DeltaV Neural overview

Use the DeltaV Neural application to create neural networks, which you can use as intelligent sensors to predict variables that are difficult to measure or otherwise impossible to determine in a timely manner. Such intelligent sensors are used in many industries for replacing and cross-checking physical sensors/analyzers and act as inexpensive, on-line predictors of lab test results.

The NN function block implements neural networks in the DeltaV system. When a module that contains the NN function block is downloaded, all inputs and outputs are assigned to the Continuous Historian. The DeltaV Neural application uses this historical data to train and test the neural network.

The DeltaV Neural application:

• Provides an interface from which you select historical data to train and test a neural network
• Preprocesses the selected historical data to ensure that it can be used to train the neural network
• Determines which process inputs have an effect on the process output and calculates time delays
• Trains the neural network using the selected data
• Verifies the accuracy of the model using the selected data

The application can perform most of these tasks automatically. However, if you are an expert user you can perform these steps manually and adjust results as required.