Research On Damage Recognition And The Influence Of Pitting Corrosion On Mechanical Properties Of Hull Structural Plate

Marine structures,such as ships,have long been in a harsh marine environment,and will inevitably suffer from alternating loads,stress concentrations,corrosion,fatigue,etc.The surface of structures is prone to varying degrees of damage.Damage can lead to degradation of mechanical properties,affectting the service life of the structure,and even catastrophic accidents.Therefore,it is extremely important to detect and recognize damage on the surface of structures.Based on the development of computer vision technology,this paper attempts to detect and recognize the surface damage of hull structural steel plate by convolutional neural network(CNN)for the first time,and studies the influence of multi-parameter pitting corrosion on steel plate's mechanical properties.Main researches are as follows:The research status of surface damage detection is reviewed.The advantages and disadvantages of traditional non-destructive testing technology and machine vision detection technology are compared and analyzed.The research status of convolutional neural network based on visual technology and influence of pitting corrosion on mechanical properties of hull structural plate are expounded.The basic structure of the convolutional neural network,the network training process,the optimization algorithm used in the training process and the over-fitting method are elaborated.These provide theoretical basis and practical guidance for the following surface damage detection of hull structural plate using convolutional neural network.Firstly,the AlexNet for SCDI network was used to train the corrosion classification model.The Overlap-Scanning sliding window algorithm was used to detect and recognize the surface damage.The network model was improved and the Inception was added to realize the detection and recognition of polymorphic damage of hull structural plate.The generalization ability of the model was discussed,and it can be concluded that increasing images under light changing,images with different degrees of blur and images with shaded area can improve the generalization ability of the model.The size and shape of the sliding window can also be resized to improve the performance of the damage detection classifier model.Finally,the detection and recognition process was summarized,and the damage detection mode and crowdsourcing mode were proposed.The damage detection research method in this papercould provide a new idea and guidance for the detection in the ship field,and promote the development of damage detection in the ship field to the direction of artificial intelligence.Aiming at hull structural plate subjected to pitting corrosion damage,the influence of multi-parameter pitting on the mechanical properties was studied by mechanical drilling to simulate pitting corrosion and uniaxial tensile test.The damage degradation laws of elastic modulus,yield strength,ultimate strength and peak strain under different pitting damage degrees were studied respectively.The bilinear generalized stress-strain curve constitutive model and linear-exponential generalized stress-strain curve constitutive model were established respectively.The bilinear generalized constitutive model is simple,and the linear curve model is used to replace the nonlinear segment,which has a certain safety margin.The linear-exponential generalized constitutive model overcomes the shortcomings of the bilinear generalized constitutive model for inaccurate fitting of inelastic segments.The linear-exponential model can provide an accurate and simple constitutive model for structural safety studies based on the degree of different pitting damage under the influence of multiple parameters.The influence of multi-parameter pitting corrosion on the mechanical properties provides a new way to study the effect of pitting corrosion.