Feature extraction and fault detection in a non-stationary process through unsupervised machine learning

A method comprising:

• receiving training data regarding a field device in an industrial process control and automation system, wherein the training data includes a dynamic process mean;
• extracting a meaningful feature using a first portion of the training data related to the operation of the field device, the meaningful feature extraction including determining a moving mean index, a first order incremental index and a second order incremental index;
• performing an unsupervised classification using a second portion of the training data to determine a health index for the meaningful feature;
• identifying a faulty condition of real-time data using the health index of the meaningful feature;
• and
• performing a rectifying operation in the industrial process control and automation system for correcting the faulty condition including shutting off or ramping down a process of the field device.

A non-transitory machine-readable medium encoded with executable instructions that, when executed, cause one or more processors to:

• receive training data regarding a field device in an industrial process control and automation system, wherein the training data includes a dynamic process mean;
• extract a meaningful feature using a first portion of the training data related to the operation of the field device, the meaningful feature extraction including determining a moving mean index, a first order incremental index and a second order incremental index;
• perform an unsupervised classification using a second portion of the training data to determine a health index for the meaningful feature;
• identify a faulty condition of real-time data using the health index of the meaningful feature;
• and
• perform a rectifying operation in the industrial process control and automation system for correcting the faulty condition including shutting off or ramping down a process of the field device.