Client: A company producing an innovative fuel additive.
Objective: The client wished to reliably and accurately determine the concentration of their fuel additive, which was in the range 0-300 parts per billion (ppb), based on a SERS analysis. Comparing peak ratios, they had only been able to achieve an average accuracy of 50 ppb and worst-case errors of 150 ppb.
Our Solution: Using Analyze IQ Lab on a small set (<50) of reference SERS spectra supplied by the client, we experimented with various pre-processing and analysis methods and found that best results were obtained using First Order Derivative preprocessing and a Support Vector Machine with Linear Kernel. We also identified that two of their spectra were mislabelled.
Result: We delivered an Analyze IQ chemometric model which enabled the client to determine the concentration to within 7 ppb, a 7-fold improvement on the client’s own method. This made the overall analysis of fuels feasible.