DeltaV Neural – Lab Entry Automatic update based on lab or analyzer entry of new samples Estimates future value of measurement based on current upstream conditions, crystal ball approach Diagnostics to detect deviation in estimate from new sample value
Future Prediction The trained neural network provides a predicted output into the future NN block is configured with a ‘FUTURE’ Output and an ‘OUT’ Allows user to perform what-if studies Valuable in processes with large dead times
Future Prediction ‘FUTURE’ is calculated by setting the delays with associated inputs to zero, that is give the steady state solution for the given input values
Lab Entry Function Block Lab Entry (LE) allows lab analysis to be entered and collected in the DeltaV Historian Time difference between the sample taken and the analysis is entered automatically to reflect the DELAY parameter Lab value screening and time delay in processing the sample are provided in the LE block
Configure LE Open the EX2_SIM with the control studio Remove the existing input, TT2-4, from the neural network block. Add LE block, TT2-4LE, and wire as shown
Lab Entry, LE, Block Both OUT and DELAY should be wired
Lab Entry Modes Manual – Lab entry is passed through to the output Out of Service – Output is set to Bad OUT and DELAY parameters will normally be Good Constant when block is in MAN When new analysis is entered, the status of OUT to Good Not Limited for two executions to indicate a new sample
Save and Download both The EX2_SIM control model and the Historian Complete the wiring Save and Download both The EX2_SIM control model and the Historian
Add LE block to the Operator Display Open lab2sim display, quick edit and add the LE block. Browse to locate the specific block Configure the point with OUT and CV, or current value
LE Display Configuration Save the display and exit Open the display – Not how to make entries in the LE block
Detail Faceplate, Push Lab Entry to enter the data
LE experiment Make changes in the inlet water temperatures and hot flow rate. Note the temperature out and make the lab entry at various times Watch for AM and PM What about daylight saving time? Retrain the network and check the results
Plant LIMS can interface with DeltaV through OPC OPC, or OLE for Process Control, offers an alternative by defining a standard set of objects, interfaces and methods for use in process control and manufacturing automation applications to facilitate interoperability. LabWare is a member of the OPC Foundation and LabWare LIMS fully supports the Data Access Specification. LabWare LIMS Process Windows can display "live" data that automatically updates when it is changed. The data sources can be either LIMS results, or OPC tag values.
Neural Network Feed Forward Experiment We will now develop a neural network to compensate for changes in exchanger We will develop a network that can anticipate the cold water rate required The network will act as a feed forward calculation
Neural Network Feed Forward Experiment Delete the LE block and the NN1 block before you configure the new block Be sure to download the Historian before you train the new network!
Configure the inputs as shown, Why do we just want the SP and not the PV?
Wire Output of NN block To FF_VAL of TC2-4
For TC2-4 FF_VAL scale 1 to 1.0 GPM FF_GAIN of 1.0
Training from a file In this lab you will train from an existing data file The file is EX2_SIM NN_FF 20040423 save.dat
Auto generate from file