Modeling of the velocity of an instrumented PIG using artificial neural networks.
PIG, neural networks, velocity.
The passage of a PIG is a technique quite used in inspection of big length and principally buried pipes using the pressure differential on it to impulse itself. But, during the inspection, one of the problems that may occur is the stop of the PIG because of severe incrustations or fabrication/installation defects of the pipes, doing the halt of the instrument and its posterior release with high velocities due to the accumulate of pressure at back part. This work purpose the use of neural networks in order to model the relation between the differential pressure on the PIG and its velocity during your path in the tube. Therefore, it was used a supervisory system to capture the pressure data along the test pipe and an odometer coupled to the PIG for the velocity data. It was considered two neural network models, in the case the MLP and NARX networks, the latter being a recurrent network. The training and validation results showed that the models by neural networks were efficient to estimate the velocity of the PIG.