Each autonomous technological process requires an early fault detection system as an integral part of its infrastructure. Complex fault diagnostics system must be able not only to quickly detect any unwanted deviation from nominal function, but also to make a competent estimate of the type and source of possible failure. The article presents the fault diagnostics system based on Markov model. It uses an innovative dynamically oriented classifier of fault states. The other enhancement represents the regression vector used in the diagnostics system which combines directly measured process data with their components obtained using Empirical Mode Decomposition (EMD).