Research On Intelligent Evaluation Technology Of Acoustic Emission Detection For Storage Tank Floor Corrosion

With the rapid economic growth of China,the demand and consumption of petroleum energy are increasing continuously.As a specialized storage container for petrochemical products,atmospheric storage tanks are extremely important to ensure their long-term safe and stable operation.Among all parts of the storage tank,the bottom plate is the most susceptible to corrosion and the most difficult to detect.Without opening the tank,conventional non-destructive testing methods are ineffective in detecting and evaluating the tank bottom plate corrosion.At present,acoustic emission testing is one of the main technologies for online inspection of the corrosion state of the storage tank bottom plate,which can realize early detection and defect diagnosis of corrosion.However,there are many and complex acoustic sources in the corrosion process of the storage tank bottom plate,and some key parameters in acoustic emission analysis still rely on experts’ experience,which brings uncertainty factors to the evaluation of the corrosion state of the storage tank bottom plate and hinders the application and dissemination of acoustic emission technology in the storage tank bottom plate corrosion detection.In view of the above problems,this paper studies the identification of acoustic source types and intelligent evaluation of corrosion grades for storage tank bottom plates.Firstly,the acoustic source mechanism of the corrosion on the tank bottom was studied.Based on the Knopoff point stress pulse theory,the intensity of the acoustic source caused by bubble rupture was calculated.The results showed that both bubble rupture and metal dissolution can be used as acoustic emission sources for corrosion on the tank bottom.In addition,acoustic emission monitoring experiments were conducted on Q235 steel,which is a commonly used material for tank bottoms,to obtain corrosion acoustic emission signals caused by bubble rupture and metal dissolution.Three types of interference signals were also collected as noise.Secondly,the acoustic emission signals’ time-frequency parameter features were obtained,and the continuous wavelet transform time-frequency plot,Mel spectrogram,and Gramian angular field transformation were analyzed to examine the feature distribution of different signals.Based on the parameter and image features,BP neural network and improved residual neural network were respectively used for recognition.The results showed that the model based on Mel spectrogram and improved residual network had the highest accuracy in sound source recognition and better noise resistance.It could identify corrosion types effectively,and was utilized for the identification of on-site tank bottom corrosion acoustic emission signals.The identification results indicated that the tank bottom corrosion acoustic emission signals were mainly caused by bubble bursting.Finally,to tackle the problem of scarce and uneven data samples for the corrosion level classification in the existing storage tank bottom plates,a combined approach of Synthetic Minority Over-sampling Technique and grey wolf optimization algorithm-based Support Vector Machine was used for intelligent evaluation.The corrosion level of the tank bottom plate was predicted based on the sound source identification results.The findings indicate that this model can effectively improve the accuracy and reliability of identification for small sample and imbalanced data.