Using the Transport Type option, the target sediment load can be selected.
Using the DataList option, the desired child list for creating the model can be selected. The data in the selected list will be displayed in the Selected DataList section.
To select training and testing data:
- After choosing a DataList, click on the Selection column to specify how each data row will be used (for training or testing).
- Additionally, by right-clicking on a specific data row in the data table, you can manually assign it as training or testing data using the options Select for Train or Select for Test.
- By using the Auto Train/Test Selection option, you can automatically categorize the data into training and testing sets by selecting different algorithms and specifying the percentage of data allocated to each set.
Using the Number of Inputs option, the number of independent parameters can be selected in the input parameters table.
In the ANN Output Unit section, the output unit of the neural network must be specified:
The symbol [*] indicates that the corresponding parameter is dimensionless.
After specifying the output unit of the artificial neural network, the independent parameters (inputs to the model) must be selected in the corresponding parameters table:
By selecting all desired parameters and clicking on the Check Inputs option, the software will prepare all the data in a formatted view and present it to the user. Therefore, the user can copy the provided data and import it into software such as MiniTab to perform statistical analyses and draw necessary conclusions.
The Dropout System option, the Neural Network Layer Configuration Table, and the Artificial Neural Network Settings accessible via the ANN Settings option, are all related to artificial neural networks in the STE software and are explained in detail in the “Artificial Neural Networks” section.
This operation involves three types of configurations, and the results depend on these settings. These configurations are available to the user in the Tools menu, under the Options section.
- Accuracy Settings:
For more information on configuring this section, please refer to the “Accuracy Evaluation Criteria” in the user manual. - Genetic Algorithm Settings
To learn how to configure the genetic algorithm in the STE software, please refer to the “Genetic Algorithm” section of the manual. - Artificial Neural Network Settings
To learn how to configure artificial neural networks in the STE software, please refer to the “Artificial Neural Networks” section of the manual.
By clicking the Train button, the STE software will train the selected neural network using the intelligent genetic algorithm based on the specified settings. The results will then be displayed on a separate page as follows:
By clicking on the plotted graph, the results will be organized in a table and displayed to the user.
By right-clicking on the plotted chart and selecting Export Chart, you can save the chart as an image file.
By right-clicking on the plotted chart and selecting Export Data to Excel, you can export the calculation details to Microsoft Excel.
In the Description section, the user can add notes about the created model in addition to the automatically generated descriptions. By clicking on Save This ANN, the created model will be saved into the project file.
Using the Memory option located on the Artificial Neural Network Training page, trained models can be recalled and deleted.
Selectable Input Parameters for Training Artificial Neural Networks: #
| Parameter | Unit/Description |
|---|---|
| Flow Discharge (Q) | Cubic meters per second (m³/s) |
| Flow Depth (h) | Meters (m) |
| Day | Recorded day of data (value between 1 to 31) |
| Normalized Day | Dimensionless (value between 0 and 1) |
| Binary Day | Binary format (5 binary inputs representing the day will be fed into the neural network) |
| Month | Recorded month of data (value between 1 to 12) |
| Normalized Month | Dimensionless (value between 0 and 1) |
| Binary Month | Binary format (4 binary inputs representing the month will be fed into the neural network) |
| Year | Recorded year of data (integer value) |
If the Transport Type is set to Bed Load (i.e., training the ANN to estimate bed load), the following parameters will also be available for selection:
| Parameter | Unit/Description |
|---|---|
| Suspended Load (Qs) | Kilograms per second (kg/s) |
| Suspended Load (Qs) | Cubic meters per second (m³/s) |
| Suspended Load Concentration (Qs/Qw) | Dimensionless concentration [*] |
| Suspended Load Concentration (Qs/Qw) | Parts per million (ppm) |
