Quantification of oxygenates, sulphides, thiols and permanent gases in propylene. A multiple linear regression model to predict the loss of efficiency in polypropylene production on an industrial scale

Abstract

This research developed a new methodology using a gas chromatograph, a pulsed discharge helium ionization detector (PDHID), and two mass spectrometric (MSD) detectors. The detectors worked simultaneously, using 8 columns and 7 valves. This new proposal for simultaneous analysis, with a single injection and analysis time of 36 min, allowed the quantification of 10 oxygenated compounds (alcohols, ketones and carboxylic acids), 3 permanent gases, 3 sulphides and 4 thiols, which are aggressive inhibitors of the Ziegler-Natta catalytic systems. The RSD (n=6) for repeatability of the peak area of the 20 compounds analyzed, and the retention time were less than 0.59 and 0.23% respectively. The RSD (n=6) for intermediate precision for the peak area was less than 0.85% and, for the retention time, less than 0.35%. 95% of the inhibitors analyzed showed relative errors inter and intra-day less than 3%. The inhibitors detected and quantified were: formic acid (2 to 45.32 ppm), acetic acid (1 to 25.32 ppm), acetone (5 to 72.67 ppm), methanol (1 to 39.93 ppm), isopropyl alcohol (2 to 74.88 ppm), ethanol (0.1 to 57.51 ppm), 1-propanol (0.1 to 92.36 ppm), 1-butanol (5 to 92.36 ppm), 2-butanol (5 to 95.15 ppm), tert-butanol (5 to 90.22 ppm), CO2 (0.5 to 5.0 ppm), CO (0.002 to 5.049 ppm), O2 (0.5 to 6.5 ppm), CH3(CH2)3SH (0.025 to 2.238 ppm), CH3CH2SH (0.025 to 1.595 ppm), COS (0.025 to 1.477 ppm), CH3SH (0.025 to 1.223 ppm), CH3(CH2)2SH (0.025 to 1.880 ppm), CS2 (0.025 to 1.929 ppm), H2S (0.025 to 0.847 ppm) and tert-butylmercaptan (0.025 to 2.283 ppm). These compounds generated reductions of between 5 and 20% in polypropylene production, representing losses of several million dollars. Therefore, a multilinear regression model was developed to predict this percentage of production loss in a fluidized bed reactor, based on the quantified inhibitors. The model has a correlation coefficient of 0.91 and a standard deviation of 1.12, allowing comparisons between actual and predicted experimental values. P values are less than 0.05, indicating that each inhibitor has a statistically significant effect on the model.