Research On Debonding And Crack Identification Of Laminate Structure Based On Distributed Optical Fiber

As the aircraft faces complicated aerodynamic heating during service,the structure must be thermally protected to ensure the normal operation of the structure.Thermal protection materials are usually bonded to the metal skin of aircraft.Faced with ultra-high-speed environments during service,stress cracking may occur on the surface,and the bonding interface is prone to delamination,resulting in a decrease in structural performance and even serious consequences.Therefore,in order to ensure the safety and stability of the thermal protection structure,damage monitoring is required.The distributed optical fiber sensor is small in size,and the embedded structure hardly affects the mechanical properties of the measured structure.With its strong anti-interference performance,it can meet the measurement requirements of high density and high sensitivity.So it can be used to monitor the thermal protection structure.Because of the large amount of strain data collected by the distributed optical fiber sensor,manual analysis is difficult.Deep learning algorithms can analyze ultrahigh data volumes,thus providing the possibility of identifying high-density data.Aiming at the thermal protection bonding structure,this paper takes the laminated board structure formed by bonding phenolic resin board and aluminum alloy board as the research object.First,the two classic damage modes(debonding and cracking)of the laminated board are theoretically deduced and numerically analyzed.Through simulation,the strain field of the laminated plate model under the cantilever beam is obtained,as well as the strain characteristics of debonding and cracking.Then based on the principle of distributed optical fiber sensor,the embedded distributed optical fiber sensor is used to carry out prefabricated debonding and cracking.Cantilever beam loading test of damaged laminate specimens.Combining the strain characteristics of the aforementioned different types of damage,the strain data collected by the sensor is analyzed,and the damage location is successfully realized and the two types of damage are identified.Finally,a convolutional neural network is used to construct a one-dimensional neural network model suitable for high strain data.The strain data in the experimental study is trained to identify the strain data of the last load level,and finally the damage location and classification are realized.The work in this paper is of great significance for batch processing of high strain data and identification of debonding and crack damage through strain data,and provides a reference for intelligent health monitoring.