Research Of Concrete Structure Damage Identification Based On Improved SOM Neural Network Using Laser Scanning Technology

In the structural damage identification and monitoring based on laser scanning technology,the structural damage signal is captured by visual sensor,and some algorithm is used to find,locate and even quantify the damage.It has the advantages of real-time on-line monitoring,fast and regular detection,saving human resources and so on.Since the end of 1970s,structural health monitoring has received widespread attention and made remarkable progress,and a variety of structural damage identifications and monitoring methods had been proposed.However,it still faces many challenges,such as many restrictive factors,the accuracy of damage identification model and the sensitivity of damage monitoring index need to be improved.This paper mainly carries out the research on the core issues of structural health monitoring of laser scanning technology:damage identification and damage monitoring.The research focuses on the development of laser scanning equipment without illumination and multiple devices integration.A structural damage identification method which is insensitive to the detection environment but sensitive to damage.A damage networks model that can effectively improve the accuracy of identification.And extract the damage sensitive characteristic indexed to monitor the health state of the structure.The main research contents are as follows:(1)A high precision 3D laser scanning(3D Laser Scanning)system)is developed independently.The 3D laser scanning system is integrated by linear slider module and 2D laser ranging sensor.It controls the fine movement of the sliding screw through the stepper motor to realize the accurate up-down scanning of the laser ranging sensor.The design scheme for laser scanning system is described in detail.The principle of structural measurement and calculation under this system is introduced.The fabrication method of the structural test measurement system is described in detail by adding an one-dimensional transmission actuator to the 2D laser ranging sensor.Based on the structure testing example,the practicability of the scanning system measurement and the accuracy of the scanning results is verified and analyzed.(2)The texture characteristics of damaged images of different dimensions are studied.Combined with the measurement principle of three-dimensional laser scanning system and the properties of damage image obtained by scanning,an effective method to reduce the amount of point cloud data is explored.The method of structural damage image enhancement is introduced in detail.And carry out the related research on the quality evaluation of damage image optimization of concrete structures.(3)The Gray Level Co-occurrence Matrix,which characterizes the texture characteristics of structural damage is established to describe the structural damage quantitatively based on statistical method.Based on the generation criterion,which is the core of GLCM,a method of improving gray level co-occurrence matrix is proposed,and a gray level co-occurrence matrix is constructed to comprehensively analyze the combination form generation criterion.At the same time,the damage characteristics of concrete structures characterized by damage characteristic parameters are discussed deeply based on the improved GLCM.The extraction of damage characteristics and the selection of damage indexes was carried out.(4)The construction method of Self Organizing Map,SOM model for damage identification of concrete structures is introduced.It includes the construction of damage identification network model,learning process and algorithm flow and so on.Based on the SOM topology,a method to optimize the network is proposed.The method is applied and optimized network model to carry out structural damage identification research.(5)Carry out the health state monitoring method based on G-S-G damage state index of concrete structures.Based on the characteristics of damage texture,a damage state index which can identify and monitor structural damage information is proposed.According to the sensitive characteristics and characterization ability of the index,the index is screened and the reliability is verified.Based on the self-developed laser scanning measurement device,combined with the damage index extracted by GLCM,an improved SOM concrete structure damages identification model is established,which effectively identifies the damage type and the damage index is sensitive to the damage state,which is of great significance for the timely detection of early damage in the structure.