Research On Mach-zehnder Interferometric Intelligent Optical Fiber Sensor Embeded With TLPFGs

With the rapid development of modernization and information military construction and major state construction projects, optical fiber sensing technology represented by interferometric sensor has unique advantages and has attracted wide attention in fronter defence early warning, special pipe monitoring, tunnel safety supervising, safety engineering monitoring, regional protection and other fields. In this thesis, a interferometric intelligent sensor system was established based on Tilted Long Period Fiber Grating (TLPFG) and Mach-Zehnder interferometer to implement the high-precision and intelligent recognition for invasion event. The main achievements of this thesis are listed as follows:1. Perfectly matched layer(PML) technique and total reflection boundary conditions were applied to establish the tilted long period fiber grating complex mode coupling model, the characteristics of TLPFGs were investigated through the examination of resonant wavelength and minimum transmissivity versus the tilted angle and grating period, which show the TLPFGs was sensitive to pressure. Utilizing the new resonant peak, multi-dimensional measurements for sensing parameters could be realized, therefore, the measurement accuracy could be improved.2. Transmissivity and resonant wavelength of the TLPFGs with different polarizations under specific parameters were analyzed, the intensive polarization dependence of TLPFGs introduced by tilted angle were discussed, such that provide a theoretical basis for the design of polarizing filters and other polarization dependent devices.3. An improved notching method with derivation of power spectrum and mean denoising was presented to realize a sensing system with good frequency response property. Retention of the envelope characteristics and smoothing of the envelope curve is aim to enhance the identity of the slope mutant. And the system relative positioning error was1.5%; polarization controller was further adopted to adjust the polarization of Sagnac interferometric sensor, the results show the system position error was decreased to0.721percent.4. A phase extraction method based on Mach-Zehnder interferometer combined with improved3×3coupler was proposed. With the pre-correction for three channels and demodulation for AC signals, the phase characteristics carrying event information were obtained, reducing the requirements for the symmetry of three-port phase information output by the coupler comparing with the traditional demodulation technology; an improved correlation method for location based on sectioning was proposed to avoid DC drift caused by the circuit noise and polarization change of optical signal, and to improve positioning accuracy for the invasion event.5. A weak signal detection system was developed. Soft threshold wavelet reconstruction was used to restore the useful information from noise, and the FFT high-pass filter was further used to smooth the signal curve. The excellent selection of signal feature was implemented and the identification accuracy was increased.6. A novel model of feature selection based on regression analysis was established and two simplified models of feature selection for two types of events and multiple events were designed, respectively. The single feature vector selecting method utilizing histogram and time-domain characteristic were introduced to achieve the simplified model of feature selection based on histogram by mapping a clutter multi-dimensional sample space to a simple one-dimensional sample space, and then to an ordered multi-dimensional sample space; a model of multi-feature vector selection based on multiple linear regression was established, several dependent variables were set and predictive value of the feature vectors were calculated to achieve significance test and feature selection.7. A pattern recognition model for invasion event was established. Weights were optimized by utilizing finite feedback and recognition arithmetic based on perception standards, the result show recognition accuracy for two types of invasion events and three types of invasion events were91.33%and83.33%, respectively.8. A structure of Mach-Zehnder interferometric optical fiber sensor embedded with TLPFGs was proposed. The theoretical analysis show that the sensor can not only greatly improve the location accuracy for target event; but also can solve the problem of accurate location for multiple events.