Research On Recognition Algorithm Of Adhesion State Of Anticorrosive Coating Of Plate Structure Based On Ultrasonic Guided Wave

The plate structure with anti-corrosion coating has been widely used in large equipment,aeronautics and astronautics,solid rocket motor housing.In order to ensure the safe and reliable operation of the attached anti-corrosion coating plate structure,it is very important to identify the operation state of the coating by using the detection method based on the integrity of the anticorrosive layer structure.Ultrasonic has the advantages of fast speed,strong penetration and high sensitivity,and is an effective method to detect debonding defects of plate structure,and it can detect the position and size of debonding defects quickly and accurately.Therefore,the detection of debonding defects of plate structure based on ultrasonic guided wave has become an important research direction in the field of nondestructive testing.In this thesis,the propagation model of ultrasonic guided wave in plate structure is established,and the propagation characteristics of guided waves in double layer bonding medium are analyzed.The multimodal,dispersion and energy transfer characteristics of guided waves are analyzed theoretically.The change of material parameters of double layer plate will cause the change of bonding state,and then the echo signal transmission,dispersion and energy transfer will be changed.The ultrasonic echo signals of the board structure in good bonding condition,weak bonding state based on density change and partial debonding state based on thickness change are obtained.The wavelet lifting algorithm based on soft threshold is used to denoise the signal.The echo signal is analyzed from the three aspects of the maximum amplitude,energy and time-frequency distribution of the signal,so as to establish the relationship between material parameters and echo signal The corresponding relationship of changes.The multi-dimensional feature extraction is carried out for the acquired ultrasonic echo signals,which respectively extract the maximum amplitude,waveform coefficient,kurtosis coefficient and deviation coefficient of the signal in the time domain,the center of gravity frequency and the frequency standard deviation feature,the signal energy,wavelet energy spectrum entropy and wavelet singular spectrum entropy in the time frequency domain.Based on the extracted 9-dimensional feature,the thesis classifies the bonding state of plate structure by using naive Bayesian classifier and random forest classification algorithm,and compares the recognition rate of the two classification algorithms.The results show that the change of the dielectric parameters of the plate bonding structure will cause the change of the bonding state,and then cause the change of the ultrasonic echo signal.The characteristic quantity of ultrasonic echo signal is used to express the change of bonding state,so as to establish the corresponding relationship between material parameters and echo signal.Based on the 9-dimensional features extracted from time domain,frequency domain and time-frequency domain,the recognition rate of bonding state using naive Bayes classifier is 94%,and that using random forest algorithm is 92%.