Know what’s going to happen and when with our Event Predictive Archetypes
We can ingest data at more than a million data points per second, and render scores in a few milliseconds. Simularity uses machine learning and statistical probabilities in a patent-pending process to learn what “normal” looks like in any time series data set. We can tell when things aren’t acting normally, and we can also identify event signatures for specific events that can provide context and prescriptive maintenance instructions.
This example is a hard drive failure predictor. Based on data from 53 different sensors on the hard drives, and a sample set of failed drives and a set of normal drives, we created a Time Series Predictive Archetype for hard drive failures. With this Predictive Archetype, we can monitor all sensors on all live drives in real time, and score them, determining whether or not they will fail, how soon we expect them to fail, and what type of failure it might be. Our dashboard indicates the real time status of the drives, and flags those that are showing signs of failure. Alerts are set for when drives exceed score thresholds.
A Predictive Archetype for time series data consists of a set of Event Signatures. Each event signature represents a set of sensor patterns, that when seen toegther, indicate a particulare type of failure. There are many ways a hard drive can fail. By looking at the event signature that a drive is exhibiting, and comparing that to know event signatures for the various different ways a drive can fail, we can prescribe the appropriate corrective action (i.e. which part to replace).
The snapshot below is of the main dashboard and summary details on a drive that is predicted to fail. The dashboard shows the failure score (higher means more likely to fail), and also predicts the hours to failure. The summary details show thumbnails for the 5 sensors that are most indicative of a pending failure, and the predictive strength of each sensor. This summary detail is the Event Signature.
In this video, Simularity’s AI is used to detect anomalies in the sensors attached to gravel mining equipment. Deployment of this solution took just 3 days.
This is our anomaly detection technology integrated with Splunk. This particular example uses airline fare response times. Contact Us to learn more about how Simularity can help you find anomalies in your logs and more effectively manage your data center.
Simularity Brings Real Time Deep Learning To The Edges Of The Internet Of Things
The IoT is all about time series data. Simularity is the only company effectively doing real time deep learning on massive amounts of time series data. We’ve developed innovative new methods you can’t get anywhere else.
- Run on huge networks of commodity devices, both large and small, in a cooperative system
- Analyze and act in real time, regardless of computing, connectivity, and bandwidth restraints
- Automatically learn normal individual sensor behavior, including both short and long term cyclic behavior
- Adaptive compressed reporting reduces bandwidth consumption, storage, and analysis time
- Smart alerts based on deep learning mean fewer false alarms and less time spent determining what is happening
- Deployment is fast and easy: no need to create rules or write code
Contact Us to learn more about how Simularity can help you put your connected device data to work