Pipeline corrosion defect prediction method based on optimized neural network

Abstract

The invention discloses a pipeline corrosion defect prediction method based on an optimized neural network. The method comprises the following steps: collecting corrosion defect information in a pipeline, conveying medium conditions, a pipeline body and operation parameters; integrating, cleaning and converting the data to form a corrosion defect data training and testing sample; initializing a BP neural network model; adaptively and dynamically adjusting and optimizing the neural network model based on hidden layer neuron value optimization, a finite storage BFGS algorithm and a learning rate, and obtaining an optimal neural network prediction model through sample training and testing; conveying medium conditions, a pipeline body, operation parameters and the like of a to-be-predicted pipeline are input into the prediction model, and the annular distribution and the size of the corrosion defects in the pipeline are obtained through prediction. According to the method, corrosion influence factors in the pipeline are fully considered, the neural network structure, the storage space, the iteration speed and the stability are optimized, and a reliable reference basis is provided for residual life prediction, corrosion failure risk assessment, repair and the like of the in-service natural gas pipeline.

Description

technical field [0001] The invention belongs to the technical field of oil and gas pipelines, in particular to a pipeline corrosion defect prediction method based on an optimized neural network. Background technique [0002] In the oil and natural gas industry, corrosion usually causes a large number of defects such as thinning and cracks on the surface of the pipeline, which not only reduces the pressure bearing capacity of the pipeline, but also easily causes major safety accidents such as gas leakage or explosion, which seriously threatens the safety of the natural gas pipeline itself. Accurately grasping the location distribution and size characteristics of corrosion defects in natural gas pipelines in service has important guiding significance for the risk assessment of pipeline corrosion failure and the adoption of reasonable repair measures. [0003] At present, the corrosion defects of in-service pipelines are mainly obtained by means of magnetic flux leakage testi...

AI Technical Summary

Problems solved by technology

For the BP neural network, the setting of its structure has not yet formed a complete theoretical guidance system, especially for the number of neurons in the hidden layer, most of them rely on empirical settings. If the number is set too much, the training efficiency will be reduced and over-fitting will occur. convergence phenomenon, and too few settings may cause the algorithm not to converge

In addition, the traditional BP neural network model generally uses the gradient descent method to update the weight value and threshold of the network. Its iterative calculation process often needs to store a large number of high-dimensional matrices, which reduces the convergence speed of the algorithm; and the learning rate is used to adjust the convergence speed of the algorithm. One of the important factors, if the value is too small, it will also reduce the convergence speed of the algorithm, but if the value is too large, it will cause the instability of the neural network model

Images