The same sensor-based IoT platform technology that has connected devices and appliances in our homes, is also used in large equipment like pumps and engines. If they have a sensor, they can stream information, send telemetry data out to the cloud, have the data processed and stored in a database, and send notifications back to the user based on any number of desired outputs. 

The goal of this sensor output data is to provide business intelligence that helps you make smarter decisions. But many large equipment industries collect reams of telemetry data from a wide range of devices and ultimately do nothing with it. This happens because they receive thousands of data points with no simple way to prioritize the signals showing problems.  

To combat this, companies – especially those that use large equipment that can require expensive maintenance – need to identify the specific data streams that indicate developing failure conditions. Artificial intelligence-driven microservices make processing and analyzing your asset’s data easier.


What’s an AI microservice?

AI microservices are used to complete specific tasks like data processing and data aggregation rather than using a single monolithic application. They’re independent, smaller solutions that are easier to deploy and integrate with third-party tools or tool sets. 

For example, our Amber solution is a serverless, cloud-to-cloud application that taps into your legacy applications, performs data-related tasks on your behalf, and serves up the data outputs in an easy-to-use format.

In other words, your days of sifting through thousands of data points to monitor your equipment’s condition are over.  


How do AI microservices help you get more out of your IoT platform?

Let’s say you have several worksites, each with multiple high-value assets connected through IoT sensors. It’s challenging (if not impossible) for your operators to monitor every piece of data from each of your assets. A better option is to have your operators monitor assets only when their data measurements start to indicate problems.

An AI-based solution like Boon Logic’s Amber monitors asset data for you and watches for complex data patterns that indicate when a condition exists within your asset that’s inconsistent with how it’s previously behaved.

When you use AI in this way, it performs the data analysis for you and selectively raises red flags alerting your domain expert where to take a closer look. It also helps you avoid unplanned production downtime and proactively choose the most opportune time to make critical repairs.

Instead of sifting through thousands of data sets, you’re looking at one integrated dashboard that clarifies when your domain expert needs to take action.

Here are two examples of how an AI microservice helps you get more out of your IoT platform.


Anomaly reporting

Traditional anomaly detection approaches look for minimum/maximum excursions of individual sensors beyond a statistically defined normal range.

But what happens if each individual sensor value is within the statistical limits of its normal operation, but there is a relational anomaly between sensors? These relational anomalies give the earliest indication of asset non-compliance.

This is where the unsupervised machine learning (a subset of AI) capabilities of Boon Logic’s Amber solution come into play. We leverage this technology because it’s especially effective at detecting anomalies and making predictions based on deviations from normal operation, even when those states can have a significant amount of expected variation.

Amber consumes sensor data values as it builds its machine learning models and monitors assets and processes. It creates a “snapshot” of the state of that asset at that moment and produces an entire array of analytic outputs, including an anomaly index.

This means your domain experts need only monitor one high-level metric to identify when something has changed within your asset. Values range from 0 to 1000, with values close to zero signifying patterns that are the most common and values close to 1000 considered more anomalous. Platform providers can easily integrate this measurement into SMS/email alerting systems used by their maintenance teams.


Real-time warning levels

Another analytic output Amber produces is a real-time warning level. This is possible because during its set-up phases, Amber builds high-dimensional models that are individualized to each asset being monitored.

This configuration works on a wide variety of assets and processes, including single-feature and multi-feature processing (sensor fusion). In fact, Amber is most powerful when it processes sensor fusion vectors streaming from a multi-sensor asset.

When monitoring your assets, each sensor fusion vector processed by Amber produces analytic outputs, including a warning level based on the frequency of anomalies within recent history. These warning levels are expressed as 0 (normal), 1 (asset is changing), and 2 (asset is critical).

Warning levels are then reported to your domain experts in a simple green/yellow/red metric that helps them quickly identify when an asset needs attention. When that happens, Amber automatically generates root cause analysis directing maintenance teams to the sensors implicated in the warning.


AI doesn’t have to be hard (we can help)

If you have a historical data set, Amber can pre-train your data, construct a model that analyzes months’ worth of data, and show you where a failure would have been detected within that data.

Even better, this process is fast – 1000x faster than alternative algorithms – and takes only a few seconds once the historical data set is fed into Amber. Once the model is constructed, you’re in a monitoring state with Amber analyzing new data and providing you with real-time analytics (rather than waiting for data results from a computer located somewhere else).

If your Amber warning level is green, you’re good to go. And if some condition within your asset starts to change, the anomaly index informs you immediately.


Readiness Assessment CTA


Want to learn more?

Boon Logic’s AI-based predictive maintenance condition monitoring tool can help you protect your high-value assets and significantly reduce unplanned downtown.

Amber has prebuilt integrations for several popular platforms. If customization is needed, Amber uses a Rest API programming interface and also provides software development kits with developer resources in various programming languages, including Python, C++, C#, JavaScript and others. That means subject matter experts can use it to produce rich operational insights in a few days.

If you want to learn more about using AI microservices to get more out of your IoT platform, we’d appreciate the chance to have an introductory conversation. Contact us today.