ADAMOS Advanced Streaming Analytics

Nowadys machines generate big volumes of data through integrated sensors which have to be analyzed instantly. The analysis allows to identify anomalies which help to detect, report and prevent defects. Therefore use cases such as condition monitoring or predictive maintenance can be implemented easily.

Streaming analytics is all about extracting business value from data in motion in real-time. It involves the ability to constantly calculate statistical analytics on data streaming from different data sources such as for example machines, devices, sensors, applications, social media, and more.

ADAMOS Advanced Streaming Analytics embeds an industry-leading streaming analytics engine which allows you to address exactly these use cases. The engine contains patented Hypertree technology that provides highest throughput and shortest response times to ensure decisions are made the instant an event occurs. The natively built engine has a small footprint making it suitable for running in large cloud deployments as well as on small edge devices.

Streaming Analytics Design

The ADAMOS streaming engine comes with a native Event Processing Language which allows developers to create arbitrary analytics on both discrete events and event streams. There is support for a wide range of processing models for data in motion, including time- and location-based windows, streaming calculations, anomaly detection and event relationships. Data can be blended together from any number of sources.

The engine allows to detect patterns across multiple streams and correlate events from multiple streams. Event streams can also be enriched with historic and contextual data-at-rest where necessary to improve accuracy of automated decisions.

Visual Streaming Analytics Design

In addition to programming analytics logic using the Event Processing Language, the Analytics Builder allows to design streaming logic using a graphical web editor. The workflow-like design approach lends itself ideally to power users and even business users.

The Analytics Builder comes with a collection of pre-defined analytics building blocks which allows to eliminate manual coding, cross-checking and error correction that comes with custom code or proprietary solutions. As such it speeds up solution development significantly.

Machine Learning Execution

The platform features embedded machine learning and predictive model execution capabilities. This includes advanced analytics formulated in the vendor-neutral PMML industry standard to execute models that are exported from tools such as SAS, IBM SPSS, KNIME, or open source languages like R and Python.

Supported algorithms range from simple regression to complex statistical and machine learning models, including model ensembles and deep neural networks. The PMML-based machine learning model execution is currently only supported in on-premise deployments.

Alarm Management Integration

The streaming engine seamlessly integrates with the Alarm Management capability of ADAMOS Core. Whenever an analytics rule is violated it can trigger an alarm which in turn sends a command to a machine or an SMS or E-Mail to an operator, or even execute a service integrating with a 3rd party system to, for example, open a support ticket.

Advanced Streaming Analytics - Good to know

Condition Monitoring and Predictive Maintenance are the most common industrial themes that come up in this context.

Think about a shop floor of an automotive manufacturer on which raw car bodies are painted, for example. Applying an anomaly detection algorithm across data streams of paint nozzle pressure and paint viscosity originating from a paint robot as well air temperature and humidity from according sensors mounted in the room, you could identify and even prevent potential quality issues during the manufacturing process. As a result, you could increase the first run rate and avoid expensive post processing quality measurement and correction stations.