Refineries: crude switches, product giveaway and old controllers

Crude switches upset the unit

What happens. When the crude changes, the distillation unit sees a different mix of light and heavy fractions. Temperatures, flows and product qualities drift for hours until the operators have caught up by hand.

What it costs. Off-spec product during the transition, yield lost to the wrong cut points and energy spent on correcting. Crude switches are frequent, so the losses add up.

Why the usual fix fails. Operators wait for the lab or the analyser to tell them what the new crude is. The controllers are tuned for the old feed and fight the change.

How a fixed-price remote diagnosis checks it. From assay data, feed records and unit logs of past switches, I check whether the properties of the new crude can be estimated at once and used as a feedforward to the unit. The fixed-price remote diagnosis tells you what a transition could gain.

Background: Control that adapts to drift and wear

Distillation qualities come from an analyser every half hour

What happens. The quality of the products is known only when the analyser reports, or when the lab result arrives. In between, the unit runs on the last value. A disturbance stays unseen until the next value.

What it costs. Giveaway: the products are made better than the specification as a safety margin, which costs yield of the valuable fraction and energy.

Why the usual fix fails. More frequent analyser readings do not exist; the analyser is slow by nature. Controlling on tray temperatures alone ignores pressure and composition.

How a fixed-price remote diagnosis checks it. From temperatures, pressures, flows and the analyser and lab values, I check how well the quality can be computed continuously, and how much margin a faster value could release. The fixed-price remote diagnosis answers for your column.

Background: Computed measurements

Fired heaters run with excess oxygen as a margin

What happens. The furnace burns fuel with more air than the flame needs, to be sure there is no unburnt gas or smoke. The oxygen in the flue gas is held at a safe level, even when the load, the fuel gas or the weather change.

What it costs. Fuel spent on heating air that goes up the stack. A stack loss that is higher than necessary and extra emissions per tonne of product.

Why the usual fix fails. Lowering the oxygen setpoint by hand risks carbon monoxide when something changes. A single oxygen probe in the stack does not tell about each burner.

How a fixed-price remote diagnosis checks it. From flue-gas oxygen, carbon monoxide, fuel and air flows and temperatures, I check how close combustion can run to the limit and what protections it needs. The fixed-price remote diagnosis tells you what fuel saving is plausible.

Background: Predictive control of process plants

Catalyst deactivation shortens the run of a hydrotreater

What happens. A catalyst slowly loses activity. The operators raise the reactor temperature to keep the product quality, until the maximum is reached and the unit has to be stopped. How active the catalyst really is, nobody measures directly.

What it costs. A shorter run between catalyst changes, extra stops, energy for higher temperatures and a quality that nears the limit near the end of the run.

Why the usual fix fails. Raising the temperature as a rule of thumb is conservative early and too late at the end. Schedules for catalyst changes by calendar ignore the feed and the load.

How a fixed-price remote diagnosis checks it. From temperatures, pressures, hydrogen use, feed data and product quality over past runs, I check how well the activity can be estimated and the temperature profile planned along the run. The fixed-price remote diagnosis tells you how much run length can be gained.

Background: Control that adapts to drift and wear

Coke growth in cracker furnaces is unknown until the decoke

What happens. In cracking furnaces, coke grows on the inside of the tubes. It cuts heat transfer and raises the pressure drop and the tube temperature. The furnace is decoked when the readings say so or when the calendar says so, and nobody knows how much coke is in the tube.

What it costs. Run length is shorter than necessary when the furnace is stopped early, and tubes are damaged when it is stopped late. Each decoke costs production and energy.

Why the usual fix fails. Using the pressure drop alone is ambiguous because it depends on the feed and the load. A fixed run length is a compromise.

How a fixed-price remote diagnosis checks it. From tube temperatures, pressure drop, feed and steam rates and records of past decokes, I check whether coke growth can be estimated and the decoke planned on it. The fixed-price remote diagnosis tells you what to expect.

Background: Computed measurements

Product blends carry giveaway

What happens. Fuels and other products are blended from several components to meet a specification. The blend is calculated from lab values of the components, which are old by the time they are used, and then corrected. To be sure, the blend is made better than needed.

What it costs. Expensive components used where cheaper ones would pass, and blends that miss and have to be reworked. Over a year, the giveaway adds up to a lot of margin.

Why the usual fix fails. Blend recipes from the planning office are based on the last lab data. A new analyser in the blend line gives a value but not a decision.

How a fixed-price remote diagnosis checks it. From component qualities, the blend history and the analyser and lab values, I check how well qualities can be estimated and a blend optimised against the specification. The fixed-price remote diagnosis tells you how much giveaway could be removed.

Background: Economic predictive control

The hydrogen balance is managed by hand

What happens. Hydrogen is made in one unit and used in many others. The balance between producers, consumers and compressors is managed by the operator in the control room, who reacts when the pressure in the network drops or rises.

What it costs. Hydrogen vented to the flare when there is a surplus, extra hydrogen produced or bought when there is a shortage, and compressors that run harder than they need to.

Why the usual fix fails. Operating rules for a fixed order of consumers cannot follow the changes of the units. A bigger buffer in the network is paid for in cost.

How a fixed-price remote diagnosis checks it. From network pressures, flows, unit loads and plans, I check how hydrogen could be distributed to give the lowest cost and least venting, with forecasts of consumption. The fixed-price remote diagnosis tells you what saving is plausible.

Background: Economic predictive control

Gas processing keeps large margins to the dew point

What happens. Gas must leave the plant with a dew point low enough for the pipeline. The dew point is measured by an analyser with delay, or calculated from a few signals. The plant runs with a margin so that it does not go off-spec.

What it costs. Throughput below what the plant could do, extra cooling or regeneration energy and gas that is dried and cooled further than needed.

Why the usual fix fails. A bigger margin is safe and costly. A faster analyser does not remove the dependence on the load and the inlet gas.

How a fixed-price remote diagnosis checks it. From inlet gas composition, temperatures, pressures and dew-point measurements, I check whether the dew point can be estimated well enough to run closer to the limit. The fixed-price remote diagnosis tells you how much margin could be released.

Background: Computed measurements

Compressor stations keep large surge margins

What happens. A compressor can surge, which is violent and can destroy it. To stay clear, the control keeps the operating point well away from the surge line, and opens an anti-surge valve early.

What it costs. Gas recirculated through the valve is compressed twice for nothing, which costs power. The compressor works at a point with lower efficiency than it could.

Why the usual fix fails. Moving the surge line closer by hand is a risk nobody likes to take. A map from the maker is valid for new machines and for the design gas.

How a fixed-price remote diagnosis checks it. From compressor data (flows, pressures, temperatures, speeds, valve positions) I check how closely the surge limit can be estimated for the present state and gas. The fixed-price remote diagnosis tells you how much power could be saved.

Background: Computed measurements

Flaring during process upsets

What happens. When a unit is upset, gas goes to the flare. The cause is often a chain of events across several units, where one operator acts and another is surprised. The flare burns, with smoke and noise that neighbours notice.

What it costs. Lost product and fuel, emissions, complaints from the neighbourhood and trouble with the permit. Every flaring event has to be reported.

Why the usual fix fails. Alarms come when it is already happening. Each unit’s control acts on its own and does not know what the next unit will do.

How a fixed-price remote diagnosis checks it. From alarm logs, flow and pressure records of past events, I check which upsets can be foreseen and how the units could be coordinated to lower the load on the flare. The fixed-price remote diagnosis tells you what a coordinated control would address.

Background: Predictive control of process plants

Tank movements are planned by hand

What happens. The tank farm receives crude and products from ships and pipelines and delivers to units and customers. Planners fit the movements into tanks, pumps and lines by hand in a spreadsheet, and change the plan whenever a ship is late.

What it costs. Demurrage when a ship waits, mixing errors when the wrong product goes into a tank and idle capacity because of a cautious plan.

Why the usual fix fails. A spreadsheet checks one plan, but cannot look for a better one. More tank capacity as a buffer is the most costly remedy.

How a fixed-price remote diagnosis checks it. From tank data, ship schedules, line and pump constraints and past movements, I check how an optimiser could suggest schedules and re-plan after a delay. The fixed-price remote diagnosis tells you what it could take out of the cost.

Background: Optimal paths and setpoint plans

Electric furnaces for steam crackers bring new dynamics

What happens. Plans to heat crackers with electricity change how the furnace behaves. Electric heating reacts faster than burners and follows the power supply. The existing control was designed for gas firing.

What it costs. Uncertainty about stability and product quality before the first unit is built, and a risk of expensive changes after start-up.

Why the usual fix fails. Taking the control concept of the gas furnace and swapping the actuator ignores the different dynamics. Trial and error on a full-scale furnace is out of the question.

How a fixed-price remote diagnosis checks it. From furnace design data and the process, I check what a model of the electrically heated furnace shows about the dynamics and a suitable control concept. The fixed-price remote diagnosis tells you what to clarify before the design is frozen.

Background: Predictive control of process plants

Bio feed co-processing changes how the unit behaves

What happens. Refineries add vegetable oils or other biological feed to the normal crude cuts. The mix of molecules is different: more oxygen, more water, other contaminants. The unit, tuned for fossil feed, responds differently to each batch.

What it costs. Faster catalyst ageing, product quality that moves and temperatures and pressures that the operators have no feel for yet.

Why the usual fix fails. Treating the bio feed as one more cut with a fixed share does not follow the differences between batches. Tests on the real unit are expensive and slow.

How a fixed-price remote diagnosis checks it. From feed analyses, unit logs and product qualities with bio shares, I check how the unit responds and whether an adaptive control can follow the batches. The fixed-price remote diagnosis tells you what you could expect.

Background: Control that adapts to drift and wear

Analysers are expensive and often down

What happens. Online analysers are costly to buy and to maintain. They need sample lines, conditioning and calibration, and they break down. When one is out, the loops that depend on it are put on manual.

What it costs. Control that goes blind for days, margins that rise during the outage and maintenance staff who spend a lot of time keeping instruments alive.

Why the usual fix fails. Buying a second analyser doubles the maintenance. Manual operation relies on the lab value of the shift.

How a fixed-price remote diagnosis checks it. From the process variables that keep running and the analyser’s own history, I check whether a computed value can stand in for the analyser during outages and be corrected when the analyser returns. The fixed-price remote diagnosis tells you which analysers are good candidates.

Background: Computed measurements

Hundreds of control loops, many of them poor

What happens. A refinery has hundreds or thousands of control loops. Many oscillate, sit in manual, or hit a valve limit. The control engineers cannot look at all of them, so the poor ones stay poor.

What it costs. Variability in the process that forces margins on quality and energy, and loops that cause trouble in other units. Valuable benefits of the advanced control layer are lost when the base loops fail.

Why the usual fix fails. Checking loops by hand on a complaint finds the loudest ones, not the most expensive ones. More engineers for loop work are not available.

How a fixed-price remote diagnosis checks it. From the loop logs of the control system (setpoints, measurements, outputs, modes) I rank loops by what they cost and find the likely causes. The fixed-price remote diagnosis gives you a list of the loops worth fixing first.

Alarm floods hide the real fault

What happens. When a unit trips, dozens of alarms come within a minute. The operator in the control room cannot tell which is the cause and which are the consequences. Many are standing alarms that nobody has cleared.

What it costs. Slow reaction to the real fault, wrong decisions under stress and a safety risk. Plants that tolerate many alarms every day train operators to ignore them.

Why the usual fix fails. Raising alarm limits to reduce the count silences the useful ones too. A rule list for every combination is never complete.

How a fixed-price remote diagnosis checks it. From alarm and event logs, together with the process structure, I check which alarms follow from which cause and where a model could point to the first one. The fixed-price remote diagnosis tells you what could be removed or grouped.

Old multivariable controllers are switched off and the benefit is lost

What happens. Advanced process control was installed years ago. The plant changed, the people who built it have left, and the operators have switched the controller off because it behaves oddly. The unit runs by hand again.

What it costs. The benefits it once gave in yield and energy are lost, and nobody has them in the accounts as a loss. A new project has to start from scratch.

Why the usual fix fails. Retuning the old controller without the know-how rarely works. The software licence and the old model are often outdated.

How a fixed-price remote diagnosis checks it. From the history of the controller, logs from when it ran well and the unit’s present data, I check what has changed and whether a re-identification of the models can bring it back. The fixed-price remote diagnosis tells you what a renewal would take.

Background: Predictive control of process plants

Site steam and power are not optimised together

What happens. The site makes steam in boilers and turbines, buys or sells power and runs the units’ demands. Steam headers, turbines and boilers are operated by rules for each, and nobody looks at the whole.

What it costs. Steam vented at one pressure level while another is short, turbines running at a poor point and power bought at a high price while own generation is idle.

Why the usual fix fails. Rules per header cannot see the cost of power at the moment. A better boiler brings less than a better operating plan for what is installed.

How a fixed-price remote diagnosis checks it. From steam balances, turbine and boiler data, power prices and unit demands, I check how a combined plan of steam and power could lower cost. The fixed-price remote diagnosis tells you what saving to expect.

Background: Economic predictive control

Emission analyser outages put compliance at risk

What happens. Emissions from stacks and furnaces are measured by analysers that must run continuously. When one fails, the plant has gaps in its record and must follow the permit’s rules for replacement values or a stop.

What it costs. Penalties and reporting effort, replacement values that are conservative and plant load limited by the missing measurement.

Why the usual fix fails. A second analyser doubles the equipment and the maintenance. Manual samples do not fill the gap.

How a fixed-price remote diagnosis checks it. From process data, fuel and flue-gas signals and the analyser’s history, I check whether emissions can be computed to a quality the permit accepts as a substitute. The fixed-price remote diagnosis tells you what proof the authority would likely expect.

Background: Computed measurements

Plants converted to new feeds behave in unknown ways

What happens. Units are converted to process new feeds, such as waste oils, bio-based oils or hydrogen. The pipes and vessels are the old ones, the chemistry is not. No operator has run it before.

What it costs. Surprises at start-up: unstable temperatures, product off specification, fouling. Start-up takes longer and the first months are not profitable.

Why the usual fix fails. Relying on the lab pilot ignores the scale of the plant. The old operating instructions describe the old behaviour.

How a fixed-price remote diagnosis checks it. From design data, lab results and the existing unit’s records, I check what a simulation of the converted unit shows about stability and limits, before the plant is touched. The fixed-price remote diagnosis tells you what to clarify first.

Background: Predictive control of process plants