Detection of glass in RDF using NIR spectroscopy

Detta är en Master-uppsats från Mälardalens högskola/Akademin för ekonomi, samhälle och teknik

Sammanfattning: Purpose of this study was to investigate the possibilities of using Near-infrared (NIR) spectroscopy to detect glass in refuse derived fuel (RDF) as well as what on-line data of glass content could be used for in terms of boiler operation and performance determination. Sample configurations were done with dried RDF (to prevent mass loss due to moisture and spectroscopic disturbance) and increasing concentrations of colored soda-lime glass, total of 100 samples. Glass was randomly scattered among the RDF by shaking the added glass and RDF in a bucket to generate representative samples of real life conditions. NIR-spectra acquisition was done between 12000 and 4000 cm-1, at 8 cm-1 resolution and average of 32 scans. The determination of boiler performance was done in accordance to Swedish standards for acceptance testing and heat loss due to glass was treated as slag. Resulting performance calculations showed boiler efficiency via indirect method matching efficiency calculated via direct method (deviating at maximum 2 %) which validates the summarized losses (including due to glass). The heat loss due to glass was calculated to 0,068 MW/%glass, which equated to average of 0,16 MW for 2,37 % glass. Total heat loss was amounted to an average of 11,53 MW. The developed models were not satisfactory in their quality of regression prediction. Although some had, through pre-processing, good development of explained variance at increasing factors, but still had a “Not Applicable” coefficient of determination by regression prediction. The poor quality of models can be explained by poor glass detection (poor representation) of the spectroscopic instrument due to a combination of glass being randomly scattered in the background material and sometimes covered by RDF as well as that the NIR-spectroscopy light beam only hits a small area. By increasing the number of samples upwards 300 -500, the effect of random scatter of glass can be mitigated and acceptable models could be acquired.

  HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)