Advanced

Development of a FT-NIR model for analysing water content in rapeseed

Nilsson, Ludvig LU (2015) KLG820 20142
Food Technology and Nutrition
Abstract
A FT-NIR (Fourier-Transform Near Infrared) model for moisture content has been developed for whole seeds of Swedish rapeseed (Brassica napus) by the conditions given during the harvest of 2014. Compared to Internal standardized (ISO) method 665, the final model showed very small differences in terms of repeatability and measurement uncertainty even though the measurement time was four times faster calculated on a daily basis. If only one sample is analysed the FT-NIR is 430 times faster than ISO 665. Compared to the Swedish national Near Infrared Transmission (NIT) model used in official testing of rapeseed varieties, both the measurement quality of obtained values and time of analysis were twice as good for the developed FT-NIR model.... (More)
A FT-NIR (Fourier-Transform Near Infrared) model for moisture content has been developed for whole seeds of Swedish rapeseed (Brassica napus) by the conditions given during the harvest of 2014. Compared to Internal standardized (ISO) method 665, the final model showed very small differences in terms of repeatability and measurement uncertainty even though the measurement time was four times faster calculated on a daily basis. If only one sample is analysed the FT-NIR is 430 times faster than ISO 665. Compared to the Swedish national Near Infrared Transmission (NIT) model used in official testing of rapeseed varieties, both the measurement quality of obtained values and time of analysis were twice as good for the developed FT-NIR model. From a Lantmännen perspective implementation if this model will save over 30 working days on a yearly basis if the same number of samples are analysed as in 2014. Faster analysis time will also generate several important advantages compared to current practice. These include almost real-time quality assurance of on-line harvesters NIR according to ISO 12099. A shorter time between sampling and analysis will also minimise possible change in sample composition not uncommon in samples of rapeseed.

The model was calibrated and validated using samples from different locations throughout the rapeseed growing region of Sweden. Resulting model generated a SEP (Standard Error of Prediction)=0.166 with a large sample cup (11 g sample) and SEP=0.260 with the smaller autosampler (2 g sample). For comparison, a typical SEP in the literature is 0.15, 0.37 and 0.5. In addition, the external Swedish national NIT models (ESNNIT) have a SEP between 0.33 and 0.63 with a mean of 0.40 between the years 2006 to 2014. The measurement uncertainty of ISO 665 is sample and laboratory dependent, but expressed as internal reproducibility for the laboratory in Svalöv, 0.3 typically corresponds to a 95% confidence interval. The corresponding value for the FT-NIR model expressed as 1.96*SEP (95% confidence internal) was found to be 0.33. The findings of this thesis clearly indicates that a change from ISO 665 to the FT-NIR methodology developed can be made without compromising with the risk of an increased measurement uncertainty. (Less)
Popular Abstract (Swedish)
Lantmännen Lantbruk har i ett projekt tillsammans med Lunds Tekniska Högskola utvecklat en modell för analys av vattenhalt i svenskodlad raps med tekniken Fourier Transform Nära Infraröd (FT-NIR). Kvaliteten i de mätvärden som modellen levererar är mycket lik kvaliteten i den metodik (referensmetodik) som försökte efterliknas när modellen skapades. Även ur ett internationellt perspektiv presterar utvecklad metodik i topp vid jämförelser av nyckeltal vid NIR-analys av raps.

Det finns olika instrument som använder sig av nära infraröd (NIR) teknik för att analysera vattenhalten i raps. Utvecklad metodik jämfördes parallellt med samma prov, med två av de allra vanligaste förekommande NIR modellerna i Sverige. Resultaten visar att... (More)
Lantmännen Lantbruk har i ett projekt tillsammans med Lunds Tekniska Högskola utvecklat en modell för analys av vattenhalt i svenskodlad raps med tekniken Fourier Transform Nära Infraröd (FT-NIR). Kvaliteten i de mätvärden som modellen levererar är mycket lik kvaliteten i den metodik (referensmetodik) som försökte efterliknas när modellen skapades. Även ur ett internationellt perspektiv presterar utvecklad metodik i topp vid jämförelser av nyckeltal vid NIR-analys av raps.

Det finns olika instrument som använder sig av nära infraröd (NIR) teknik för att analysera vattenhalten i raps. Utvecklad metodik jämfördes parallellt med samma prov, med två av de allra vanligaste förekommande NIR modellerna i Sverige. Resultaten visar att utvecklad metodik gav mest övertygande resultat, den är både snabbare och mer specifik. Det går dock inte med säkerhet att avgöra varför utvecklad modell verkar prestera bättre. Eventuella skillnader i referensmetodik vid framtagande av andra NIR-modeller, skillnad i databearbetning samt skillnader i använd teknik tillhör några av de vanligaste förekommande orsakerna som brukar nämnas. Med hög sannolikhet är det en kombination av ovan faktorer som möjliggjort framsteg.

Fördelar med utvecklad metodik jämfört med referensmetodik (som fortfarande används mycket i Sverige) är många. På daglig basis kan nämligen fyra gånger fler prov analyseras, och om bara ett prov ska analyseras, gör den nya tekniken det 430 gånger snabbare jämfört med studerat laboratoriums nuvarande kapacitet. Möjligheterna är många medan svagheterna är få. Att mätosäkerheten dessutom krymper ner mot vad som kan förväntas med referensmetodik gör att implementering av utvecklad modell kan utföras med en mycket liten ökning i risk jämfört med nuvarande rutiner inom Lantmännen. En av många andra fördelar med byte av metodik inkluderar möjligheter till analysarbete enligt ”lean production” – mer värde för mindre arbete. (Less)
Please use this url to cite or link to this publication:
author
Nilsson, Ludvig LU
supervisor
organization
alternative title
Utveckling av en FT-NIR modell för vattenhaltsanalys i svenskodlad raps
course
KLG820 20142
year
type
H2 - Master's Degree (Two Years)
subject
keywords
FT-NIR, water content, brassica napus, rapeseed, raps
language
English
id
5038601
date added to LUP
2015-02-05 15:55:19
date last changed
2015-02-05 15:55:19
@misc{5038601,
  abstract     = {A FT-NIR (Fourier-Transform Near Infrared) model for moisture content has been developed for whole seeds of Swedish rapeseed (Brassica napus) by the conditions given during the harvest of 2014. Compared to Internal standardized (ISO) method 665, the final model showed very small differences in terms of repeatability and measurement uncertainty even though the measurement time was four times faster calculated on a daily basis. If only one sample is analysed the FT-NIR is 430 times faster than ISO 665. Compared to the Swedish national Near Infrared Transmission (NIT) model used in official testing of rapeseed varieties, both the measurement quality of obtained values and time of analysis were twice as good for the developed FT-NIR model. From a Lantmännen perspective implementation if this model will save over 30 working days on a yearly basis if the same number of samples are analysed as in 2014. Faster analysis time will also generate several important advantages compared to current practice. These include almost real-time quality assurance of on-line harvesters NIR according to ISO 12099. A shorter time between sampling and analysis will also minimise possible change in sample composition not uncommon in samples of rapeseed. 

The model was calibrated and validated using samples from different locations throughout the rapeseed growing region of Sweden. Resulting model generated a SEP (Standard Error of Prediction)=0.166 with a large sample cup (11 g sample) and SEP=0.260 with the smaller autosampler (2 g sample). For comparison, a typical SEP in the literature is 0.15, 0.37 and 0.5. In addition, the external Swedish national NIT models (ESNNIT) have a SEP between 0.33 and 0.63 with a mean of 0.40 between the years 2006 to 2014. The measurement uncertainty of ISO 665 is sample and laboratory dependent, but expressed as internal reproducibility for the laboratory in Svalöv, 0.3 typically corresponds to a 95% confidence interval. The corresponding value for the FT-NIR model expressed as 1.96*SEP (95% confidence internal) was found to be 0.33. The findings of this thesis clearly indicates that a change from ISO 665 to the FT-NIR methodology developed can be made without compromising with the risk of an increased measurement uncertainty.},
  author       = {Nilsson, Ludvig},
  keyword      = {FT-NIR,water content,brassica napus,rapeseed,raps},
  language     = {eng},
  note         = {Student Paper},
  title        = {Development of a FT-NIR model for analysing water content in rapeseed},
  year         = {2015},
}