20.500.12556/DKUM-55187 Napovedovanje kemične sestave krme z NIR spektroskopijo Predicting chemical composition of forage using NIR spectroscopy V diplomskem delu smo preučevali uporabnost prenosnega NIR spektrometra za napovedovanje kemične sestave krme. V raziskavo smo vključili 244 vzorcev travnih silaž, mrve in zelene krme, v katerih smo z weendsko analizo določili vsebnost suhe snovi (SS), surovih beljakovin (SB), maščob (SM), vlaknine (SV) in pepela (SP) ter prostornino plinov (ki je pomembna za izračun energijske vrednosti krme). Dobljene kemične analize smo povezali s spektralnimi podatki. Pri razvijanju in validaciji modelov smo testirali vpliv območja valovnih dolžin, uporabe odvodov osnovnega spektra absorbcije ter načina validacije modelov. Kakovost modelov smo ovrednotili z determinacijskim koeficientom (R2CV) in napako navzkrižne validacije (seCV) ter parametrom RPD (razmerje med seCV in standardnim odklonom referenčnih vrednosti). V splošnem smo dobili boljšo kakovost modelov brez uporabe odvodov na celem in NIR spektru. Način validacije ni imel vpliva na rezultate. Zelo dobre rezultate smo dobili za SB (R2CV >0,90, RPD > 3,0) in SV (R2CV > 0,85, RPD>2,5) ter srednje dobre za SM, SP in prostornino plinov (R2CV = 0,65-0,75, RPD = 1,5-2,0). SS nismo uspeli zadovoljivo napovedati. Za primerjavo smo modele razvili še z laboratorijskim spektrometrom. Kakovost modelov pri laboratorijskem inštrumentu je bila večja kot pri prenosnem spektrometru. Razlog za to je delno v slabši natančnosti prenosnega inštrumenta (optična sonda in kabel) in delno v časovni razliki med izvedbo kemijske in spektralne analize pri prenosnem inštrumentu. Upoštevajoč slednje lahko zaključimo, da ima prenosni NIR spektrometer velik potencial za napovedovanje kemične sestave krme, vendar bi bilo za praktične namene potrebno napovedne modele pripraviti ponovno z novimi vzorci krme. The aim of the present study was to assess the ability of portable NIR spectrometer to predict chemical composition of forage. The study comprised 244 samples of grass silage, hay and green grass, in which dry matter, protein, fat, raw fibres, ash and gas volume (which is important to calculate energy value) were analysed using Weende analysis. When developing and validating models different parameters were tested: spectral region, derivatives of original spectrum and the type of validation. The quality of models was assessed through coefficient of determination (R2CV) and standard error of cross-validation (seCV) and RPD parameter (ratio between seCV and standard deviation of reference values). In general, the highest quality of models was obtained without using derivatives on the whole or NIR spectral range. The type of validation did not affected the results. Very good results were obtained for protein (R2CV > 0.90, RPD >3.0) and raw fibres (R2CV >0.85, RPD>2.5) and a little less accurate for fat, ash and gas volume (R2CV =0.65-0.75, RPD =1.5-2.0). Dry matter content could not be satisfactory predicted. Obtained results were compared with those obtained with laboratory spectrometer. The latter gave models of better quality. This could be partly explained by lower accuracy of portable instrument (optical sonda and cable) and party by time span between chemical and spectral analysis in case of portable NIR instrument. Considering the latter, it could be concluded that portable NIR spectrometer has great potential for the prediction of forage composition. However, for practical purposes the models could be prepared again using new forage samples. NIR spektroskopija krma kemična sestava prenosni spektrometer NIR spectroscopy forage chemical composition portable spectrometer true true false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2015-10-05 12:24:25 2015-10-19 13:33:38 2022-06-08 03:11:00 2021-05-13 03:39:51 2015 0 Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 URN:SI:UM:DK:QRI33ASS VS_Stalekar_Alenka_2015.pdf VS_Stalekar_Alenka_2015.pdf 1 9BA045F068CA5448BBF693A3223916F4 ad4a8bca65d4746c528f2c231d34566f7f6a9cc070b36a27ea85479495771e2f 3e3471e3-7c07-11eb-bb7a-00155d0001ca https://dk.um.si/Dokument.php?lang=slv&id=82291 Fakulteta za kmetijstvo in biosistemske vede 0 0 0