Worlds of new opportunities are opening up for safety-critical and real-time machine learning applications. The general pattern? Gather sensor signals so machine learning platforms can make predictions and trigger corrective actions. These types of applications are easier to implement than ever, thanks to the release of the Oracle Cloud Infrastructure Anomaly Detection Service.
Whether it’s for building management, energy markets, nuclear reactor safety, manufacturing, or something else, this kind of ML system starts with sensor data. For example, Oracle has an IoT data capture service as well as partners, such as SS Global, who offer solutions. Add to that OCI Anomaly Detection Service, the AI component for making inferences on incoming time series data. These inferences become the basis for an action to be taken.
What kind of action? It could be a dispatch to the field, a service request, or an overhead cable fault alert sent via Oracle Utilities Customer Cloud Service. It’s critical to have a wide set of options for the most important part of this sequence: the intelligent action.
These use cases are straight forward. Now, let’s push the envelope.
Imagine you have several warehouses fitted with fixed fire detection sensors. Of course, you want to reduce potential injury, the catastrophic costs of fire, and damage from fire retardant. But wouldn’t it be better to avoid all this by acting before a fire starts? With OCI Anomaly Detection Service, you can monitor multiple variables simultaneously, such as temperature, humidity, and various wood or plastic fire emissions such as carbon monoxide, fluorhydric acid, and nitrogen oxide.
Any single value may not be critical, however. Monitoring for scalar signal thresholds can lead to increased false alarm probability (FAP). Also, the various combinations that can result in a fire are vast, which can lead to higher missed alarm probabilities (MAP). OCI Anomaly Detection Service is based on multivariable prediction machine learning, which evaluates the many different sensor signals to predict a possible fire in a warehouse — with better results due to lower false and missed alarms.
You can leverage the Oracle Cloud platform to do far more than just send an alert. Consider a fire danger prediction from OCI Anomaly Detection Service. That prediction can trigger the dispatch of a drone or terrestrial robot to gather additional sensor data and also take video. You have now reduced possible injury by sending an uncrewed vehicle to the scene. It can also make further predictions based on what its mobile sensors detect (vapors, liquids, or other hazards) combined with readings from the on-site sensors.
Thus, you’ve integrated enterprise notifications, service records, and dispatch with OCI Anomaly Detection Service in a single platform that gathers signals, makes predictions, and acts intelligently.
Here is a conceptual view of a complete intelligence system. Reading from left to right:
Machine Learning is rapidly changing industries across the board. For example, energy markets can combine OCI Anomaly Detection Service and trading platforms to enable real-time, optimized transaction of complex assets, including batteries, solar, wind, pumped hydroelectric, distributed energy resources, and demand response applications. A solar array can intelligently retain and release stored energy based on predictions from OCI Anomaly Detection Service around market volatility that come from probabilistic time series of optimal price forecasts.
Oracle’s OCI Anomaly Detection Service is ready for developers, scientists, and data engineers at startups, enterprises, and public entities to put their imaginations to work and build solutions that take action based on data signals.
Bill is a senior director in Oracle’s cloud engineering team and has over 30 years of experience in cloud native development, enterprise architecture, strategic planning, and business architecture. He is a frequent speaker at The Open Group and Oracle conferences, the inventor of the DVD Rewinder, and named on three patents.