A deep learning with knowledge distillation scheme for lateral lane-level vehicle identification based on ultra-weak fiber Bragg grating (UWFBG) arrays is proposed. Firstly, the UWFBG arrays are laid underground in each expressway lane to obtain the vibration signals of vehicles. Then, three types of vehicle vibration signals (the vibration signal of a single vehicle, the accompanying vibration signal, and the vibration signal of laterally adjacent vehicles) are separately extracted by density-based spatial clustering of applications with noise (DBSCAN) to produce a sample library. Finally, a teacher model is designed with a residual neural network (ResNet) connected to a long short-term memory (LSTM), and a student model consisting of only one LSTM layer is trained by knowledge distillation (KD) to satisfy the real-time monitoring with high accuracy. Experimental demonstration verifies that the average identification rate of the student model with KD is 95% with good real-time capability. By comparison tests with other models, the proposed scheme shows a solid performance in the integrated evaluation for vehicle identification.

Intrusion identification has been an intractable task for perimeter security. One of the primary challenges is to possess high identification rate over a long-distance range monitoring. This paper proposes an intrusion identification scheme based on ultra-weak fiber Bragg grating (UWFBG) arrays. The scheme is acquired by the combination of a Gaussian mixture model (GMM) and a hidden Markov model (HMM). The time dependencies are obtained by the analysis of relevant sensors in UWFBG arrays from the procedure of intrusions. The features extracted from vibration signals with time dependencies are used as the input of GMM-HMM. The GMM-HMM simultaneously analyzes features and time dependencies to identify intrusion. Experimental demonstration verifies that the proposed scheme can identify three intrusions (walking, knocking and climbing) and two non-intrusions (heavy truck passing and wind blowing) with the average identification rate of 98.2%. By the comparison tests with other six classifiers, the proposed GMM-HMM scheme shows a solid performance in the integrated evaluation for intrusion identification.

By fixing two FBGs on the surface of a rotating shaft along the direction of ±45° and using dynamic wavelength demodulation technology, we propose an optical fiber sensing system to monitor the driving torque and torsion angle of the rotating shaft. In theory, the dependence relation of the dynamic difference of central wavelengths on the torque and torsion angle of the rotating shaft has been deduced. To verify an optical fiber sensing system, a series of sensing experiments have been completed and the measured data are approximately consistent with the theoretical analysis. The difference of two central wavelengths can be expressed as the sum of two parts: a “DC” part and a harmonic “AC” part. The driving torque or torsion angle is linear with the “DC” part of the difference of two central wavelengths, the harmonic “AC” part, meaning the torsion angle vibration, illustrates that periodic vibration torque may be caused by inhomogeneous centrifugal forces or inhomogeneous additional torques produced by the driving system and the load.

For long-distance unattended metro tunnel monitoring systems, accurate identification of abnormal disturbances is of great significance for underground safety. Many abnormal disturbance data will be generated in the ultraweak grating sensing system along the metro tunnel. The classification and labeling of these abnormal disturbance signals is a very large and time-consuming task. Therefore, we propose an active learning method using a transductive relevance vector machine based on a Gaussian mixture model (GMM-TRVM) to classify and identify those signals. First, we analyzed the dynamic response signals of fiber Bragg grating sensors in the metro tunnel. Then, we extracted the time-domain and frequency-domain features of the abnormal disturbance signal. Taking the characteristics of abnormal disturbance signals as input and corresponding disturbance categories as output, the training of the GMM-TRVM model was completed under the active learning framework. In the circular query process, the model was evaluated after each update, and the evaluation index was assigned an F1 score. The final results show that the proposed GMM-TRVM active learning method achieves good classification and recognition.

The foundation of an intelligent highway network is the construction of a high-density distributed strain monitoring system, which is based on sensing elements that can sensitively capture external information. In this research, the development and application for the structure of a novel strained optical fiber cable based on the weak fiber Bragg grating (wFBG) arrays are discussed. A modulation and demodulation solution of wavelength division multiplexing combined with time division multiplexing is developed by utilizing the property by which the wavelength of the strained optical fiber cable is periodically switched. Further, the strain transfer model of the optical cable is analyzed hierarchically using the theory of elasticity. The strain transfer coefficients of the overhanging region and the gluing region are combined to deduce the sensitivity model of the strained optical fiber cable. Moreover, the finite element technique is integrated to optimize the structural parameters of the optical cable for high-sensitivity or large-scale range. The strained optical fiber cable based on wFBG arrays is applied to a steel-concrete composite bridge. The static and dynamic loading tests show that the sensing optical cable can be monitored for strain variation in order to realize the functions of lane identification, weighing vehicle tonnage as well as velocity discrimination.