Collaboratively administrate empowered markets via plug-and-play networks.
Adding Knowledge to IoT
IoT connectivity and Artificial Intelligence services provide almost unlimited new ways of capturing and analysing a wealth of structured & unstructured data. By using a single, common, knowledge automation tool as the brain of the data ecosystem, real business value can be achieved in the form of expert automated decisions.
In this sample
In this sample, there are two Knowledge Bases authored in XpertRule Decision Author: the first knowledge base monitors real time data streams from industrial equipment to detect patterns and anomalies indicating a potential operating problem and the second knowledge base runs on demand to support troubleshooting and diagnosis of problems. The specific technologies include:
XpertRule Decision Author (Rules Authoring & Data Mining Environment)
XpertRule Node.js Runtime Environment (Decision Automation server installed on a Bluemix Instance)
Architecture & Overview
This industrial machinery condition monitoring application continuously monitors values from IoT sensors and generates diagnostic fault conditions to alert end users and technical support personnel of abnormal equipment operation to minimise unplanned downtime and reduce the risk of equipment failure.
The application uses two Knowledge Bases authored in XpertRule Decision Author: the first knowledge base monitors real time data streams from the equipment to detect patterns and anomalies indicating a potential operating problem and the second knowledge base runs on demand to support troubleshooting and diagnosis of problems. The system architecture is shown below. In this case the diagnostics are running on IBM Bluemix but a local server or alternative Cloud based provider can be used.
The demonstration monitors key equipment measurements such as values of classifier, rotor and fan currents and the feed flowrate. Values are updated on demand by selecting refresh or on a continuous basis by Auto Update and example fault conditions are generated for the different components of the milling system.
The demonstration cycles through a sample of operating data and repeats the same operating scenario after 5 mins. Alerts are shown next to the relevant values on the display of the current equipment state and where there are multiple alerts an assessment of the likely problem is shown at the top. An explanation of any required follow up action is available by selecting an alert.
Interactive guidance for troubleshooting equipment problems and alerts is available via the Troubleshoot option. The feedback forms in the demonstration have been disabled but are typically used to gather feedback from field based user experience.
Predictive analytics use data mining models to detect anomalies in the values of key aspects of equipment performance. The example data mining tree below calculates the expected value of the operating amps or current to an item of equipment under specified operating conditions. An alert is raised if the measured amps deviate significantly from the expected value.
Adding rule based intelligence to Predictive analytics enables faster diagnosis of operating problems. In the decision trees below we take into account the equipment status before issuing an alert and determine the likely underlying cause of a combination of active alerts.
The main decision tree for the troubleshooting knowledge provides recommended checks for different operating problems selected by the end user. Each symptom has a hierarchy of decision trees providing guidance to resolve the problem.
In the decision tree below we show the recommended steps to resolve one of the problems shown in the main tree.
End users can submit feedback on troubleshooting guidance so that the knowledge base is updated with field based experience.