Manual assessment of articulation errors by Speech Language Pathologists (SLP) is a complex process requiring assimilation of various information regarding the patient. Automatic assessment of articulation errors can assist an SLP in maximizing the efficiency of therapy. Our work focuses on building an automatic assessment method for articulation errors at phone level and classifying a patient utterance as either correct, substitution, omission, distortion or addition (CSODA). Identification of the error at phone level is essential to provide the patient with actionable feedback for correction. The objective of our work is to be able to improve the recognition ability of the ASR to identify articulation errors through improved classification of consonants. In this paper, we propose an automatic speech recognition (ASR) based method to identify substitution errors for consonants, using a rule based language model (LM) as well as tuning of acoustic models (AM) for consonants under consideration. As the first step, we evaluated the proposed method using normal speech. The changes to AM shows a significant improvement in overall recognition, by 17.29% for normal speech.