Expert Fault Modeling

Expert Fault Modeling application (EFM) is part of Predictive Maintenance Workplace in APM suite of ABB Ability™ Genix, which allows you to comprehensively track the health of all the assets in a system and works as a standalone expert system to predict the faults and monitor the performance of a complex system.

The prediction is based on the diagnostic and performance models developed within Expert Fault Modeling application. Diagnostic model is a data science driven model with a set of rules to identify the health of an asset/system. Performance model is a set of rules to identify the deviations in the Performance of an asset/unit at a system or fleet level with respect to a reference condition. First principle analysis with Machine Learning (ML) methods are used to ensure robust analysis of a system. These methods provide a multi-criteria decision making tool, which is easy to customize, scale and provide expert recommendations related to health of a complex system, which are multivariate in nature.

The EFM application provides health and performance summary and helps in real time decision making by providing best alternatives among possible damages. It also provides non-dimensional parameters that localize the faults in terms of probability and severity.

The purpose of EFM application is to provide all the information to define the best maintenance strategy for the plant/equipment/items without being dependent on individual experience.

EFM provides a heuristic approach to predict the failure mode with a probability and how severely the failure can affect an equipment or system. The advantage of this approach is that, its independent of type of equipment, installation, manufacturer and can be used for any type of failure modes.

The general methodology involves building a machine learning diagnostics model of the equipment or system to derive a reference value for parameters associated with the equipment or system. The output from Expert Fault Modeling application provides Health Indicator values against a preset threshold. Similarly, the Fault Indicator of EFM provides information regarding the probability and severity of the failure modes. System level information provides opportunity to access multiples assets and take required actions.

The healthy model creates the following three types of information, where each information is expressed as a normalized percentage:

  • Key Diagnostics Indicator (KDI): The current equipment properties are compared to the model to find the closest operating point. Each property is then compared to this “reference” and the deviation is calculated as a KDI.

  • The equipment properties are compared to the fault model to derive the following two types of fault index:

    • Key Probability Indicator, which indicates the likelihood of failure.

    • Key Severity Indicator, which indicates the magnitude of failure.

  • A forecast predictor is used to provide a future prediction of each index.

If the difference between reference and current values of properties are statistically significant for a predetermined period, or exhibits a statistically significant data pattern, rules are triggered to advise about the potential for a failure to happen. The result is used to provide recommendations, which can be directly integrated with recommendation management system or work management system.

To access the Expert Fault Modeling application, go to Genix Platform > Digital Apps Center > APM Model Configuration and Deployment > Model Configuration and Deployment > Expert Fault Modeling.