GB/T 22394.2-2021 Condition monitoring and diagnostics of machines—Data interpretation and diagnostics techniques—Part 2:Data-driven applications

Scope

This part of GB/T 22394 provides the process for implementing data-driven monitoring and diagnostic methods to assist professionals, especially those in monitoring centers, in their analysis work. Although some steps are already embedded in existing tools, it is still necessary to pay attention to the following steps for better use: -- Selection of assets, key faults, and available process parameters; -- Data cleaning and resampling; -- Model development; -- Model initialization and adjustment; -- Model performance evaluation; -- Diagnostic process. Implementing these steps does not require comprehensive knowledge of statistical methods, but it requires the ability to first establish a training model and apply it to monitoring and diagnostic processes. Train a data-driven monitoring model on a machine in normal operating conditions. The principle of fault monitoring is to compare observed data with estimated data. The difference between the observed value and the expected value of a parameter (called the residual) indicates the presence of an anomaly, which may be related to the equipment or instrument. Train a data-driven diagnostic model on a machine in normal and faulty operating conditions. The principle of diagnostic methods is not to detect parameter deviations, but to identify faults by comparing the observed conditions with the faults learned during the training phase. Commonly used techniques include pattern recognition and pattern classification. The data can be obtained from the historical data of a distributed control system (DCS) or from a specific monitoring system.

Development Information

Referenced Standards

Adopt standards

Related Standards