FB_MllPrediction

Syntax

Definition:

FUNCTION_BLOCK FB_MllPrediction
VAR_INPUT
    stPredictionParameter : ST_MllPredictionParameters;
    eTraceLevel           : TcEventSeverity;
END_VAR
VAR_OUTPUT
    bBusy                 : BOOL;
    bError                : BOOL;
    ipResultMessage       : I_TcMessage;
    hrErrorCode           : HRESULT;
END_VAR

Inputs

Name	Type	Description
stPredictionParameter	ST_MllPredictionParameters	General Prediction parameters
eTraceLevel	TcEventSeverity	Eventlogger trace level. Default = Critical. Must be set before Configure method is called

Outputs

Name	Type	Description
bBusy	BOOL	True if a asynchronous action is taking place
bError	BOOL	Indicates error in method
ipResultMessage	I_TcMessage	Contains the last invoked error
hrErrorCode	HRESULT	Unique error code

Methods

Name	Description
CheckPreferredIODataTypes	Checks if the supplied I/O data types are preferred by the currently loaded model.
CheckSupportedIODataTypes	Checks if the supplied I/O data types are supported by the currently loaded model.
Configure	Configures the FB with the given parameters. It returns true if the configuration or reconfiguration is done \| Please call this method only from one task context at the time.
GetCustomAttribute_array	Retrieves a custom array attribute (int8/16/32/64 or fp32/64) from the machine learning model, which can be of any data type (requires casting after retrieval) and can be an array.
GetCustomAttribute_fp64	Retrieves a custom floating-point attribute (converted to 64-bit floating point) from the machine learning model.
GetCustomAttribute_int64	Retrieves a custom integer attribute (converted to 64-bit signed int) from the machine learning model.
GetCustomAttribute_str	Retrieves a custom string attribute from the machine learning model.
GetEngineIdFromRef	Retrieves the numeric engine id from its (optional) reference name.
GetInputDim	Retrieve the size of the input dimension from the currently loaded model.
GetMaxConcurrency	Retrieves the maximum number of threads that can concurrently use the loaded model via individual EngineIds.
GetModelName	Retrieves the name of the loaded model.
GetOutputDim	Retrieves the size of the output dimension from the currently loaded model.
Predict	Performs a prediction using the currently loaded model, expecting that the input data has been already written to the input buffer prior to call and output data will be retrieved from the output buffer after prediction.
PredictRef	Performs a prediction using the currently loaded model, expecting that the input data has been already written to the input buffer prior to call and output data will be retrieved from the output buffer after prediction.
Reset	Resets an error
SetActiveEngineOptions	Allows to set special engine options to the currently active engine.

General information

The FB_MllPrediction is a central Function Block for the usage of TC3 Machine Learning in the PLC. The Function Block offers a variety of Methods as described above. Basically, the FB_MllPrediction offers the functionality to load and to execute ML models. Hence, it is an interface to the TwinCAT 3 integrated inference engine (ML Runtime).

Error handling

Note that all methods of FB_MllPrediction return a BOOL which indicates if the execution on the method caused any error, e.g.

bFailed := fbprediction.GetInputDim(nInputDim);

The evaluation of the return value of the methods is equivalent to the evaluation of the Output bError of FB_MllPrediction.

Further, FB_MllPrediction references the TwinCAT 3 EventLogger and thus ensures that information (events) is provided via the standardized interface I_TcMessage. The trace level can be adjusted using TcEventSeverity.

Sample Code

Sample code for the usage of the Function Block is available here.