CHALLENGING THE INSPECT TECHNOLOGY
INSPECT is an innovative tool for online and non-destructive analysis along the manufacturing line. It is intended to be installed in production plants performing real-time measurements that are currently carried out only on representative samples in the lab and it identifies chemical compositions of the examined target.
The device is equipped with light-sources of various energy and intensity, which illuminate the target. As a result the reflected characteristic spectrum describes the intrinsic properties of the analysed material. These information are processed by an AI-based algorithm capable of instantaneously compare the data with a multi-threshold pattern recognition gating prior to store the information in a classification database. Moreover, deep-learning techniques allow to calibrate the device and to extend its functionalities.
You can find more about this technology here
DE.TEC.TOR. is currently performing a feasibility study with the aim to identify three different organic molecules in tomato pulp.
Considering the very heterogeneous sample matrix, the study is considered as particularly challenging for the contaminant concentrations ranging between 1 ppm and 0.01 ppm.
1) Sample probing
To extract the material spectroscopic peculiarities, an optical fiber guides the radiation provided by a light-source and wavelength-specific silicon sensors detect the signal resulting from the interaction with the sample under investigation.
2) Signal extraction
We adopt our patented technique combining NIR and X-ray fluorescence to classify the material thanks to its peculiar reaction to the radiation stimuli expressed in overtones and combination bands of fundamental molecular vibrations and emission of characteristic “secondary” (or fluorescent) X-rays.
The material spectrum is digitised and the information is stored for data analysis purpose.
3) Data elaboration
Various data manipulation technique are adopted, including:
- Data pre-processing (normalisation, zeroing…)
- Spectrum region selection (literature indication about functional group spectral wavelength, gross correlation lack with reference sample…)
- Pattern recognition (PCA, first and second derivative significative trends…)
4) Data-analist-guided contaminant detection
Being able to detect all the three organic molecules in their different concentrations, would successfully complete the feasibility study.
From this point start the complete system-prototype development consisting essentially in tw0 macro-steps:
5) Development of the Machine Learning algorithm
6) AI-guided online autonomous classification
Con la presente si informa che i dati raccolti verranno esclusivamente utilizzati all’interno della banca dati di DE.TEC.TOR.. La persona autorizza la DE.TEC.TOR. ad usare i suoi dati per inviare eventualmente materiale informativo riguardante le attività o qualsiasi iniziativa dI DE.TEC.TOR.. I dati non saranno per nessuna ragione divulgati o ceduti a terzi senza una Sua esplicita autorizzazione. I dati personali che Vi riguardano non saranno utilizzati a fini di invio di materiale pubblicitario commericiale o di vendita diretta o per il compimento di ricerche di mercato o di comunicazione commerciale. In ogni momento, ai sensi dell’art.7 del D.lgs.196/2003 e del Regolamento Europeo 2016/679, potrà consultare, modificare o chiedere la cancellazione di tali dati scrivendo a: DE.TEC.TOR. – Lungo Dora Voghera 36/A, 10153 Torino – Italy