Back to search

Cornell University

IMPROVING THE ACCURACY OF MID-INFRARED DETERMINATION OF FAT, PROTEIN, LACTOSE, AND UREA IN MILK

Abstract

dc:description.abstract

Mid-infrared (MIR) instruments for milk analysis are used throughout the dairy industry for payment testing, herd management, and product manufacturing, as they provide a rapid/reliable method for measuring milk components. Milk MIR analysis is a secondary method requiring statistical prediction models and calibration for each measured component. Two prediction model approaches exist for measuring milk components from milk MIR spectra: basic filter models with intercorrection factors (IF) (for major milk components) and partial least squares (PLS) models (for major and minor milk components). Fat, protein, and lactose can be measured by basic filter models, while milk urea nitrogen (MUN), a metric of interest in herd management, is measured by PLS models. The current method for determination of IF (instrument-specific coefficients in basic filter models that account for background component variation) lacks robustness. Additionally, current MUN PLS models are often influenced by background matrix components, and, to achieve a complete system for MUN MIR measurement, a chemical reference method (for reference MUN values for calibration) with interlaboratory results approved by the Association of Official Analytical Chemists (AOAC) is needed. Therefore, this thesis had the following objectives: (i) develop a method to determine/optimize IF for MIR basic filter models using modified milk samples; (ii) conduct a collaborative study for AOAC-approval of an enzymatic/spectrophotometric reference method for MUN; and (iii) develop an improved PLS model for MUN MIR measurement. For objective (i), a multiple linear regression model followed by optimization was developed to calculate IF. IF were determined for nine MIR instruments and resulted with smaller magnitudes and lower variability among instruments than those reported previously using the traditional method. For objective (ii), a collaborative study was conducted with 10 laboratories and the enzymatic/spectrophotometric method for MUN showed improved repeatability and reproducibility (RSDR<1%) compared to other reported methods for milk urea. Finally, for objective (iii), a PLS model for MUN was developed using a variety of farm milks, individual-cow milks, and modified milks to train the model. The developed model showed good prediction accuracy at a wide range of MUN concentrations during cross-validation (RPD=5.3). External model validation is now required as the next step for implementation of the new PLS model in the industry.

Degree

thesis:*
Name thesis:degree_name
Ph. D., Food Science and Technology
Level thesis:degree_level
Doctor of Philosophy
Discipline thesis:degree_discipline
Food Science and Technology
Grantor
Cornell University
Year dc:date.issued
2023

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Portnoy, Matilde
Chair dc:contributor.chair
  • Barbano, David
Committee members dc:contributor.committeemember
  • Van Amburgh, Michael
  • Wolczanski, Peter

Subjects

dc:subject × 4

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Bibid
16176677
Dc Identifier Other
ProQuest Submission ID: 13476
ProQuest Publication ID: 30249712

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Portnoy, Matilde. IMPROVING THE ACCURACY OF MID-INFRARED DETERMINATION OF FAT, PROTEIN, LACTOSE, AND UREA IN MILK. Doctor of Philosophy thesis, Cornell University, 2023. https://hdl.handle.net/1813/114118