Multi_R_Designer Tutorial

---

MEXA is a tool helping the user to find the best model predicting hydrodynamics, mass or heat transfer properties providing a range of operating conditions such as velocities, fluid properties and packing properties (common model inputs). If the user wants to predict, for example, total liquid holdup for specific operating conditions, he can assign a range for the liquid and gas velocities, particle dimensions and choose one or several models available in the Multi_R_Designer simulator. MEXA will find the number of experimental data in the Multi_R_Designer integrated databank matching the specified conditions and will divulge the performance of the selected models on these data. With this tool, the user can determine the model or correlation exhibiting the best prediction capability of the studied property for the operating condition setup.

In the example presented here, we will determine the best model predicting the total liquid holdup in trickle beds for a set of operating conditions which can be either stored in a file or introduced in the MEXA using the data range table.

If the first option (operating conditions stored in a file) is used for MEXA inquiry, it can be done after a simulation.The appropriate procedure for creating a user's file and carrying out simulations is presented in sections Frame for user's file and manual simulation respectively. Once the simulation is done, start a MEXA inquiry from STEP 4 below.

If the inquiry is based on a range of operating conditions, the following steps are used:

STEP 1:

• Select Simulation/Start new simulation from the main menu . The user is introduced to the Simulator: Trickle bed dialog window. The red space bar at the bottom of the dialog window warns that no model is selected. For Total liquid holdup in the Model field, 0/15 means that none of the 15 available models are selected.

STEP 2:

• Click on the pencil  in the Model field. Models available to predict total liquid holdup are depicted in a Total liquid holdup dialog window.
• All models can be selected by clicking on the Use heading while the check boxes can be used to select only some models. In this example, all available models are considered. TR_ANN08 is the neural network correlation developed at Laval University for predicting total liquid holdup. Symbol N is the number of experimental data in the Multi_R_Designer integrated databank on which the models were simulated, andassociated to the models refer to the Average Absolute Relative Error (AARE) and standard deviation on absolute relative error, respectively.

STEP 3:

• Information about the selected model is available in Info.
• In the General tab for TR_ANN08 (figures below), the model reference (Name) is given as well as the Type (Neural network correlation), the model Input list in dimensional form (liquid density, gas density ...) and the Output also in dimensional form.
• In an Applicability tab, the domain of validity for each model input is given in the unit of your choice. All these information can be printed or saved in a .html file by clicking on Export. These information gives a quick overview of the models in order to determine if they are applicable based on te studied operating conditions (i.e. if they are valid only for high interaction flow regime or non-Newtonian fluids).

STEP 4:

• Select MEXA at the bottom of the Simulator:TrickleBed dialog window. The list of models selected in STEP 2 is given and the user can start its investigation to determine the best model/correlation predicting total liquid holdup regarding the specified operating conditions.
• Select a model (TR_ANN08 for this example) and click on Process.

STEP 5:

• The user is introduced to the MEXA: TrickleBed dialog window. The model's inputs in dimensional form are given with their range of variation within the simulation file. Otherwise,  introduce the minimum and maximum values for each variables of interest. First, specify the unit for each variable in the Unit list, then write the Min and Max values. The number of experimental data in Multi_R_Designer integrated databank fulfilling the specified range is displayed in the corresponding N Pts entry.
• For a liquid density varying between 900 to 1200 kg/m³, MEXA has found 6239 experimental data which lies within this range in the Multi_R_Designer integrated databank. Resulting data count gives the resulting number of data sets in the Multi_R_Designer integrated databank matching all input specification ranges (Liquid density, Liquid viscosity, Surface tension, Liquid superficial velocity, Gas density, Gas viscosity, Gas superficial velocity, Diameter equivalent to volume sphere, Column diameter, Sphericity factor, Bed porosity).
• In this example, 45 experimental total liquid holdup data which fulfils all input specifications are found. Obviously, this number would become greater if less inputs were specified (deselect the inputs using the check boxes in the Use field).

• Dimensionless inputs are given, when possible, in the Dimensionless tab. For the TR_ANN08 correlation, there is ten dimensionless numbers as inputs and 60 experimental total liquid holdup data matching the dimensionless input specification ranges.

STEP 6:

• The parity plot of the simulated versus experimental total liquid holdup found in the Multi_R_Designer integrated databank can be visualized for each type of selection (dimensional or dimensionless) by pressing on Parity.
• To have the cumulative distribution error of the selected model on these experimental data, press on Error CDF.
• For the statistic, press on Statistics tab.

To compare the models performances, repeat the procedure with a new model/correlation starting from STEP 4: "select a model and click on Process"  with the same variable range specifications.

STEP 7:

• The variation of the extracted total liquid holdup data versus operating conditions can also be visualized by selecting the Plot button at the bottom of the MEXA: Trickle bed dialog window, where the selection of the variable range specifications is done (STEP 5). Other experimental properties in Multi_R_Designer matching the range constraints specified above can also be graphically represented.

Multi_R_Designer Tutorial