This thesis project was proposed by Caseificio Elda s.r.l with the aim of understanding the origin of the variability of the pH value of the product known as ‘spreadable dairy cheese’. The product in the various production batches shows a pH between 4,40 and 4,85: this range is considered too wide as it affects the quality of the product making standardisation difficult from a sensory point of view and limiting shelf-life. pH analyses were carried out on the finished product and the raw materials, with the aim of assessing which of the raw materials used has the greatest influence on the pH of the finished product. For the study of the problem, a pilot plant was used to reproduce the spreadable dairy production process on a small scale. During the period between February 2024 and July 2024, 197 samples were produced and analysed; this allowed discrimination of which of the raw materials used has the greatest influence on pH. Once a largely sufficient amount of data had been obtained, two statistical models were developed to predict the pH of the finished product by analysing the pH of the raw materials alone. Predictive models, such as the general linear model (GLM) and polynomial linear regression (RLP), simplify the company's production management and lay the foundation for solving the instability problem. Furthermore, through an in-house consumer test, the substitution of the ingredient citric acid with lactic acid for the production of spreadable dairy products was evaluated in order to test whether lactic acid was perceived to be less acidic than citric acid. Regarding the statistical models, they proved reliable in predicting the pH values of the finished product. The RLP has an R2 up to 0.9342 on the test datasets, while the GLM has an R2 up to 0.9209 on the test datasets, and for both, the average error was only 0.02 pH points that is an average % error of 0.53% (RLP) and 0.45% (GLM). By predicting the pH of the spreadable dairy product using predictive models, the company will be able to handle batches outside the pH range of 4,40 – 4,60, for example by adjusting the pH during production or by allocating them to customers who do not need a product with a shelf-life calculated in the range of 4,40 – 4,60. Finally, the results of the consumer test showed that there was no statistically significant difference in the perception of acidity between the sample modified with lactic acid and the standard sample with citric acid, therefore, it will not to be necessary to modify the standard recipe.
Il presente progetto di tesi è stato proposto dal caseificio Elda s.r.l con l’obiettivo di comprendere l’origine della variabilità del valore di pH del prodotto denominato “latticino spalmabile”. Il prodotto nei diversi lotti produttivi mostra un pH compreso tra 4.40 e 4.85: questo range è considerato troppo ampio in quanto incide nella qualità del prodotto rendendo difficoltosa la standardizzazione dal punto di vista sensoriale e limitando la shelf-life. Sono state eseguite analisi di pH sul prodotto finito e sulle materie prime, con lo scopo di valutare quali, tra queste ultime utilizzate, influenzi maggiormente il pH del prodotto finito. Per lo studio del problema è stato utilizzato un impianto pilota che permettesse di riprodurre su scala ridotta il processo di produzione del latticino spalmabile. Nel periodo compreso tra febbraio 2024 e luglio 2024 sono stati prodotti ed analizzati 197 campioni; ciò ha permesso di discriminare quale tra le materie prime utilizzate influenza maggiormente il pH. Una volta ottenuto un numero ampiamente sufficiente di dati, sono stati sviluppati due modelli statistici che permettessero di predire il pH del prodotto finito attraverso la sola analisi del pH delle materie prime. I modelli predittivi, come il general linear model (GLM) e la regressione lineare polinomiale (RLP), semplificano la gestione della produzione da parte dell’azienda e pongono le basi per la soluzione al problema di instabilità. Inoltre, attraverso un consumer test all’interno dell’azienda, è stata valutata la sostituzione dell’ingrediente acido citrico con l’acido lattico per la produzione latticino spalmabile allo scopo di verificare se l’acido lattico risultasse percepito meno acido rispetto all’acido citrico. Per quanto riguarda i modelli statistici, sono risultati affidabili nella predizione dei valori di pH del prodotto finito. L’RLP ha dato un R2 che sui dataset di test arriva fino a 0,9342, mentre con il GLM si è ottenuto un R2 che, sui dataset di test, arriva fino a 0,9209, e per entrambi, l’errore medio è stato di soli 0,02 punti di pH ed equivale ad un errore medio % dello 0,53% (RLP) e 0,45% (GLM). L’azienda prevedendo il pH del latticino spalmabile attraverso i modelli predittivi, potrà gestire i lotti fuori dall’intervallo di pH di 4,40 – 4,60, per esempio aggiustandone il pH durante la produzione o destinandoli a clienti che non necessitano di un prodotto con la shelf-life calcolata nell’intervallo 4,40 – 4,60. Infine, i risultati del consumer test hanno dimostrato che non vi è differenza statisticamente significativa per quanto riguarda la percezione dell’acidità tra il campione modificato con acido lattico ed il campione standard con acido citrico, quindi, non sarà necessario modificare la ricetta standard.
CASO STUDIO: MODELLIZZAZIONE DEL pH DI UN LATTICINO SPALMABILE E ANALISI SENSORIALE
MARCHI, EMANUELE
2023/2024
Abstract
This thesis project was proposed by Caseificio Elda s.r.l with the aim of understanding the origin of the variability of the pH value of the product known as ‘spreadable dairy cheese’. The product in the various production batches shows a pH between 4,40 and 4,85: this range is considered too wide as it affects the quality of the product making standardisation difficult from a sensory point of view and limiting shelf-life. pH analyses were carried out on the finished product and the raw materials, with the aim of assessing which of the raw materials used has the greatest influence on the pH of the finished product. For the study of the problem, a pilot plant was used to reproduce the spreadable dairy production process on a small scale. During the period between February 2024 and July 2024, 197 samples were produced and analysed; this allowed discrimination of which of the raw materials used has the greatest influence on pH. Once a largely sufficient amount of data had been obtained, two statistical models were developed to predict the pH of the finished product by analysing the pH of the raw materials alone. Predictive models, such as the general linear model (GLM) and polynomial linear regression (RLP), simplify the company's production management and lay the foundation for solving the instability problem. Furthermore, through an in-house consumer test, the substitution of the ingredient citric acid with lactic acid for the production of spreadable dairy products was evaluated in order to test whether lactic acid was perceived to be less acidic than citric acid. Regarding the statistical models, they proved reliable in predicting the pH values of the finished product. The RLP has an R2 up to 0.9342 on the test datasets, while the GLM has an R2 up to 0.9209 on the test datasets, and for both, the average error was only 0.02 pH points that is an average % error of 0.53% (RLP) and 0.45% (GLM). By predicting the pH of the spreadable dairy product using predictive models, the company will be able to handle batches outside the pH range of 4,40 – 4,60, for example by adjusting the pH during production or by allocating them to customers who do not need a product with a shelf-life calculated in the range of 4,40 – 4,60. Finally, the results of the consumer test showed that there was no statistically significant difference in the perception of acidity between the sample modified with lactic acid and the standard sample with citric acid, therefore, it will not to be necessary to modify the standard recipe.File | Dimensione | Formato | |
---|---|---|---|
MARCHI_EMANUELE.pdf
accesso riservato
Dimensione
1.33 MB
Formato
Adobe PDF
|
1.33 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/70706