This article proposes a framework for smoothing of and Kα doublet elimination from X-ray diffraction scans in an automated fashion. Smoothing is performed using penalized likelihood estimation, while for doublet elimination the model is extended using the composite link model.