Digital Twin was named one of top ten strategic technology trends by Gartner for 2017. There are several technologies being developed under this name. Mitek Analytics provides Operational Digital Twins that include building asset models from historical data. These models characterize variation across the asset fleet and trends along the time.
Our Performance Digital Twin software analyzes sensor time series data, such as aircraft flight recorder data, collected during normal operation of the IIoT assets. Our Reliability Digital Twin software analyzes reliability event data, such as replacement and repair dates, collected in maintenance and repair operation (MRO) for the IIoT assets.
In the example illustrated in the figure, the Performance Digital Twin analytics is applied to operational data collected in the jet engine fleet. The Predictive Analytics train the models on big data sets. The Prescriptive Analytics realize business value using these predictive models. The jet engine fleet monitoring application is one of the most demanding in the Industrial IoT because turbomachines must be maintained every few thousands hours, are mission critical, and must operate with high performance. Performance Digital Twin can be deployed for a variety of assets. Our Digital Twin analytics are driven by IIoT data collected in operation for fleet of assets; these are Operational Digital Twins. Historical data is used to build predictive models. These models are then used to prescribe actions, provide decision support for current data. Mitek has developed Analytics Engines for Performance Digital Twin and for Reliability Digital Twin.
Performance Digital Twin
The Analytics Engine for Performance Digital Twin (PDT) provides unique functionality for multi-level analysis of the sensor data. It analyses large sets of data images recorded across the asset fleet over many time periods; each image contains multivariable time-series data for a time period. The time period could be a single aircraft flight, car trip, one day of industrial plant operation, etc.
The PDT Predictive Analytics builds models for asset fleets based on historical operational data. This is done in two stages. First, a baseline performance model of a single asset is implemented and tested. Such multivariable model could rely on existing engineering knowledge specific to the customer or shared by the industry and published. In many applications Mitek Analytics can provide the baseline performance model; examples include fixed wing aircraft, wind turbines, substation transformers, automotive vehicles, and gas turbines for power generation. Second, the RDT Predictive Analytics scale up the baseline performance model to obtain a multi-level model that characterizes the variation across asset population and trends along the time. This data-driven model could be used for monitoring the performance of the assets. PDT analytics includes data integrity check and data cleansing, which must be done before training the model. Up to 10% of all data records might have bad data that could distort the model training.
The PDT Predictive Analytics uses innovative mathematical algorithms for Big Data batch processing. The algorithms perform optimal Bayesian estimation of model parameters using convex formulations. The model parameters are obtained by using convex optimization methods to provide fast, precise, and scalable estimation with guaranteed performance. The trained RDT Predictive Model characterizes variations of asset performance parameters across the fleet (population) and their trends along the time. In one deployment example the RDT Predictive Analytics estimate a multi-level model with ten million parameters from one trillion data points.
The PDT Prescriptive Analytics produces added value by enabling of fleet-wide energy efficient performance, condition-based maintenance, availability, safety, and other desirable outcomes.
Reliability Digital Twin
Our Reliability Digital Twin (RDT) analytics extracts valuable information from reliability event data. The RDT analyzes reliability event data, such as replacement and repair dates, collected in maintenance and repair operation (MRO) for the assets. The RDT Analytics Engine provides unique functionality for multi-level analysis of the MRO data for asset fleet collected over many part repair and replacement periods.
The RDT analytics have been proven in Air Force applications. The data on reliability events (failures, part replacements, repairs) are analyzed with methods that are largely independent of the application. As a result, our analytics can be used in many vertical applications. The data-driven modeling facilitates the discovery of trends and causes of substandard reliability performance and the identification of remedial actions. In the sustainment of military aircraft, the reliability event data has been available for a few years. The Industrial IoT enables collection and analysis of the reliability event data in many other up and coming applications.
The baseline reliability models that are currently used in many industries describe the part reliability through a single parameter, Mean Time Between Failures (MTBF). Mitek’s Predictive Analytics for Asset Fleet Reliability connect individual histories of parts and systems in a unified multi-level model of fleet reliability that scales up the MTBF model. Our Predictive Analytics account for the time-varying reliability parameters across the fleet and with usage time.