Anomaly detection and failure prediction for predictive monitoring of industrial equipment and industrial measurement equipment

A method of anomaly detection and failure prediction for predictive monitoring of ultrasonic flow meters of an associated process, the method comprising:

• retrieving a detection model generated from historical measurements of a measurement signal group associated with the ultrasonic flow meters;
• receiving real-time diagnostic measurements associated with the ultrasonic flow meters of the associated process;
• scoring the received real-time diagnostic measurements using the detection model to generate performance metric scores for each of a major space projection and a null space projection of the detection model of the detection model as a percentage of variability from the detection model;
• identifying one of the performance metric scores for each measurement signal group to detect anomalies or predict potential failure of the ultrasonic flow meters, selected from a maximum score from one of the major space projection or the null space projection;
• and
• displaying the identified performance metric score and associated parameter primarily contributing to the performance metric score for a measurement signal group of the ultrasonic flow meter.

A system comprising:

• a processor;
• a network interface coupled to the processor for receiving diagnostic measurements from a plurality of industrial equipment or industrial measurement equipment associated with a process;
• and
• a memory containing instruction which when executed by the processor perform:
• receiving a plurality of historical diagnostic measurements of respective measurement signal groups associated with a respective industrial equipment or industrial measurement equipment of the associated process;
• generating a detection model for each measurement signal group associated with the industrial equipment or industrial measurement equipment, defining statistical thresholds for operation of each measurement signal group derived from the plurality of historical diagnostic measurements;
• receiving real-time diagnostic measurements from each measurement signal group associated with the respective industrial equipment or industrial measurement equipment;
• scoring the received real-time diagnostic measurements using the generated detection model to generate performance metric scores for each of a major space projection and a null space projection of the detection model as a percentage of variability from the detection model;
• and
• identifying one of the performance metric scores for each measurement signal group to detect anomalies or predict potential failure of the industrial equipment or industrial measurement equipment selected from a maximum score from one of the major space projection or the null space projection.