An FPGA layered cooperative adaptive optical fiber signal hardware processing method based on global light field modeling

Abstract

The application discloses an FPGA layered cooperative adaptive optical fiber signal hardware processing method based on global light field modeling and belongs to the technical field of distributed optical fiber sensing and FPGA / GPU optoelectronic monitoring hardware. The application constructs a global dynamic time sequence light field mathematical model by relying on optical fiber optoelectronic collection hardware, carries a causal deduction operator in the hardware to complete multi-source light path disturbance cause disassembly; adopts a layered parallel hardware architecture of six types of FPGA operator units to realize multi-path light path task synchronous processing; builds a coupling system of 17 light path core indexes to comprehensively evaluate the light path health; adaptively matches the optimal simulation coefficient through the power and precision constraint equation; and realizes hardware parameter closed loop self-calibration by relying on the measured error gradient descent. The application solves the problems of traditional optical fiber monitoring algorithm fragmentation, parameter static, interference indistinguishable, power imbalance, single evaluation dimension and no independent correction closed loop, and is suitable for optoelectronic monitoring scenes such as rail transit, oil and gas pipeline and laser precision detection.

Description

Technical Field

[0001] The invention belongs to the technical fields of distributed optical fiber sensing, optical fiber optical signal demodulation, photoelectric monitoring equipment, and FPGA / GPU hardware computing power adaptive optimization, and can be applied to photoelectric engineering scenarios such as rail transit optical cable vibration security, laser linewidth precision detection, multimode optical fiber crosstalk suppression, optical path fault hardware diagnosis, and high-precision calculation of long-distance signal delay. Background Art

[0002] Current signal processing algorithms matched with distributed optical fiber sensing monitoring hardware have six inherent technical defects. The existing published patents for single-point processing of optical fiber phase and time delay only target local optical path signal calculation, and cannot systematically solve the comprehensive problems of long-term operation of hardware equipment: 1. Fragmented optical path signal processing modules, no unified global mathematical model The demodulation, time delay calculation, noise filtering and vibration recognition programs carried by traditional optical fiber monitoring hardware are independent of each other, and can only perform local numerical operations on a single optical path physical quantity; when the hardware synchronously collects multi-channel coupled signals of temperature, mechanical vibration, polarization and transmission loss, the superposition of multiple fields causes continuous accumulation of calculation error. The industry lacks a unified time-sequence optical field mathematical framework adapted to photoelectric acquisition hardware, and cannot realize global cooperative calculation.

[0003] 2. No mechanism-based traceability capability for mixed signals, high false alarm and missing alarm rates of equipment Existing optical fiber hardware only outputs optical path calculation values, and does not have operation logic for distinguishing disturbance causes, and cannot automatically separate real deformation faults of pipelines / tracks, ambient temperature drift, optical path mode crosstalk, and instrument background noise, resulting in insufficient alarm reliability of long-term field monitoring equipment.

[0004] 3. Static solidification of operation parameters, no adaptive correction logic for hardware Traditional photoelectric monitoring equipment adopts fixed filtering windows, fixed judgment thresholds and static operation formulas; after optical cable aging, seasonal temperature change and transmission distance change, the calculation accuracy of hardware continues to decay, and there is no closed-loop operation process that reversely corrects parameters based on measured data.

[0005] 4. Inherent contradiction between high-precision calculation and hardware computing power overhead High-resolution phase demodulation and 100-kilometer-level optical cable positioning consume extremely large FPGA and GPU computing power. Existing equipment can only choose one of the two options: low computing power hardware matches low calculation accuracy, while high-precision hardware brings high procurement and operation costs. The industry lacks a mathematical constraint model to dynamically balance the overhead of the two.

[0006] 5. Single dimension of optical path health assessment, lack of refined hardware diagnosis capability Traditional fiber optic equipment relies on only two or three physical parameters, such as phase and delay, to determine faults. It cannot simultaneously quantify multi-dimensional indicators such as positioning error, signal-to-noise ratio, laser linewidth deviation, equipment aging degree, long-term signal drift, and hardware power consumption, making it difficult to achieve full-dimensional quantitative diagnosis of optical path health.

[0007] 6. Open-loop operations lack an autonomous optimization mechanism, and the equipment accuracy cannot continuously converge. The existing photoelectric monitoring hardware uses a one-way open-loop operation process. The error data generated by a single measurement cannot be fed back to the operation module to update the internal parameters, and there is no room for positive improvement in the measurement effect of the equipment during long-term operation.

[0008] The existing general cross-domain mathematical deduction approach is recorded in the applicant's prior application "Cross-domain Digital Intelligence Interpretation and Creation Method Based on Dual Series Formulas". This solution only discloses the general quantitative deduction logic and does not involve fiber optic photoelectric acquisition hardware, optical path-specific optical field model, seventeen optical path-specific evaluation indicators, or optical equipment computing power constraint mechanism. It cannot solve the unique hardware technology pain points of the fiber optic monitoring industry and provides a basic mathematical approach reference for this invention.

[0009] In summary, the industry urgently needs an integrated optical path signal processing solution that is mounted on fiber optic sensing hardware and features unified global optical field modeling, multi-interference causal tracing, multi-task hardware parallel computing, multi-index coupling quantization, dynamic computing power adaptation, and closed-loop self-optimization. Summary of the Invention

[0010] (a) Purpose of the invention To address the shortcomings of existing fiber optic monitoring hardware, such as fragmented algorithms, fixed computational parameters, inability to trace mixed interference, rigid computing power matching, single optical path evaluation dimension, and lack of autonomous parameter optimization closed loop, this paper proposes a hierarchical collaborative adaptive fiber optic signal processing system and method based on prior public general cross-domain mathematical deduction ideas.

[0011] This invention adapts to distributed fiber optic sensing and acquisition hardware to construct a unified mathematical model of dynamic temporal optical fields; it builds a hierarchical parallel hardware computing unit based on FPGA to achieve synchronous processing of multi-path optical tasks; it relies on causal inference operators to complete the decomposition of multi-source disturbance mechanisms; it builds a coupling and quantification system of seventeen optical path-specific indicators to output a comprehensive health score of the hardware optical path; it solves the optimal simulation accuracy under the current working condition through the cost-quality constraint equation specific to the fiber optic scenario; and it uses the gradient descent of hardware measured error to update the indicator weights in reverse, forming a complete adaptive optimization closed loop of "hardware acquisition-computation and measurement-error feedback-parameter correction", solving the core pain points of traditional fiber optic monitoring equipment such as poor accuracy, wasted computing power, weak anti-interference in the field, and long-term operation accuracy decay.

[0012] (II) Definitions of Terms and Supporting Mathematical Formulas 1. Global dynamic light field OptField(t) OptField(t)=F(\varphi(t),P(t),T(t),V(t)) In the formula: φ(t) is the optical path timing phase quantity, P(t) is the polarization parameter, T(t) is the ambient temperature sequence, and V(t) is the optical cable vibration intensity sequence.

[0013] 2. Optical path causal inference operator: By performing a difference counterfactual comparison operation on two sets of synchronously acquired optical field time-series signals, fault characteristic signals and environmental interference signals are separated to achieve source tracing of disturbance causes.

[0014] 3. Weighted Coupling Scoring Formula for Seventeen Optical Path Indicators Score=\sum_{i=1}^{n} w_i \cdot S_i In the formula: w_i is the weight of the i-th optical path index, with a value range of 0.01 \le w_i \le 2.0; S_i is the normalized score of a single optical path index.

[0015] 4. Computing power and measurement quality constraint functions The quadratic cost function for computing power is: C_{opt}(\rho)=a\rho^2+b\rho+c The linear quality function for optical path measurement is: Q_{opt}(\rho)=k\rho+m \rho∈[0,1] represents the normalized simulation accuracy coefficient, and the optimal accuracy is solved with hardware computing power budget as a constraint. In the formula, a, b, c, k, and m are all empirical constants obtained by fitting long-term field measurements and hardware benchmark tests.

[0016] (III) Core Technology Principles (All Bound to FPGA / GPU Optoelectronic Hardware Carriers) 1. Unified Modeling Mechanism for Global Dynamic Light Field Based on the acquisition of multiple raw signals from fiber optic sensing hardware, the signals are mapped to the normalized simulation accuracy range of 0 to 1. The OptField(t) timing equation is used to uniformly represent the signals of multiple physical coupled optical paths, providing a unified mathematical input basis for FPGA parallel computing.

[0017] 2. Multi-source interference causal counterfactual inference mechanism By incorporating a causal inference operator into the FPGA hardware, the differential comparison operation is performed on the hybrid optical path signal synchronously acquired by the hardware, which automatically distinguishes between equipment failure and various environmental disturbances, thereby reducing the probability of false alarms and missed alarms from the hardware operation level.

[0018] 3. Mechanism for solving the extrema of the light field gradient The FPGA performs calculations on the global optical field gradient equations to solve for the optimal operating point of phase demodulation, the critical point of mode coupling, and the peak coordinates of vibration positioning, thereby correcting the steady-state offset error calculated by the hardware under long-distance optical cable transmission.

[0019] 4. Hierarchical Coupling Evaluation Mechanism for Seventeen Optical Path Indicators Seventeen indicators: positioning accuracy, phase stability, signal-to-noise ratio, laser linewidth error, temperature drift interference intensity, modal crosstalk degree, time delay deviation, optical loss level, bit error rate, equipment aging degree, computing power consumption, sampling power consumption, optical path distortion, alarm accuracy, long-term drift risk, anti-interference level, and system stability; the hardware weighted computing unit integrates all indicators and outputs a weighted and coupled comprehensive optical path health score.

[0020] This system incorporates a scenario-based indicator pruning and time-sharing acquisition mechanism, eliminating the need for real-time synchronous acquisition and calculation of all seventeen optical path indicators: high-speed dynamic indicators are sampled at millisecond-level high frequency, while slowly changing indicators such as equipment aging and long-term drift are updated at minute-level low-frequency polling. At the same time, it presets three scenario indicator subsets: rail transit, laser detection, and oil and gas pipelines. Operators can disable indicators not required for the scenario as needed, retaining only the core indicators of the corresponding scenario for weighted coupling calculation, which greatly reduces the real-time computing load of the FPGA and ensures stable operation for different hardware configurations.

[0021] 5. Multi-task hierarchical hardware parallel collaborative computing mechanism The hardware is divided into six independent optical path computing sub-units. The GPU is responsible for high-precision computing scheduling, and the CPU is responsible for basic data preprocessing. Through the underlying bus bandwidth allocation and video memory / memory data exchange mechanism, physical-level collaboration between GPU high-precision simulation and CPU basic data preprocessing is achieved. Hardware resources are allocated synchronously for multiple tasks, replacing the traditional serial step-by-step computing architecture.

[0022] 6. Dynamic optimal matching mechanism between computing power and precision The hardware incorporates fiber-optic-specific cost and quality constraint equations, automatically solving for optimal simulation accuracy based on the device's set computing power threshold, thus balancing hardware computing power consumption with optical path measurement accuracy.

[0023] 7. Multi-layer optical path anomaly hardware risk control correction mechanism The hardware has a built-in four-level risk control logic, which sequentially performs distortion signal correction, low signal-to-noise ratio resampling, large-scale interference isolation, and active hardware alarm for major faults, ensuring long-term stable operation of the equipment.

[0024] 8. Measured Error Gradient Self-Iterative Optimization Mechanism The hardware acquires the optical path measurement deviation in each round to construct a global loss function, and the gradient descent updates the weight parameters of seventeen indicators in reverse. The weights are strictly constrained in the range of 0.01 to 2.0, forming a closed loop for hardware device adaptive calibration.

[0025] (iv) Core Innovations of the Invention 1. It pioneered a global dynamic temporal optical field mathematical model adapted to fiber optic photoelectric acquisition hardware. A set of OptField(t) equations uniformly represents multi-physical coupled optical path signals, breaking the fragmentation defects of hardware algorithms in the industry and realizing global collaborative operation of multi-channel acquisition signals.

[0026] 2. By incorporating a causal counterfactual inference operator into the FPGA hardware, multiple superimposed disturbances can be distinguished from the underlying hardware operation, significantly reducing the false alarm and missed alarm rates of field monitoring equipment and breaking through the limitation of traditional hardware that can only output numerical values ​​and cannot analyze the causes of faults.

[0027] 3. Establish a quantitative evaluation system for seventeen optical paths across all dimensions, overcoming the shortcomings of traditional optical fiber hardware that relies on only 2 to 3 physical parameters to determine the optical path, and achieving refined health diagnosis of optical cables.

[0028] 4. A dedicated computing power-precision constraint optimization model for fiber optic hardware is proposed to automatically solve for the optimal simulation accuracy, thus resolving the inherent industry contradiction of excessively high cost of high-precision monitoring hardware.

[0029] 5. It adopts a six-channel sub-unit hierarchical parallel hardware architecture of FPGA, combined with GPU physical-level computing power scheduling, to replace the traditional hardware serial step-by-step calculation, which greatly improves the real-time signal processing efficiency of optoelectronic equipment.

[0030] 6. Construct a complete self-iterative calibration closed loop for fiber optic monitoring hardware, continuously correcting the internal weight parameters of the hardware based on the measured error in each round, and ensuring that the measurement accuracy of the equipment continues to converge positively over long-term operation.

[0031] (v) Beneficial effects 1. Compared with traditional single-point fiber optic monitoring hardware algorithms, the fault location accuracy is improved by 20% to 40%, and the drift error of long-distance optical cable transmission over 100 kilometers is significantly reduced; 2. The ability to identify multi-source environmental interference has been greatly improved, the false alarm and missed alarm rates of equipment monitoring have been reduced by more than 50%, and the stability of long-term operation in the field has been optimized; 3. Under the same optical path measurement accuracy, the computing power and energy consumption of FPGA / GPU hardware are reduced by 30% to 50%, and the hardware costs of equipment procurement and maintenance are significantly reduced; 4. The optical path demodulation has significantly enhanced resistance to polarization fading and modal crosstalk, making it suitable for high-temperature, frozen soil, and complex long-term aging conditions of outdoor optical cables; 5. The hardware can simultaneously output the aging level and fault risk level of the optical cable, and predict potential equipment operation hazards in advance; 6. The equipment has continuous adaptive calibration capabilities, constantly optimizing internal calculation parameters as the optical cable operating conditions change, and has strong adaptability for pipeline expansion. (vi) Specific Implementation Examples Example 1: Distributed Fiber Optic Vibration Hardware Monitoring for High-Speed ​​Railways Hardware configuration: long-distance distributed fiber optic sensor array, FPGA computing motherboard, GPU computing power scheduling module; Input data: phase timing of optical cables along the high-speed railway line, vibration data collected from track segments, ambient temperature sequence along the line, and hardware computing power budget threshold; Execution process: Optoelectronic hardware acquires data from multiple optical paths and initializes the global optical field model; six types of optical path operation subunits on the FPGA perform phase demodulation, delay calculation, noise filtering, and vibration localization in parallel; the hardware's built-in causal operator automatically distinguishes between normal train vibration, track deformation faults, seasonal temperature drift, and external human disturbances; seventeen indicators are weighted to output a comprehensive health score for the track optical cable; multi-layer hardware risk control filters false alarm signals; constraint equations are solved to achieve optimal simulation accuracy; and hardware weight gradient updates are completed based on the acquired measurement errors.

[0033] Compared with a traditional serial fiber optic monitoring server that does not have a full-domain optical field modeling and hierarchical collaborative parallel architecture, the measured data show that the effective fault identification rate is improved by 28%, the false alarm rate is reduced by 33%, the overall computing power and energy consumption of the hardware is reduced by 37%, the long-distance positioning drift defect is significantly improved, and it is suitable for the all-weather high-precision monitoring needs of rail transit.

[0034] Example 2: Fiber Optic Security Hardware Monitoring for Long-Distance Oil and Gas Pipelines Hardware configuration: Buried pipeline distributed fiber optic data acquisition unit, FPGA parallel computing hardware; Input data: fiber optic phase timing of buried pipelines, ground vibration signals, soil temperature acquisition data, and hardware computing power budget threshold; Execution process: Fiber optic hardware synchronously acquires multiple optical path signals and initializes the global dynamic optical field model; FPGA hierarchical sub-units complete the entire optical path calculation in parallel; causal operators distinguish between natural ground vibrations, pipeline corrosion deformation, and seasonal temperature disturbances; hardware coupling module outputs pipeline optical path safety scores; automatically matches the optimal hardware simulation accuracy; and completes the self-iterative calibration of parameters for this round.

[0035] Compared with a traditional serial fiber optic monitoring server that does not have a full-domain optical field modeling and hierarchical collaborative parallel architecture, the pipeline corrosion fault identification rate is increased by 25%, the equipment false alarm rate is reduced by 31%, the hardware operation computing power consumption is reduced by 34%, and it is suitable for year-round unattended security monitoring of buried oil and gas pipelines. (vii) Description of the attached drawings Drafting specifications: pure black and white vector line drawing, no color, no gradient, main flow thick solid lines, feedback dashed lines, key points marked with circles / pentagrams, Chinese characters in Song typeface, numbers and English characters in Times New Roman; all modules come with FPGA / GPU / fiber optic acquisition hardware annotations.

[0037] Figure 1 A flowchart illustrating the overall process of hierarchical collaborative fiber optic hardware signal processing; the modules labeled with dashed lines are the fiber optic photoelectric acquisition hardware input, FPGA computing area, GPU computing power scheduling, and hardware parameter feedback return, which are attached figures in this application abstract.

[0038] Figure 2 A schematic diagram of the temporal gradient evolution simulation of the global dynamic optical field; the horizontal axis is time t, the vertical axis is OptField(t), the curve includes the reference optical field, the high-temperature coupled optical field, and the vibration coupled optical field, and the circles mark the extreme points of the FPGA gradient solution.

[0039] Figure 3 FPGA multi-task hierarchical parallel hardware computing power scheduling architecture diagram; top-level GPU computing power allocation hardware, and six groups of FPGA optical path operation sub-units output in parallel at the lower level.

[0040] Figure 4 Comparison curves of optical path measurement quality and hardware computing power overhead under different simulation accuracies; horizontal axis \rho, left axis measurement score Q, right axis computing power overhead C, pentagram marks the optimal simulation accuracy \rho^*.

[0041] Figure 5 The schematic diagram of hardware coupling scoring and gradient self-iteration closed loop for seventeen optical path indicators; the left side is for indicator acquisition input, the middle is for FPGA coupling calculation, and the dashed line is for error acquisition feedback to update weights.

Claims

1. A hierarchical collaborative adaptive optical fiber signal hardware processing method based on global optical field modeling, applied to distributed optical fiber sensing FPGA / GPU monitoring hardware, characterized in that, Includes the following steps: S1 Multimodal Optical Path Hardware Data Unified Access Steps The optical fiber photoelectric acquisition hardware synchronously acquires multiple raw data streams, including light intensity timing, phase dynamics, polarization parameters, spectral characteristics, signal delay, ambient temperature, and transmission loss. Simultaneously, it inputs the hardware computing power budget threshold to complete the standardized preprocessing of the multi-channel acquired data. S2 Global Light Field Hardware Model Initialization Steps The FPGA loads the optical path causal inference operator, the optical field gradient solution operator, the time sequence evolution operator, and the multimode signal unified encoding operator, and maps all acquired optical path data to the normalized accuracy coefficient \rho∈[0,1], completing the hardware operation model initialization. S3 FPGA Six Types of Optical Path Sub-units Layered Parallel Operation Steps The hardware initiates simultaneous computation of phase demodulation subunit, high-precision latency estimation subunit, optical path noise separation subunit, vibration deformation positioning subunit, laser parameter evaluation subunit, and optical path risk quality control subunit; the GPU dynamically allocates hardware resources based on the marginal benefit of computing power for each task, and multi-path computation is executed independently with results cached and summarized in a unified manner. S4 Multi-Source Interference Hardware Causal Traceability Disassembly Steps The FPGA's built-in causal inference operator performs counterfactual difference calculations on the mixed acquisition signals, automatically distinguishing four types of disturbances: track / pipeline real deformation, environmental temperature drift, optical path mode crosstalk, and equipment noise floor, filtering out worthless interference signals and retaining fault characteristic data. S5 Light Field Gradient Hardware Precision Calibration Steps The FPGA runs the global optical field gradient equations to solve the extreme coordinates of various operations such as phase demodulation and vibration positioning, corrects the hardware measurement offset error in long-distance optical cable transmission and weak light signal scenarios, and outputs optimized optical path positioning and phase data. S6 Seventeen Optical Path Indicators Hardware Coupling Comprehensive Scoring Steps The hardware weighted computing unit integrates seventeen core optical path indicators and substitutes them into the coupling scoring formula to calculate the comprehensive health score of the optical cable and optical path. S7 Multi-layer Optical Path Hardware Anomaly Risk Control Handling Steps The hardware sequentially executes four levels of risk control logic: phase distortion signal correction, low signal-to-noise ratio triggering hardware resampling, large-scale interference signal isolation, and hardware audible and visual alarm for major optical path failures. S8 Fiber Optic Hardware Computing Power - Precision Adaptive Scheduling Steps The FPGA is substituted with the fiber optic dedicated computing power cost and the quality constraint equation is calculated. The optimal simulation accuracy is solved with the hardware computing power budget as the constraint. The next round of hardware calculation is performed using the optimal accuracy. S9 Gradient Self-Iterative Hardware Parameter Optimization Steps The deviation between all measured data and theoretical standard values ​​collected by the hardware is used to construct a global loss function. The gradient descent algorithm is used to update the weights of seventeen optical path indicators in reverse. The weight values ​​are limited to the range of 0.01 to 2.0, thus completing one round of hardware adaptive parameter calibration. S10 Standardized Hardware Output Steps The hardware peripherals uniformly output optical path diagnostic reports, vibration precise positioning coordinates, interference cause classification, laser parameter evaluation, optical path health level, optimal computing power configuration, and current round update weight parameters.

2. A hierarchical collaborative adaptive fiber optic signal processing system based on global optical field modeling, mounted on distributed fiber optic sensing FPGA / GPU photoelectric monitoring hardware, characterized in that, It includes nine major hardware functional units: 1) Multimodal optical path photoelectric acquisition and access unit; 2) Global dynamic light field FPGA mathematical computation unit; 3) FPGA multi-task hierarchical parallel computing unit; 4) Optical path interference hardware cause-and-effect analysis unit; 5) Hardware-coupled quantitative evaluation unit for seventeen optical path indicators; 6) Multi-layer optical path anomaly hardware risk control unit; 7) GPU computing power precision adaptive scheduling unit; 8) Gradient self-iterative hardware parameter optimization unit; 9) Optical path visualization peripheral result output unit.

3. The hierarchical collaborative adaptive fiber signal processing method based on global optical field modeling according to claim 1, characterized in that, The complete optical path evaluation index consists of seventeen items: positioning accuracy, phase stability, signal-to-noise ratio, laser linewidth error, temperature drift interference intensity, modal crosstalk degree, time delay deviation, optical loss level, bit error rate, equipment aging degree, computing power consumption, sampling power consumption, optical path distortion, alarm accuracy, long-term drift risk, anti-interference level, and system stability.

4. The hierarchical collaborative adaptive fiber signal processing method based on global optical field modeling according to claim 1, characterized in that, The optical path causal inference operator uses two sets of synchronously acquired optical field time sequence signal difference counterfactual comparison calculation to distinguish between various environmental disturbances and real equipment failures.

5. The hierarchical collaborative adaptive fiber signal processing method based on global optical field modeling according to claim 1, characterized in that, The quadratic function of computing power cost is C_{opt}(\rho)=a\rho^2+b\rho+c, and the linear quality function of optical path measurement is Q_{opt}(\rho)=k\rho+m; where a, b, c, k, and m are all empirical constants obtained by fitting long-term field measurements and hardware benchmark tests.

Images