Research On The High Speed Demodulation Of The Sensing Signal Of Bridge Health Monitoring Based On FBG

Bridge health monitoring is an important means of building infrastructure maintenance.Once the bridge structure was damaged,the monitoring system would generate an abnormal warning signal.Due to the advantages of the fiber Bragg grating(FBG)sensor,such as low cost,small volume,anti-electromagnetic interference and high precision,it becomes a hotspot in the field of bridge health monitoring.Signal demodulation is a key link of FBG sensing system.Usually,the sensing signal contains some noise or some datas might be lost due to the interference of environment or aging of equipment,that will seriously affect the health and safety assessment of structure.Therefore,both the signal de-noising and the data recoverying are necessary.The study possesses a certain academic significance and application value.The effect of noise on demodulation is analyzed.To address the problem of pseudo modes generated by EMD decomposition,the pseudo modes cancellation is introduced.To solve the problem of selecting effective components,mutual information is used to estimate the critical point of high and low frequency components according to the frequency domain characters of signal and noise.So that a precise self-adaptive algorithm of selecting relevant modes based on Empirical Mode Decomposition(EMD)and mutual information is proposed.The influence of data loss on demodulation accuracy is analyzed.According to the case of data loss of static FBG spectrum signal,a repaired method based on sparsity adaptive reconstruction algorithm is proposed.For the problem of data loss on FBG vibration signal,a data recovery method based on learning dictionary is presented.For the issue of that sparsity should be a priori condition in compressed sensing,an adaptive threshold function matching the characteristics of FBG signal is proposed.The suitable observation matrix and sparse dictionary are selected according to the characteristics of signal,so as to improve the precision of data recovery.To solve the problem that complex characteristic vibration signals can not be adequately represented by orthogonal base,a method of establishing dictionary is presented.The Euclidean distance is introduced to obtain a training set that is similar to original signal,after then the AK-SVD algorithm is used to obtain adaptive dictionary.The data acquisition system is established to collect the spectrum signal in different temperature.The proposed noise de-noising algorithm and data recovery method are used to process the signal.Experiments and simulations validate the efficacy of these two methods.At last,the temperature calibration and error analysis are completed to verify the feasibility and effectiveness of proposed methods.