Research On Structural Health Monitoring Method Based On Multi-Sensor Data Fusion

At present,structural health monitoring technology has been widely used in the field of aviation,which is of great significance for the safe use of aircraft.As a technology gradually moving from the laboratory stage to practical application,structural health monitoring technology is faced with many challenges,such as large area of actual engineering structure monitoring,complex configuration,multiple types of sensors arranged,and the information obtained by each sensor is local and incomplete.Therefore,the urgent problem is how to fuse the information of different types of sensors to make a comprehensive and effective assessment of the health status of the structure,so as to obtain more accurate life prediction results.In order to fully monitor the health of the structure,this paper uses piezoelectric transducers(PZT)for active monitoring and resistance strain gauges for passive monitoring,combines the passive monitoring method that can continuously monitor structural damage and the active monitoring method sensitive to minor damage to comprehensively monitor the damage,thereby improving the monitoring accuracy and achieving more accurate structural damage identification and life prediction.Firstly,the related theory of Lamb wave is introduced,the dispersion equation is solved,the group velocity and phase velocity curves of Lamb wave are drawn,and the propagation characteristics of Lamb wave are obtained.This provides a theoretical basis for the determination of the center frequency of the Lamb wave excitation signal in the experiment.Then introduced the crack propagation mechanism and strain-based monitoring principle,the monitoring experiment plan of active and passive fusion is determined.Piezoelectric transducers and strain sensors were used to monitor fatigue crack growth and bolt loosening.Extract the damage characteristic parameters of Lamb wave S0 mode data and strain data,and establish the relationship between them and the damage degree respectively.Secondly,the pattern recognition algorithm of the two types of damage is studied.Back Propagation(BP)neural network algorithm,Deep Belief Network(DBN)algorithm and Random Forest(RF)algorithm are used for pattern recognition.The extracted Lamb wave and strain characteristic parameters were taken as two sets of samples respectively,and the pattern recognition study of single parameters was carried out based on three algorithms.Comparing the pattern recognition results of the three algorithms,the random forest algorithm has the highest pattern recognition accuracy.In order to further improve the accuracy of pattern recognition,give full play to the advantages of active and passive monitoring methods,reflect damage information from multiple angles,and perform data fusion on the recognition results of the two sensors.Taking the recognition results of the strain sensor and piezoelectric transducer obtained by the random forest algorithm as fusion objects,D-S evidence theory and fuzzy logic theory are used for decision-level fusion.After data fusion,a more accurate and reliable identification result is obtained than a single sensor.Comparing the results of the two fusion methods,it can be seen that the recognition accuracy after fusion by D-S evidence theory is higher.Finally,the life prediction of the structure is made.Finally,the life prediction of the structure is made.Based on the single feature parameters,the particle filter(PF)algorithm is used to predict the life of the structure,and then the prediction results are fused by the weighted average method to obtain the final prediction result.At the same time,a random forest algorithm is used to predict life based on multiple feature parameters.By comparing the prediction results of two different methods,we can see that the prediction results of the particle filter algorithm are more accurate.