Research On Oil Pipeline Leakage Monitoring Method Based On Information Fusion

Pipeline transportation is one of the main methods of transporting products like gas,oil products and so on,which has played an important role in the protection of energy demand of industrial and social and economic development.After the pipeline leakage,the petroleum products will not only bring to the enterprise loss,but also will pollute the environment.It can be concluded that the pipeline leak detection technology has an important significance of saving natural resources,protecting the natural environment and people's life safety and ensuring the safety of state property.At the same time,the development of pipeline leak detection technology could reduce the economic loss,enhance the competitive ability of enterprises.Firstly,this paper studies the method of multi-source signal denoising based on wavelet transform.Wavelet analysis is a signal analysis method based on time domain and frequency domain.This analysis method has adjustable time analysis window and frequency analysis window.In the process of low frequency signal analysis,the time window can be very large,it also has high frequency resolution characteristics.In turn,in the process of analyzing high frequency signals,the time window is very small,the frequency resolution also gets lower.Usually low frequency signals tend to last longer,while high frequency signals last shorter,this happens to correspond to the wavelet transform property,therefore,the wavelet analysis method is suitable for the analysis of normal signals.Then the pressure signal and flow signal are collected to carry out the experiment,the denoising effect was detected.The experimental results show that the wavelet denoising achieves good effect.Secondly,the paper studies the pipeline leak detection method based on the least square support vector machine and the extreme learning machine algorithm multi-source information fusion.The multi-source signal used in the experiment is derived from the experimental data between two workstations A and B of a product pipeline,multi-source information sampling frequency is 20 Hz,the pressure signal sensor and the flow signal sensor collected the continuous signal for 15 hours respectively,200 sets of data were extracted in 12 minutes and 30 seconds.Feature extraction formula is used to extract the data,8 sets of eigenvalues were obtained,it forms a 200 by 8 eigenmatrix,these data include normal signals,regulating valve signals and leakage signals.In turn,the pressure feature matrix,flow feature matrix and fusion feature matrix are input into the least square support vector machine algorithm and extreme learning machine algorithm,training and testing,the result of classification judgment is obtained,and the comparison was made,the experiment result shows that,based on multi-source information fusion method,the accuracy of pipeline leak detection can be improved effectively.