This paper presents a new approach in fuzzy mental workload modeling by using experton concept. The mental workload is represented by a Load-Experton (LE), a multidimensional aggregating tool, calculated with subjects' judgments as a confidence intervals. Compared to averaging operator, LE has the advantage to keep information disorder. So, this instrument is subjected to two treatments. First, a fuzziness analysis using probabilistic and non-probabilistic entropy. Second, overall mental workload score is calculated using maximum, minimum, and averaging operators. A correlation analysis between mental workload dimensions and overall scores or between these scores is carried out using affinity concept and Galois lattice. The approach is applied to supervisory tasks in a petroleum manufactory.