A group of JP-5 fuels, which had passed all specification testing, were causing catastrophic turbine engine failures. Laboratory examinations of these fuels did not reveal any obvious compositional differences responsible for the unacceptable performance. Since fuel screening with full scale combustor tests are expensive and time-consuming, we investigated the use of chemometric analysis to predict fuel suitability from capillary gas chromatography. From a small group of acceptable and problem fuels, a correlation model was developed that successfully predicted fuel suitability, as long as the overall composition of the fuels was similar. To increase sensitivity and allow correlations of minor compositional changes, data preprocessing methods were developed to minimize instrumental errors and to focus the analysis on subtle compositional differences that were responsible for the unacceptable combustion performance.