Mapping of Pipe Wall Thinning with a Wireless Sensor Network
Affiliation Type:
Academia
Keywords:
Mapping of Pipe Wall Thinning with a Wireless Sensor Network
Abstract:
Wireless sensor networks (WSN) have been shown to be effective in monitoring industrial equipment and structures. The analysis of data from such networks may depend on the operator of the network. We present a neural network (NN) based approach which uses data from an ultrasonic guided wave WSN to produce reliable reconstructions of wall-thinning in a metal pipe. Data from accelerated wall-thinning experiments was recorded with a WSN consisting of eight sensors with piezoelectric transducers. Sections of a process pipe were progressively grinded to simulate wall-thinning. The A0 mode of Lamb waves was utilized due to its sensitivity to the thickness of the waveguide. To increase the amount of information available, helical propagation paths were used. Wall-thickness maps were reconstructed using a NN that was trained on simulated data. The training data was produced by randomly generating wall-thinning maps and transforming them to group velocity maps using a dispersion curve for the pipe wall. The measured shifts in wave packet ToA were then input to the NN that is able to output the wall-thickness map within milliseconds.
