Here’s a visual of the framework:
Azime opened explaining that prognostic maintenance requires starting with data—sensor data and generated data. This data is pre-processed and used to develop a detection or prediction mode by identifying condition indicators and training a model. This model is deployed and continuously refined against the continuously acquired data. Example of generated data might be to look at sensor data in the frequency domain looking for peak frequencies as condition indicators of healthy operation and faulty operation.
Anomaly detection are other conditions that can be condition indicators that feed the machine learning algorithms that build the model. The model for equipment is an estimation of the remaining useful life of the machine. Given the amount of data, pre-processing and identifying condition indicators is typically performed at the edge with the model running in the cloud for the experts who need to evaluate the information uncovered by the model.
Some of the ways to find condition indicators is by using trial and error with raw data by mean, variance, skewness, and other algorithmic manipulations. The engineering process flow begins with having the algorithm toolbox with a collection of user-defined or pre-loaded sets. The model creation process takes historical data and applies pre-processing. Net the prognostics model is built and then put into production.
Alvin explained the elements in the Intelligence Framework. These include the customizable plugins/extensions, a programmable expert system, and configurable data sources. Examples of plugin/extensions include MathWorks, FrontlineSolvers, and Ovation SmartProcess. Expert systems are rules-based systems that tries to capture the experiences of a true expert in the field. The expert system emulates the decision-making of an expert through rules & actions to achieve a desired goal.
The Intelligence Frameworks prognostics provide decision support, predictive maintenance and asset planning. The workflow for implementing an Intelligence Frameworks project is to setup the plugins, configure the data sources, develop the expert system and to design the visualization to monitor and view the results from the implementation.
Take a look at the Ovation Intelligence Framework data sheet for addition information.