110
W. Xinying et al.
10.4 DBN-Based Pipeline Corrosion Residual Life
Prediction Model
Establish a deep confidence prediction model for the residual life of corrosion
pipelines on Matlab platform. Select sensitivity count, RMS voltage, rise time, amplitude, absolute energy, ring count, duration, peak count, and experimental corrosion
time and/constructs a deep confidence pipeline residual life prediction model.
Under laboratory conditions, the acoustic emission technology and DBN algorithm are used to predict the corrosion rate of the gas pipeline. The specific steps are
as follows:
(1) Acoustic emission detection data and recording time at different times during
the pipeline corrosion process are collected and divided into test sets and training
sets.
(2) Establish a DBN-based pipeline residual life prediction model.
(3) Input the training sample into the established DBN-based pipeline residual life
prediction model, input data to the first RBM, and train the output data as the
input data of the next layer, and then cycle until All RBM learning is done.
(4) According to the tagged data and classification error in the training sample, the
parameters are fine-tuned from the top layer to the lowest layer in the DBN
model until the training of the entire DBN model is completed.
(5) Input the test sample into the trained DBN classification model for classification
performance test.
(6) The output pipe corrosion rate prediction result, combined with the time and
the pipe specification thickness to obtain the remaining life.
10.5 Implementation of Deep Confidence Model in Pipeline
Corrosion Prediction
Acquire acoustic emission signal data of a pipeline under different corrosion conditions [11]. The test pipe size is X720 mm, X10 mm, and the initial wall thickness
of the oil and gas pipeline is 8.0 mm. The section of the pipe is partially immersed
in a low-concentration acidic solvent to accelerate the simulated corrosion, and the
acoustic emission detection is performed after a certain time.
Time domain analysis was carried out on the corrosion acoustic emission signal
of the oil and gas pipeline. According to the results of time domain analysis, the
frequency domain analysis results of acoustic emission signals with other uncorroded areas show that under different corrosion states, the acoustic emission signal
frequency of the corrosion state is very different, so it can be based on the frequency
information of the acoustic emission signal. Judging the corrosion state of the oil and
gas pipeline, the corrosion of the oil and gas pipeline can be successfully detected
[12], and the specific corrosion state and speed of the oil and gas pipeline can be determined according to the analysis result of the acoustic emission signal. The corrosion
W. Xinying et al.
10.4 DBN-Based Pipeline Corrosion Residual Life
Prediction Model
Establish a deep confidence prediction model for the residual life of corrosion
pipelines on Matlab platform. Select sensitivity count, RMS voltage, rise time, amplitude, absolute energy, ring count, duration, peak count, and experimental corrosion
time and/constructs a deep confidence pipeline residual life prediction model.
Under laboratory conditions, the acoustic emission technology and DBN algorithm are used to predict the corrosion rate of the gas pipeline. The specific steps are
as follows:
(1) Acoustic emission detection data and recording time at different times during
the pipeline corrosion process are collected and divided into test sets and training
sets.
(2) Establish a DBN-based pipeline residual life prediction model.
(3) Input the training sample into the established DBN-based pipeline residual life
prediction model, input data to the first RBM, and train the output data as the
input data of the next layer, and then cycle until All RBM learning is done.
(4) According to the tagged data and classification error in the training sample, the
parameters are fine-tuned from the top layer to the lowest layer in the DBN
model until the training of the entire DBN model is completed.
(5) Input the test sample into the trained DBN classification model for classification
performance test.
(6) The output pipe corrosion rate prediction result, combined with the time and
the pipe specification thickness to obtain the remaining life.
10.5 Implementation of Deep Confidence Model in Pipeline
Corrosion Prediction
Acquire acoustic emission signal data of a pipeline under different corrosion conditions [11]. The test pipe size is X720 mm, X10 mm, and the initial wall thickness
of the oil and gas pipeline is 8.0 mm. The section of the pipe is partially immersed
in a low-concentration acidic solvent to accelerate the simulated corrosion, and the
acoustic emission detection is performed after a certain time.
Time domain analysis was carried out on the corrosion acoustic emission signal
of the oil and gas pipeline. According to the results of time domain analysis, the
frequency domain analysis results of acoustic emission signals with other uncorroded areas show that under different corrosion states, the acoustic emission signal
frequency of the corrosion state is very different, so it can be based on the frequency
information of the acoustic emission signal. Judging the corrosion state of the oil and
gas pipeline, the corrosion of the oil and gas pipeline can be successfully detected
[12], and the specific corrosion state and speed of the oil and gas pipeline can be determined according to the analysis result of the acoustic emission signal. The corrosion