New Smart Models for Minimum Fluidization Velocity Forecasting in the Tapered Fluidized Beds Based on Particle Size Distribution

The most important design parameter
of “minimum fluidization
velocity” in tapered fluidized beds is studied by robust smart
models focusing on the particle size distribution. The smart models
are developed based on the multilayer perceptron (MLP), Gaussian process
regression (GPR), and genetic programming (GP) techniques. For this
purpose, 675 data samples are measured for different geometrical dimensions
and material properties. Basically, it is shown that accounting for
the Gaussian distribution width of the particles in the models improves
their outcomes considerably. The MLP and GPR based-models show excellent
results with average absolute relative error (AARE) values of 0.36
and 0.80%, respectively, for 125 data points used for the testing
process. The data samples are also used for studying the accuracy
of available empirical models, which shows that the recent model proposed
by Rasteh et al. obtains the best outputs among the
others. The developed models also present favorable trends under different
conditions. Finally, in addition to accomplishing the sensitivity
analysis, an explicit and integrated form of empirical correlation
is also proposed by GP with an AARE value of 6.87% for total data
points.