Creating a Model

To create a new model, click on AI Models using the side navigation bar and click the ellipses in the top right corner to reveal the model list dropdown. From here click the + Create AI Model button. You will be taken to a model create form which contains the following parameters.

Name

The name should be a unique string identifier used to briefly describe this model.

i.e. Boiler #2 Model

Description

The description should be a more detailed string used to describe the model.

i.e. Boiler #2 DRL temperature control

Scan Rate

The time interval in seconds that the model will be executed.

Sample Rate

The time in seconds that the model’s sample are spaced at. This will be determined during the model training.

Scan Rate vs Sample Rate

When it comes to executing a DRL model, two timing fields that are extremely important are the sample rate and the scan rate. These two fields can work together to solve complex issues with model timing. During training, the ML engineer may need to pass in a days worth of data for a very slow reacting problem. To do this, the data is typically heavily downsampled, in this example let’s say to once an hour. Now, the scan rate is how often the model should execute, this is typically faster than the sample rate of the mode. For example, if the scan rate was set to 1 minute in the above example, the model would execute once a second, and pass in 24 samples spaced 1 hour apart.

This is used in many ML applications that have large delay times in process responses. Below is a screenshot of a model with a sample rate of 5 seconds, while the model and device is executing once a second.