An expert system for the interpretation of pyrolysis mass spectra of condensation polymers

• Published in: Analytica chimica acta 1998-02, Vol.359 (1), p.213-225
• Main Authors: Georgakopoulos, Costas G., Statheropoulos, Miltiades C., Montaudo, Giorgio
• Format: Article
• Language: English
• Subjects: Analytical chemistry; Applied sciences; Chemistry; Exact sciences and technology; Organic polymers; Physicochemistry of polymers; Properties and characterization; Spectrometric and optical methods; Structure, morphology and analysis
• ISSN: 0003-2670; 1873-4324
• DOI: 10.1016/S0003-2670(97)00632-6

The PROLOG programming language was used for the construction of the Expert System (ES) HEPHESTUS used in the interpretation of pyrolysis mass spectra of condensation polymers, such as polyamides, polycarbonates, polyethers, polyesters, polyureas, polyurethanes, polyimides, polysulphides, polysulfones, polyschiff bases, polysiloxanes and polyphosphagenes. The ES can be applied to Pyrolysis Mass Spectrometry (PyMS) spectra of polymers using Electron Impact, Chemical Ionisation or Desorption Chemical Ionisation modes. The sample is introduced and pyrolysed directly in the source of the mass spectrometer. The ES knowledge base is organised in a tree form, based on polymer structures. The inference mechanism uses empirical algorithms and the autocorrelation algorithm to determine important information present in the unknown polymer spectrum. Possible solutions are proposed using the Certainty Factor model and based on the rules of the knowledge base. The autocorrelation algorithm, used to determine the repeating unit, gave 57.3% of successes and 42.7% not best choices. Among the not best choices the 23.3% were failures. For the same purpose, the empirical algorithm gave 68.5% successes and 31.5% not best choices (among them 35.7% were failures). User interaction improves the ES's performance to propose the correct polymer (94.4% of the trials were the best choices).