for melt.
A metal-powder plant's tap-to-tap times had become unpredictable, and the carbon–oxygen ratio of the atomised powder hard to hold. Diagnostics on the heat-level data first, then a prototype for the operators.
(result to come)
See metals & mining stories →Talk to someone who knows metals →
05 · POWER & UTILITIES
Generation and grid, ahead of the fault and the fine.
Utilities answer to regulators, to the grid and to customers at once. We predict what the network and the plant will do next — an emission limit, a faulty installation, a component near the end of its life — early enough to act.
Wind · solar · hydro · biomass · grid and distribution
-
GRID OPERATIONS
Smart meters installed wrong, found late.
Installation quality is hard to verify without a technician on site. Machine learning on consumption, GIS and electrical data finds the misconnections.
Up to $8.5 million in potential lost value identified — and the shock hazards found.
Read the story -
ENVIRONMENTAL COMPLIANCE
An emission limit breached before anyone sees it coming.
A biomass plant needed to know ten minutes ahead. Prediction models forecast emissions two and ten minutes out.
Three weeks from data to delivered prediction models.
Read the story -
GENERATION FORECASTING
A forecast due every five minutes, a penalty if it is wrong.
A spot market balances supply and demand every five minutes, so wind and solar farms report production forecasts at the same pace. Arundo Edge runs the forecasting model at the farm — 43 turbines on one, 32 inverters on the other — and serves it to the market operator.
Up to AUD 1 million in grid penalties avoided by reporting accurately and on time.
Read the story -
FLEET MAINTENANCE
Offshore turbines serviced by the weather window, not the fault.
Specialist vessels make every offshore visit expensive. Fleet-wide monitoring locates problems at farm, turbine or component level and warns of gearbox faults early enough to plan the vessel.
15% average improvement in uptime across the fleet.
Read the story -
SOLAR O&M
Inverters that fail with no warning in the SCADA.
Inverter failures were a solar operator's main source of lost production, and its SCADA could not tell a production loss from a dropped connection. Models set irradiance against power to find and classify every loss, and flag the intermittent drops that come before a long outage.
8% of production losses avoided, in a live pilot on the operator's plants.
Read the story
See power & utilities stories →Talk to someone who knows power →
06 · OEM & MANUFACTURING
Your know-how, sold as a live product.
Equipment makers and technology licensors own the deepest knowledge of how their kit behaves. We turn that knowledge into a service that runs against your customers' live data — so you sell outcomes, not only equipment. Manufacturers get the same from their own lines: every machine connected, measured the same way, at every site.
Equipment makers · automation vendors · process-technology licensors · service companies · discrete and consumer-goods manufacturing
-
AFTERMARKET SERVICES
Cranes sold once, serviced on a guess.
A crane maker wanted the chief engineer on board to run its cranes better, and a closer aftermarket relationship. We co-developed the on-vessel application: more than 100 KPIs and machine-learning models that predict critical operating modes, with root cause and spare parts in three clicks.
10–15% more uptime across the vessel fleet, offshore and merchant.
Read the story -
PRODUCT DEVELOPMENT
Well rates nobody measures.
Physical multiphase meters are too costly to put on every well, so production is optimised blind. With an automation vendor, we built a virtual meter that calculates oil, gas and water rates in real time from the measurements already there.
A virtual meter the vendor sells to its own customers.
Read the story -
BATTERY STORAGE AS A SERVICE
One breakdown erases the profit of dozens of units.
Storage sold on 10–20 year contracts lives or dies on acting early on faults. A digital twin template links each unit's components, telemetry and manuals; anomaly detection runs on schedule, and an AI agent answers the service team in their chat.
100-unit fleet built from one template, models running from day one.
-
OPERATIONS EXCELLENCE
Factories that cannot see each other.
A multinational manufacturer wanted its key equipment connected across locations, with the same KPIs everywhere. Live equipment data streams to one platform, with models from machine to plant to network level.
USD 30–40 million from the efficiency boost, and USD 6–10 million from the utilisation increase.
Read the story -
PRODUCTION QUALITY
A vision model retrained for every new product.
Products were identified on the line by image analysis, but the model drifted and had to be rebuilt for every new SKU. Deep-learning computer vision now labels unknown SKUs in real time, without retraining.
100% SKU identification accuracy in a pilot.
Read the story
ACROSS ALL SIX
The same failure, in every industry.
A compressor seal, a pump, a fouled exchanger, an engine: the equipment that stops a plant is the same on an LNG carrier, an offshore platform and a chemicals site. We have predicted these failures before, in more than one industry — so your project starts from failure modes we have already modelled, not from a blank page.
-
Compressors
Oil & gas · chemicals & processing
6 use cases
-
Pumps
Oil & gas · chemicals & processing · OEM & manufacturing
6 use cases
-
Engines, generators and combustion units
Oil & gas · maritime & transportation · metals & mining
6 use cases
-
Heat exchangers and boilers
Oil & gas · chemicals & processing
4 use cases
-
Cranes and gearboxes
OEM & manufacturing · power & utilities · metals & mining
6 use cases
Problems like these now run on one platform. Arundo Foundation