A mathematical method that smooths noisy process data by fitting successive subsets of data points with local low-degree polynomials. This algorithm is highly valued in thermal process monitoring because it preserves signal trends and peak shapes better than simple averaging.