Artificial neural networks as a tool for identity confirmation of infrared spectra
Average rating
Cast your vote
You can rate an item by clicking the amount of stars they wish to award to this item.
When enough users have cast their vote on this item, the average rating will also be shown.
Star rating
Your vote was cast
Thank you for your feedback
Thank you for your feedback
Type
ReportLanguage
en
Metadata
Show full item recordTitle
Artificial neural networks as a tool for identity confirmation of infrared spectraTranslated Title
Bruikbaarheid van kunstmatige neurale netwerken voor de identititeitsbevestiging van infraroodspectraPubliekssamenvatting
Onderzoek is uitgevoerd naar de bruikbaarheid van kunstmatige neurale netwerken als criterium voor de identiteitsbevestiging van infrarood-(IR) spectra. Doel van het onderzoek is de identificatie van spectra met hoge ruisniveaus, verkregen met behulp van gecombineeerde gaschromatografie (GC)IR spectrometrie. Neurale netwerken zijn getraind op GC/IR spectra van Clenbuterol, Fluoranthene en Perylene. De resultaten zijn vergeleken met classificatie door middel van "peak matching-" en bibliotheekzoekprocedures. Peak matching bleek de meest betrouwbare methode voor de identificatie van sterk gelijkende spectra. Voorwaarde is echter een laag ruisniveau en een hoge spectrale resolutie. Dit geldt eveneens voor bibliotheekzoekmethoden. Neurale netwerken bleken minder gevoelig voor ruis en daarom meer geschikt voor bevestiging van de identiteit van IR-spectra van sporenhoeveelheden.The utility of artificial neural networks (ANN) as a tool for confirmation of the identity of infrared (IR) spectra has been investigated. The main goal of the study is the identification of spectra with relatively high noise levels, obtained from gas chromatography combined with IR spectrometric detection. Networks were trained for GC/IR spectra of Clenbuterol, Fluoranthene and Perylene as representatives of compounds for which identification in real world samples is demanded occasionally. Results have been compared with classification by peak matching and library search methods. Peak matching appears to be the most discriminative method to distinguish between closely resembling spectra, but only in case of high signal-to-noise ratio and resolution. Similar conclusions are drawn for library search identification. ANN-models are less sensitive to spectral noise and hence most suited to be used for confirmation and identification of spectra obtained intrace analysis.
Sponsors
RIVMCollections