Pharma plants: release tests slow, cycles and air run for safety

Release tests are slow and real-time quality is expected

What happens. A batch is finished, but it waits for the lab results before it can be released. The authorities and customers expect quality to be understood during the process, not only afterwards.

What it costs. Hold times, stock that is stuck, and lost batches that were already out of range an hour into the run but only found out at the end.

Why the usual fix fails. More analysers are expensive to validate and still measure only some quantities. Faster labs shorten the wait but do not bring the answer into the process.

How a fixed-price remote diagnosis checks it. From batch records, process curves and lab results I check which critical quality values can be computed from what is measured already. The fixed-price remote diagnosis tells you whether a computed value follows the lab closely enough to be trusted, and what it would take.

Background: Computed measurements

Granule moisture and size are checked offline

What happens. In granulation, moisture and particle size are taken from samples and measured off the line. The granulator runs on until the answer comes back.

What it costs. Batches that fail on size or moisture and have to be reworked or lost, and a lot of waiting for the lab.

Why the usual fix fails. A fixed end point by time or liquid amount ignores differences in the powder. Sampling at the end shows the result but cannot change it.

How a fixed-price remote diagnosis checks it. From granulator power, torque, spray rate, temperatures and the lab results of past batches I check whether moisture and size, and so the end point, can be computed during the run. The fixed-price remote diagnosis tells you how reliable this would be.

Background: Computed measurements

Fluid-bed drying runs by time, not by product state

What happens. The dryer runs for a set time or until a temperature is reached. The moisture of the product is measured by sample, if at all.

What it costs. Over-dried product and energy, or a batch that is stopped too soon and has to go back.

Why the usual fix fails. A shorter time works for a good lot and fails for a damp one. The outlet air temperature gives only an indirect hint, which depends on the load and the weather.

How a fixed-price remote diagnosis checks it. From inlet and outlet air data, bed temperature, flows and moisture results I estimate the product moisture during drying. The fixed-price remote diagnosis shows whether the end point can be controlled by the estimated state.

Background: Computed measurements

Tablet weight and hardness drift with the powder

What happens. The press is set up and runs well, then weight and hardness drift as the powder flows differently or the density changes. Operators adjust fill depth and compression force by hand.

What it costs. Rejected tablets, sorting and time spent on adjustments. Every change needs a decision on whether the product is still fine.

Why the usual fix fails. The press controls compression force, but the powder properties behind it are not measured. A fixed setting cannot follow them.

How a fixed-price remote diagnosis checks it. From force, fill, press speed and the tablet test results I estimate how the powder behaves and whether the press settings can follow. The fixed-price remote diagnosis tells you whether adaptive control on an estimated density would be an improvement.

Background: Adaptive predictive control

Coating thickness is known only after the run

What happens. In the drum, the coating is sprayed for a set time or amount. The thickness on the tablets is only known when the batch is out and tested.

What it costs. Rework on batches with too little coating, and extra coating material and time as a safety margin.

Why the usual fix fails. Weighing samples during the run is slow, and it interrupts the process. A fixed spray amount does not account for losses and for differences between batches.

How a fixed-price remote diagnosis checks it. From spray rate, air flows, temperatures, pan speed and the weights from past batches I check how well the coating growth can be computed. The fixed-price remote diagnosis shows whether the end point could be set by the computed thickness.

Background: Computed measurements

Blend uniformity is checked by sampling

What happens. Blends are tested by thief samples from the blender or from the bin. Some samples are taken, others not, and the sampling itself can disturb the powder.

What it costs. Hold times, repeated blending and lots that wait in quarantine.

Why the usual fix fails. More samples cost more time and give a picture at points only. A fixed blending time is chosen for the worst case.

How a fixed-price remote diagnosis checks it. From near-infrared spectra, blender data and the sample results I check whether uniformity can be followed in the blender with a model. The fixed-price remote diagnosis tells you how good that model is likely to be on your data.

Background: Computed measurements

Continuous manufacturing needs quality values in-line

What happens. A continuous line makes product all the time, so quality has to be known all the time. Without in-line values for the critical attributes, nothing can be released in real time and the line is run like a batch plant.

What it costs. Lost benefit of continuous manufacturing, because every lot still waits for tests, and material that goes to waste when something changes upstream.

Why the usual fix fails. A sensor for each attribute is not available or too expensive. Models built once and left alone drift as the material and the equipment age.

How a fixed-price remote diagnosis checks it. From the line’s process data, spectra and lab results I check which attributes can be estimated, how well, and with what upkeep. The fixed-price remote diagnosis tells you what is realistic before you commit to validation.

Background: Computed measurements

The end of primary drying in freeze drying is judged cautiously

What happens. In a freeze-drying cycle, nobody knows exactly when the primary drying is finished. The end is set with a long safety margin, because going to secondary drying too early can collapse the product.

What it costs. Cycle days that were not needed, energy, and an occupied freeze dryer that blocks the next batch.

Why the usual fix fails. Pressure rise tests and probes give a hint for some vials only. Shortening the cycle on experience is risky for a valuable product.

How a fixed-price remote diagnosis checks it. From shelf temperatures, chamber pressure, condenser data and product probes I estimate the product temperature and the end of primary drying. The fixed-price remote diagnosis tells you how much of the margin an estimate could safely take out.

Background: Computed measurements

Sterilisation cycles are long and fixed

What happens. Steam sterilisation runs a fixed, long cycle that guarantees the result for every load. The real heat reaching the load is not tracked as it develops.

What it costs. Throughput, since the autoclave is blocked, and energy and steam. Heat-sensitive products are stressed longer than necessary.

Why the usual fix fails. Shortening the hold time without evidence is not allowed. Validation fixes the cycle, and the data to justify another one is often missing.

How a fixed-price remote diagnosis checks it. From temperature and pressure records of past cycles I compute the lethality that each load received. The fixed-price remote diagnosis shows whether a cycle optimised within the validated limits would be shorter.

Background: Economic predictive control

Cleanroom air changes stay at maximum all day

What happens. The air handling in the cleanroom runs at full air change at all times, whether the room is in use or not. Nobody dares to change it because of the classification.

What it costs. Fan and cooling energy for air that cleans an empty room.

Why the usual fix fails. A night setback by timer is a blunt tool and can lose the classification after the restart. Adding sensors does not help if nobody uses them for control.

How a fixed-price remote diagnosis checks it. From particle counts, pressure differences, air flows and room use I check how much air the room needs in each state and how fast it recovers. The fixed-price remote diagnosis tells you whether demand-led control within your classification limits is realistic.

Background: Economic predictive control

Water-for-injection loops run hot and circulate constantly

What happens. The loops for water for injection run hot with high circulation all the time, in line with the requirements for microbial safety. The same setting is used at all hours, regardless of the demand.

What it costs. Heat, cooling and pumping energy around the clock.

Why the usual fix fails. Changing the temperature or velocity without a basis would risk the validated state. Each change seems to need a lengthy argument.

How a fixed-price remote diagnosis checks it. From loop temperatures, flows, demand profiles and heat consumption I check where energy is lost and what the loop would need at the lower end of its limits. The fixed-price remote diagnosis tells you whether optimised operation within the limits is worth a proposal.

A lab recipe fails at production scale

What happens. A process that works in the lab behaves differently in the production vessel. Mixing, heat transfer and times do not scale in a simple way.

What it costs. Delays in bringing a product to market, failed batches and months of trial runs.

Why the usual fix fails. Scaling by volume or a rule of thumb ignores the physics that change. Each trial at production scale is expensive.

How a fixed-price remote diagnosis checks it. From lab records and the first production runs I check which effects differ with scale and whether a model can predict the production behaviour. The fixed-price remote diagnosis tells you whether a scale-up model is likely to cut trials.

Near-infrared models break with new excipients

What happens. A near-infrared analyser works well until a new supplier or a changed excipient shifts the spectra. Predictions become wrong or unstable, and nobody knows since when.

What it costs. Recalibration campaigns and a loss of trust in the analyser. Meanwhile samples go back to the lab.

Why the usual fix fails. Building a new model for every change is slow. Monitoring is often just a check of the lab value from time to time.

How a fixed-price remote diagnosis checks it. From your spectra, reference values and material changes I check how the model drifts and whether a combined model with drift monitoring would stay valid. The fixed-price remote diagnosis tells you how much of the recalibration effort could be saved.

Background: Adaptive predictive control

The root cause of a deviation is hard to find

What happens. A batch deviates and the investigation starts. The data is spread over systems, and people compare a few trends by eye.

What it costs. Weeks of investigation per deviation, repeated deviations and held batches.

Why the usual fix fails. A review by a team of experts is thorough and slow, and it depends on who remembers a similar case. Simple limit checks flag the symptom, not the cause.

How a fixed-price remote diagnosis checks it. I line up batch data from good and bad batches and look for the differences that explain the deviation. The fixed-price remote diagnosis tells you whether model-based diagnosis can narrow the cause down faster for your type of case.

Equipment and cleaning are scheduled by hand

What happens. The planner fits batches, equipment and cleaning into a schedule by hand. Cleaning times and campaign rules make it complicated, and a change in one place moves everything.

What it costs. Idle equipment, waiting for cleaning and late batches.

Why the usual fix fails. A spreadsheet does not keep the rules consistent, and the planner’s own experience cannot be passed on easily.

How a fixed-price remote diagnosis checks it. From order lists, equipment data and cleaning rules I check how much idle time the current plans show and whether an optimised schedule would reduce it. The fixed-price remote diagnosis tells you what the gain could be.

Background: Trajectory optimisation and setpoint generation

Crystal form and size of active ingredients are not controlled

What happens. In crystallisation, the crystal form and size follow the cooling and the supersaturation, which are not measured. The batch goes through the same cooling plan every time.

What it costs. Failed batches, difficult filtration and drying, and a product that sometimes does not meet specification on the form.

Why the usual fix fails. A fixed cooling profile cannot respond to the real concentration. Taking samples for analysis comes too late.

How a fixed-price remote diagnosis checks it. From temperatures, concentrations, particle data and past batch results I estimate the supersaturation and test a better profile on a model. The fixed-price remote diagnosis shows whether controlled crystallisation is realistic for your product.

Background: Trajectory optimisation and setpoint generation

Solvent recovery columns are run conservatively

What happens. The distillation for solvent recovery is run with a large safety margin on purity, since purity is only known from a sample.

What it costs. Steam and cooling energy, and solvent lost to waste.

Why the usual fix fails. Fixed temperature control does not follow changing feed. More reflux is the easy answer and costs energy.

How a fixed-price remote diagnosis checks it. From column temperatures, flows, energy and lab purities I check whether purity can be computed and the column controlled at lower cost. The fixed-price remote diagnosis tells you how much of the margin could go.

Background: Economic predictive control

A transfer to a new site means re-tuning from scratch

What happens. A product moves to another site or line, with the same recipe but different equipment. The settings that worked before have to be found again by trials.

What it costs. Months before the new site makes the product reliably, and trial batches.

Why the usual fix fails. The recipe describes the product, not how the equipment behaves. Copying the settings does not give the same result.

How a fixed-price remote diagnosis checks it. From data of both sites I compare how the equipment responds and check whether the settings can be translated by a model. The fixed-price remote diagnosis tells you how much of the re-tuning could be saved.

Signals are scattered and not time-aligned

What happens. Process data sits in several systems with different time stamps and names. To answer a question someone first has to collect and align it.

What it costs. Analysis that is not done at all, and long preparation for each study.

Why the usual fix fails. A new central data system is a big project. Without cleaned and aligned data, new software has nothing to work on.

How a fixed-price remote diagnosis checks it. I take an example of your data and check what it takes to clean, align and keep it, and which analysis it would then allow. The fixed-price remote diagnosis gives you a realistic first step.

Visual inspection staff are scarce

What happens. Visual inspection needs trained people, and there are too few. Inspection is a bottleneck at the end of the line.

What it costs. Waiting product, overtime and the risk of defects passing.

Why the usual fix fails. Hiring is slow and the work is tiring. Camera systems cover a part of the defects and need their own validation.

How a fixed-price remote diagnosis checks it. From process signals and inspection results I check whether the quality of units can be estimated from the process, so that inspection can focus on the doubtful ones. The fixed-price remote diagnosis tells you how far this carries.

Background: Computed measurements

Nobody wants to change a validated control

What happens. The control works well enough, is validated, and every change means paperwork. So nobody touches it, even when it is known to be wasteful.

What it costs. Years of the same waste and variation, and a plant that cannot profit from better methods.

Why the usual fix fails. Arguments in favour of the change lack evidence, and the risk to the validated state is real. Trying out on the live plant is not an option.

How a fixed-price remote diagnosis checks it. From your data I build a simulation of the current control and test a change against it. The fixed-price remote diagnosis gives you the evidence that a change is better, and which steps would be needed for validation.

Contamination control for sterile products leaves monitoring data unused

What happens. The revised European rules for sterile manufacturing ask for a contamination control strategy, with more environmental monitoring and trend analysis (guide to the revised rules). Older plants often have no real-time data, or they have it and nobody evaluates it. Air handling is kept at maximum for safety.

What it costs. Monitoring that produces data without use, and energy for air handling set to the worst case.

Why the usual fix fails. Alarm limits on single values produce alarms without insight. More sensors add data and work.

How a fixed-price remote diagnosis checks it. From particle counts, pressure differences, temperatures and air handling data I check what trends tell and whether air handling can follow the room state within your limits. The fixed-price remote diagnosis tells you what the data supports.

Background: Economic predictive control

New and modernised active-ingredient plants must start up reliably

What happens. A proposed European Critical Medicines Act aims to strengthen the production of critical medicines and their active ingredients in Europe, with easier access to support for projects that build or modernise plants (European Parliament). These plants have to run reliably early on.

What it costs. Delays in start-up, early batches that fail and months of tuning.

Why the usual fix fails. Plants are often started with the tuning of the supplier and the know-how of few people. Continuous manufacturing adds a need for process control that many sites do not have.

How a fixed-price remote diagnosis checks it. From the design data and, where available, first runs, I check which control and which computed quality values the plant needs for a stable start. The fixed-price remote diagnosis tells you where the biggest risks are before the start-up.