Modeling and visualizing discrete IoT data using Azure Time Series Insights

Published Apr 02 2020 02:06 PM 1,743 Views

In an industrial IoT scenario, live telemetry data produced by devices, sensors, or tags can be continuous or discrete. We’re excited to share that Azure Time Series Insights can now process discrete values, enabling users to view up to 100 states. With this new feature, users can correlate state values with continuous measurements to better manage operations.


As a developer working with IoT data, you need a simple yet powerful tool to model discrete telemetry, combine it with continuous data on a dashboard to provide valuable insights to your users and customers. Azure Time Series Insights has just what you need. In this article, we will quickly review the definitions of a time series and times series data, look at how to model discrete data within the Azure Time Series Model (TSM), and, see how combining discrete and continuous data on a dashboard can provide valuable insights.


You can explore these features in Contoso Wind Farm demo environment.






Let’s review definitions that are important and provide context.


  • Time series

A time series is a collection of observations made sequentially over time. In the IoT world, most sensors produce time series telemetry. Characteristics of a time series include the following:

1. New data is always stored and recorded as a new entry.

2. Data is usually stored in chronological order.

3. All data is time stamped.

Data in a time series can be classified as one of two types: continuous or discrete.


  • Continuous time series

Continuous time series are observations that are made continuously over time. They can potentially take any value. We look at continuous time series to monitor trends. For example, we could look at the value of temperature over a six-month period.


  • Discrete time series

Discrete time series are observations taken at specific times. They take on only a finite number of values.  For example, we could look at the value of valve status over a 3-month period.


In the IoT world, we often come across use cases where customers want to analyze continuous data in conjunction with discrete data. For example, a facilities manager might want to correlate room temperature (continuous) with occupancy (discrete). A manufacturing lead might want to correlate pressure (continuous) with valve status (discrete). A technician might want to correlate vehicle speed (continuous) with the number of transmission shifts (discrete), etc. To enable such scenarios Azure Time Series Insights in extending support for analyzing discrete signal via Categorical variables.


Modeling Discrete Data


The Azure Time Series Model lets us configure categorical variables to represent discrete data. Categorical variables can take on one of a limited, fixed number of possible values while assigning each unit of observation to a group or category with up to 100 labels. This grouping also makes it easy to visualize and explore data in Azure Time Series Explorer.


Let us look at a facilities management use case where we define a categorical variable to represent the discrete values (ON, OFF) of a damper valve and then troubleshoot the cause of cold temperature in a conference room.


  • Scenario 1: Modeling Damper Position

Imagine a multi-building, multi-city campus of a high technology company. To better understand and contextualize incoming time series data from sensors across the campus, the team has built an Azure Time Series Model (TSM) of their facilities.

Some of the rooms contain a damper valve. A damper valve (or plate) stops or regulates the flow of air inside a duct, chimney, VAV box, air handler, or other air-handling equipment.

Since the damper position is either OPEN or CLOSED, its value is discrete. We use a categorical variable to represent its state. In Fig. 1 we see the categorical variable definition for Damper Position




Figure 1. Categorical Variable


The JSON code for the Damper Position variable is shown in Figure 2.




Figure 2.  JSON Code for Damper


  • Scenario 2: Troubleshooting room temperature

The facilities manager has been receiving complaints about a cold conference room. One of the technicians queries sensor data using the Azure Time Series Explorer. They pull in the data for the damper position (discrete) and the data for the room temperature (continuous). See Fig. 3 below.




Figure 3. Viewing discrete data and continuous data simultaneously


By viewing the discrete data (damper position) simultaneously with the continuous data (room temperature), they notice a pattern. The temperature is dropping by several degrees whenever the damper is open. The technician can use this information to follow up.


In this scenario, the technician used the Azure Time Series Explorer. Companies can also visualize the data by building their own application(s) on top of the Azure Time Series platform. With this approach, data can be integrated into line of business applications.


Azure Time Series Insights also integrates with Power BI. This provides another approach for visualizing time series data.


Next Steps


In IoT scenarios, live telemetry can be continuous or discrete. You can now view the discrete data in your IoT time series with Azure Time Series Insights. Begin by modeling both your discrete and continuous data with Azure Time Series Model.

Version history
Last update:
‎Sep 14 2021 09:03 AM