A Buffer-Refill Scheduling System for a Digital Small Scale Continuous Bioprocess
(2025) KETM05 20252Chemical Engineering (M.Sc.Eng.)
- Abstract
- This thesis addresses a challenge in continuous high performance liquid chromatography (HPLC): ensuring reliable, timely supply of buffer solutions. As biopharmaceutical production increasingly moves toward continuous operation, buffer management becomes critical, since chromatography requires multiple buffer types with precise compositions for washing, elution, equilibration, and regeneration. In an integrated small-scale continuous DSP operated through Orbit, a custom Python-based control platform, a digital Buffer Management System (BMS) exists, managing buffer mixing and refill order placement. Order placement occurs only when crossing a minimum volume threshold and the queue is simply first-in first-out, which creates high risks of... (More)
- This thesis addresses a challenge in continuous high performance liquid chromatography (HPLC): ensuring reliable, timely supply of buffer solutions. As biopharmaceutical production increasingly moves toward continuous operation, buffer management becomes critical, since chromatography requires multiple buffer types with precise compositions for washing, elution, equilibration, and regeneration. In an integrated small-scale continuous DSP operated through Orbit, a custom Python-based control platform, a digital Buffer Management System (BMS) exists, managing buffer mixing and refill order placement. Order placement occurs only when crossing a minimum volume threshold and the queue is simply first-in first-out, which creates high risks of delayed refills, idle systems, and buffer depletion, particularly when multiple clients operate simultaneously.
The aim of this project is to design and implement a more dynamic, predictive buffer-refill scheduling system to the BMS, that can estimate buffer depletion in advance and prioritize refill orders accordingly. This required the development of copies of each HPLC client system in Orbit in order to extract process information without interfering with the real running process.
A new buffer-depletion calculation routine was developed, consisting of three cooperating functions to extract instruction information from the active phases, pick out flask flowrates, integrate consumption over time, and calculate both the time until a minimum threshold is reached and the required refill volume at that moment in time. Instead of issuing refill orders once a low-volume threshold is crossed, clients now continuously provide the engine with depletion data every ten seconds. The engine then sorts all buffer orders, prioritizing shortest time until the threshold is reached and then largest refill volume, before putting them in the queue, which continually reorganizes itself.
The new system was benchmarked against the previous version of the BMS using three scenarios: a single client with alternating buffer consumption from two flasks, the same client with an added step-gradient phase, and two clients consuming three buffers each. In the single client cases, the new BMS maintained buffer levels well above the minimum threshold throughout 32-cycle tests, ensuring more stable buffer volumes than the old system, which often approached very low volumes before receiving a refill. A gradual downward trend in peak refill volumes was observed for the new system, attributed to the fixed dead-time required for buffer preparation.
Performance in the dual-client scenario was more mixed. Although the new BMS slightly extended the time, managing up to 12 hours, before any complete depletion occurred, high combined buffer demand, small order placements and long preparation times for certain buffer types created bottlenecks and cascading delays. The scheduler struggled when too many buffers were consumed simultaneously or when preparation times varied drastically between buffer recipes. In some cases, the new system outperformed the low volume threshold-based design, but in others the old system was superior.
The thesis concludes that the dynamic, buffer-refill scheduling system markedly improves robustness and maintains more even buffer volumes in buffer management for single client continuous DSP. It is uncertain whether it outperforms the old system during more demanding multi-client operation. The work highlights the need for enhanced prediction models (e.g., more accurate interpolation and preparation-time-aware prioritization), handling of non-time-based phases, and better support for smooth buffer gradients. Future development should also integrate the digital BMS with the physical DSP to validate its real-world performance and refine the order- and refill system further. (Less) - Popular Abstract (Swedish)
- Efterfrågan på biologiska läkemedel blir allt större, i och med att det med tiden upptäckts fler användningsområden för till exempel antikroppar, enzymer och virus. Dessa har i många fall blivit standardläkemedel för att behandla till exempel diabetes, vissa neurologiska sjukdomar och HIV/AIDS. Biologiska läkemedel känneteckas av att produceras av celler som genetiskt har modifierats. Från dessa celler kan läkemedlenas verksamma komponenter utvinnas.
Trots många stora satsningar och en stor utveckling av industrin är biologiska läkemedel fortfarande mycket dyra att producera. Det är inte bara all forskning och alla försök för att få fram exakt den komponent man är ute efter som blir dyrt, utan i många fall står uppreningen för den... (More) - Efterfrågan på biologiska läkemedel blir allt större, i och med att det med tiden upptäckts fler användningsområden för till exempel antikroppar, enzymer och virus. Dessa har i många fall blivit standardläkemedel för att behandla till exempel diabetes, vissa neurologiska sjukdomar och HIV/AIDS. Biologiska läkemedel känneteckas av att produceras av celler som genetiskt har modifierats. Från dessa celler kan läkemedlenas verksamma komponenter utvinnas.
Trots många stora satsningar och en stor utveckling av industrin är biologiska läkemedel fortfarande mycket dyra att producera. Det är inte bara all forskning och alla försök för att få fram exakt den komponent man är ute efter som blir dyrt, utan i många fall står uppreningen för den största kostnaden. I detta steget behöver den verksamma komponenten separeras från allt annat som en cell innehåller. Separationen genomförs på olika sätt, beroende på produkten och vad det är den behöver separeras ifrån. Ett exempel på en separationsmetod är jonbyteskromatografi. Här separeras alla ämnen utifrån deras laddningsskillnad.
Principen för jonbyteskromatografi är att alla ämnen blandas ut i en saltlösning som förs genom en kolonn bestående av en laddad yta. Alla ämnen, inklusive den verksamma komponenten kommer att fastna på denna ytan på grund av deras laddningsskillnad. Därefter spolas en vätska med en annan saltkoncentration, som kallas för en buffert, genom kolonnen, och beroende på ämnenas laddning kommer vissa på den laddade ytan att lossna och spolas med. Idén är att tvätta bort allt annat än den verksamma molekylen, som kan spolas ut i slutet och fångas upp.
Som med mycket annan produktion, önskas jonbyteskromatografi att köras kontinuerligt, så att man hela tiden kan utvinna den renade verksamma molekylen till sitt läkemedel. Innan denna separationsmetod kan köras kontinuerligt i stor skala behöver den först utvecklas och testas i en laborativ skala. Till detta har det blivit allt vanligare att använda sig av digitala tvillingar: digitala analoger till fysiska processer. Med hjälp av en digital tvilling kan utvecklingen i första hand ske digitalt, innan stora förändringar görs till den faktiska jonbyteskromatografi-processen.
Stora mängder buffert kan förbrukas när jonbyteskromatografi körs kontinuerligt. Därför är det viktigt att se till att den alltid finns tillgänglig i de flaskor där den tas i från. För ett system med kontinuerlig jonbyteskromatografi finns alltså ett behov för en kontinuerlig buffertbeställning och leverans.
I denna avhandling utvecklas ett dynamiskt kösystem för att fylla på buffertar i en digital tvilling till en småskalig process för kontinuerlig jonbyteskromatografi. Information om hur snabbt och hur länge olika flaskor med buffert ska tömmas används för att kontinuerligt räkna ut hur mycket buffert som behöver beställas och hur lång tid det tar innan de töms. Varje beställning prioriteras baserat på tiden tills bufferten kommer ta slut och hur mycket som behöver beställas. Beställningar läggs sedan i en kö och skickas till en buffertberedare. Kön omorganiseras kontinuerligt för att den mest prioriterade bufferten ska hamna högst upp i kön. Det utvecklade systemet testades med två fall: en maskin som förbrukade två buffertar och två maskiner som förbrukade tre buffertar var. Resultaten visar att systemet presterar väl i det första fallet och lyckas hålla en stadig volym i flaskorna. Dock har det svårare att hålla flaskorna fulla i det andra fallet och de börjar bli tomma efter ungefär 12 timmar av kontinuerlig körning. Vidare utveckling behövs för att möjliggöra robustare kontinuerlig leverans av buffert. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9216354
- author
- Holmqvist, Fredrik LU
- supervisor
- organization
- course
- KETM05 20252
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- HPLC, IEX, Digital Twin, Continuous Down-Stream Process, Buffer Management, Chemical Engineering
- language
- English
- id
- 9216354
- date added to LUP
- 2026-01-07 09:47:08
- date last changed
- 2026-01-07 09:47:08
@misc{9216354,
abstract = {{This thesis addresses a challenge in continuous high performance liquid chromatography (HPLC): ensuring reliable, timely supply of buffer solutions. As biopharmaceutical production increasingly moves toward continuous operation, buffer management becomes critical, since chromatography requires multiple buffer types with precise compositions for washing, elution, equilibration, and regeneration. In an integrated small-scale continuous DSP operated through Orbit, a custom Python-based control platform, a digital Buffer Management System (BMS) exists, managing buffer mixing and refill order placement. Order placement occurs only when crossing a minimum volume threshold and the queue is simply first-in first-out, which creates high risks of delayed refills, idle systems, and buffer depletion, particularly when multiple clients operate simultaneously.
The aim of this project is to design and implement a more dynamic, predictive buffer-refill scheduling system to the BMS, that can estimate buffer depletion in advance and prioritize refill orders accordingly. This required the development of copies of each HPLC client system in Orbit in order to extract process information without interfering with the real running process.
A new buffer-depletion calculation routine was developed, consisting of three cooperating functions to extract instruction information from the active phases, pick out flask flowrates, integrate consumption over time, and calculate both the time until a minimum threshold is reached and the required refill volume at that moment in time. Instead of issuing refill orders once a low-volume threshold is crossed, clients now continuously provide the engine with depletion data every ten seconds. The engine then sorts all buffer orders, prioritizing shortest time until the threshold is reached and then largest refill volume, before putting them in the queue, which continually reorganizes itself.
The new system was benchmarked against the previous version of the BMS using three scenarios: a single client with alternating buffer consumption from two flasks, the same client with an added step-gradient phase, and two clients consuming three buffers each. In the single client cases, the new BMS maintained buffer levels well above the minimum threshold throughout 32-cycle tests, ensuring more stable buffer volumes than the old system, which often approached very low volumes before receiving a refill. A gradual downward trend in peak refill volumes was observed for the new system, attributed to the fixed dead-time required for buffer preparation.
Performance in the dual-client scenario was more mixed. Although the new BMS slightly extended the time, managing up to 12 hours, before any complete depletion occurred, high combined buffer demand, small order placements and long preparation times for certain buffer types created bottlenecks and cascading delays. The scheduler struggled when too many buffers were consumed simultaneously or when preparation times varied drastically between buffer recipes. In some cases, the new system outperformed the low volume threshold-based design, but in others the old system was superior.
The thesis concludes that the dynamic, buffer-refill scheduling system markedly improves robustness and maintains more even buffer volumes in buffer management for single client continuous DSP. It is uncertain whether it outperforms the old system during more demanding multi-client operation. The work highlights the need for enhanced prediction models (e.g., more accurate interpolation and preparation-time-aware prioritization), handling of non-time-based phases, and better support for smooth buffer gradients. Future development should also integrate the digital BMS with the physical DSP to validate its real-world performance and refine the order- and refill system further.}},
author = {{Holmqvist, Fredrik}},
language = {{eng}},
note = {{Student Paper}},
title = {{A Buffer-Refill Scheduling System for a Digital Small Scale Continuous Bioprocess}},
year = {{2025}},
}