Prediktiv modell för att minska kasserat material vid tablettslagning
(2026) KLGM16 20261Pharmaceutical Technology (master)
- Abstract
- During start-up of tablet presses in pharmaceutical manufacturing, multiple weight adjustments are often required before the tablet weight falls within acceptable limits. This leads to delayed production and a waste of material. The aim of this work was to develop a predictive model to reduce the number of adjustments during start-ups and thereby reduce the amount of rejected material.
A linear regression model was developed based on the relationship between changes in punchy separation and resulting changes in tablet weight. The model was evaluated during 61 start-ups, of which 19 were performed with starting weights deliberately manipulated up to 80 mg from the weight limit to test the model’s robustness. A residual analysis was... (More) - During start-up of tablet presses in pharmaceutical manufacturing, multiple weight adjustments are often required before the tablet weight falls within acceptable limits. This leads to delayed production and a waste of material. The aim of this work was to develop a predictive model to reduce the number of adjustments during start-ups and thereby reduce the amount of rejected material.
A linear regression model was developed based on the relationship between changes in punchy separation and resulting changes in tablet weight. The model was evaluated during 61 start-ups, of which 19 were performed with starting weights deliberately manipulated up to 80 mg from the weight limit to test the model’s robustness. A residual analysis was conducted to identify underlying factors affecting the model’s predictive performance.
The results showed that the model accurately describes the relationship between punch separation and tablet weight, and that the model coefficients are representative for a larger sampling period. Implementation of the model would reduce the average number of adjustments from 2,7 to 1,05 per start-up, corresponding to a reduction of 61% and saved material equivalent to approximately 1,9 million tablets per year. The residual analysis identified the magnitude of step height change as the dominant factor behind the model’s limitations for starting weights far from the weight limit, while raw material, granulation machine, tablet press, compression force and rotor speed showed marginal or no influence on the model’s performance.
It is concluded that the developed model can standardize and streamline the start-up process, reduce operator dependency and significantly reduce material waste. The methodology is transferable to other pharmaceutical products, and future work may focus on improving the model’s performance for starting weights far from the weight limit, as well as developing a corresponding model for hardness adjustment. (Less) - Popular Abstract (Swedish)
- Under uppstarten av tablettpressar i läkemedelsproduktion, krävs ofta flera viktjusteringar för att uppnå de krav som ställs på tabletternas kvalitet. Detta leder till utdragen och oförutsägbar ledtid vid uppstart och framför allt, kasserat material. Syftet med arbetet var att ta fram en prediktiv formel för att reducera mängden justeringar som krävs vid uppstart av tablettpressar och därmed reducera mängden material som kasseras.
En linjär regressionsmodell togs fram baserat på korrelationen mellan skillnaden i maskininställningen steghöjd och den resulterande skillnaden i tablettvikten. Modellen testades totalt 61 gånger varav 19 av testerna var på startvikter som medvetet manipulerades till ± 80 mg från viktgränsen för att testa... (More) - Under uppstarten av tablettpressar i läkemedelsproduktion, krävs ofta flera viktjusteringar för att uppnå de krav som ställs på tabletternas kvalitet. Detta leder till utdragen och oförutsägbar ledtid vid uppstart och framför allt, kasserat material. Syftet med arbetet var att ta fram en prediktiv formel för att reducera mängden justeringar som krävs vid uppstart av tablettpressar och därmed reducera mängden material som kasseras.
En linjär regressionsmodell togs fram baserat på korrelationen mellan skillnaden i maskininställningen steghöjd och den resulterande skillnaden i tablettvikten. Modellen testades totalt 61 gånger varav 19 av testerna var på startvikter som medvetet manipulerades till ± 80 mg från viktgränsen för att testa modellens robusthet vid det yttre spannet för startvikter. Därefter gjordes även en residualanalys för att undersöka betydande faktorer på modellens prestanda.
Resultaten visade att modellen, med hög precision, beskrev förhållandet mellan steghöjden och tablettvikten, samt att modellens koefficienter var representativa över en utökad tidsperiod. Implementering av modellen skulle minska antalet justeringar per uppstart från 2,7 till 1,05 justeringar i 91,2% av alla uppstarter. Detta innebär en reduktion med 61% i antalet justeringar vilket motsvarar runt 1,9 miljoner tabletter per år i kasserat material. För resterande 8,8% av uppstarterna, för berörd produkt, där startvikten avviker långt från den godkända viktgränsen, kräver modellen istället två justeringar vilket innebär en reduktion med ~26% för dem uppstarterna. Residualanalysen identifierade storleken på steghöjdsförändringen, och därmed storleken på viktändringen som krävs, som den dominerande faktorn bakom modellens prestanda.
Slutsatserna som kan fastställas är att modellen som utvecklades kunde standardisera uppstartsprocessen och göra den mindre beroende av operatörernas erfarenhet och samtidigt reducera mängden kasserat material. Metodiken bakom framtagandet av formeln kommer även att kunna appliceras på andra läkemedel med liknande variation i startvikten. Framtida förbättringsarbeten omfattar att förbättra modellens prestanda för startvikter långt från viktgränsen samt att ta fram en liknande modell för tabletternas hårdhet. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9236704
- author
- Eliasson, Gustav LU
- supervisor
- organization
- course
- KLGM16 20261
- year
- 2026
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- tablet weight prediction, pharmaceutical manufacturing, process optimization, pharmaceutical formulation
- language
- Swedish
- id
- 9236704
- date added to LUP
- 2026-07-02 09:57:49
- date last changed
- 2026-07-02 09:57:49
@misc{9236704,
abstract = {{During start-up of tablet presses in pharmaceutical manufacturing, multiple weight adjustments are often required before the tablet weight falls within acceptable limits. This leads to delayed production and a waste of material. The aim of this work was to develop a predictive model to reduce the number of adjustments during start-ups and thereby reduce the amount of rejected material.
A linear regression model was developed based on the relationship between changes in punchy separation and resulting changes in tablet weight. The model was evaluated during 61 start-ups, of which 19 were performed with starting weights deliberately manipulated up to 80 mg from the weight limit to test the model’s robustness. A residual analysis was conducted to identify underlying factors affecting the model’s predictive performance.
The results showed that the model accurately describes the relationship between punch separation and tablet weight, and that the model coefficients are representative for a larger sampling period. Implementation of the model would reduce the average number of adjustments from 2,7 to 1,05 per start-up, corresponding to a reduction of 61% and saved material equivalent to approximately 1,9 million tablets per year. The residual analysis identified the magnitude of step height change as the dominant factor behind the model’s limitations for starting weights far from the weight limit, while raw material, granulation machine, tablet press, compression force and rotor speed showed marginal or no influence on the model’s performance.
It is concluded that the developed model can standardize and streamline the start-up process, reduce operator dependency and significantly reduce material waste. The methodology is transferable to other pharmaceutical products, and future work may focus on improving the model’s performance for starting weights far from the weight limit, as well as developing a corresponding model for hardness adjustment.}},
author = {{Eliasson, Gustav}},
language = {{swe}},
note = {{Student Paper}},
title = {{Prediktiv modell för att minska kasserat material vid tablettslagning}},
year = {{2026}},
}