Various embodiments of a method for monitoring a machine condition are provided. An embodiment of the present invention provides a method of monitoring a machine condition, comprising the following steps: modeling a normal signal model performed by detecting a signal for monitoring condition of a normal machine and converting the detected signal to a normal signal model in time domain using a hidden Markov model (HMM) algorithm; calculating a probability value data of the monitoring signal at a subject machine performed by detecting a signal for monitoring condition of the subject machine in real-time and converting the detected signal to the probability value data relative to the normal state signal model using the HMM algorithm; and determining a section having deficiency where the probability value data of the monitoring signal at the subject machinery is not maintained constantly relative to the normal signal model.