A multi-channel synchronous optical wavelength scanning test module and a data fusion system thereof

Abstract

The application belongs to the technical field of optical communication testing, and provides a multi-channel synchronous optical wavelength scanning test module and a data fusion system thereof. The multi-channel synchronous optical wavelength scanning test module comprises a scanning light source, a multi-channel scanning optical power meter, a wavelength analysis unit, a polarization controller, a trigger synchronization unit and an integrated platform. The scanning light source is configured to provide a basic light source. The multi-channel scanning optical power meter is configured to simultaneously measure the optical power of multiple channels to obtain optical power data. The wavelength analysis unit is configured to analyze the wavelength of the optical signal output by the basic light source to determine the wavelength value of the optical signal. The trigger synchronization unit is configured to realize linkage control between the scanning light source and the polarization controller to form polarization state-wavelength collaborative scanning, thereby synchronously completing wavelength scanning and switching of multiple polarization states. The integrated platform is configured to be uniformly controlled and managed. The application can realize high-precision and high-efficiency multi-channel optical wavelength scanning testing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical communication testing, and in particular to a multi-channel synchronous optical wavelength scanning test module and a data fusion system thereof. BACKGROUND

[0002] In the field of optical device (such as DWDM filter, WSS, adjustable optical attenuator) and optical system testing, wavelength scanning test combined with polarization state analysis is the core means to evaluate its key performance indicators (such as wavelength dependent loss WDL, polarization dependent loss PDL, polarization mode dispersion PMD). However, the existing technical solutions have significant limitations in realizing efficient, high-precision, multi-channel polarization state-wavelength scanning test:

[0003] First, the test efficiency is low, which is difficult to meet the batch testing demand: the mainstream scheme relies on a sequential test mode, which first completes full wavelength scanning under one polarization state, and then manually or semi-automatically switches to the next polarization state for repeated scanning. This test process leads to a linear increase in test time with the number of polarization states, resulting in low efficiency, especially for test projects that need to cover multiple polarization states; although there are multi-channel optical power meters, there is a lack of deep cooperation with polarization controllers, and it is impossible to realize real multi-channel parallel polarization state-wavelength scanning.

[0004] Second, the multi-channel measurement accuracy is limited by crosstalk and noise: when high-density multi-channel (such as ≥48 channels) is synchronized and high-speed sampling (such as ≥100k points / second) is performed, the problem of coherent crosstalk between channels is particularly prominent, which seriously distorts the optical power reading, especially affecting the accuracy of large dynamic range (80 decibels) measurement. The existing suppression methods (such as increasing physical spacing, optical filtering) have limited effect, increasing cost or reducing flexibility, and lack effective solutions at the signal processing level; when pursuing higher measurement accuracy (low noise floor), quantum noise gradually becomes a theoretical limit, and the existing test system lacks effective means to distinguish and specifically suppress quantum noise, usually using a fixed detection mode, which cannot be optimized according to the actual noise composition.

[0005] Third, the system integration is low, and the synchronization accuracy is insufficient: core functional units, such as scanning light sources, polarization controllers, optical power meters, and wavelength meters, are often deployed independently, with complex interfaces, which not only occupies space and increases cost, but more importantly, leads to difficulties in timing control between units, poor synchronization accuracy, and difficulty in achieving strict synchronization of wavelength scanning and polarization state switching, affecting the accuracy of polarization state-wavelength correlation data; at the same time, the existing scanning generally uses fixed wavelength step and fixed polarization switching sequence, which needs to traverse all preset points, including a large number of non-critical or redundant test points, and cannot optimize the scanning path according to the device characteristics or historical data.

[0006] Fourth, environmental sensitivity leads to data drift, and lack of effective compensation: high-precision wavelength scanning is extremely sensitive to environmental disturbances such as temperature, humidity and vibration, which can easily cause wavelength drift and unstable optical power readings; the existing system usually uses simple temperature compensation (may only be for the light source), lacks systematic modeling and real-time, dynamic compensation capability for the comprehensive influence of multiple environmental parameters (temperature and humidity, vibration), and it is difficult to ensure the stability and repeatability of long-term testing.

[0007] Fifth, data dispersion, lack of intelligent fusion and correction: the massive test data generated by multiple channels, multiple polarization states and wide wavelength range are scattered and isolated, the existing data processing process lacks accurate modeling of polarization state-wavelength, and lacks targeted correction means for environmental drift error and noise error (especially quantum noise), which limits the accuracy and reliability of the final analysis results.

[0008] Therefore, it is necessary to provide a multi-channel synchronous optical wavelength scanning test module and its data fusion system to significantly improve the efficiency, accuracy and reliability of optical communication device testing, and meet the growing demand for mass production and high-standard testing. SUMMARY

[0009] The present application provides a multi-channel synchronous optical wavelength scanning test module and its data fusion system, which can realize efficient, high-precision and multi-channel polarization state-wavelength scanning test of optical communication devices, and effectively overcome the many limitations of the prior art.

[0010] The present application provides a multi-channel synchronous optical wavelength scanning test module, comprising: a scanning light source, a multi-channel scanning optical power meter, a wavelength analysis unit, a polarization controller, a trigger synchronization unit and an integrated platform;

[0011] The scanning light source is configured to provide a basic light source;

[0012] The multi-channel scanning optical power meter is configured to simultaneously measure the optical power of multiple channels to obtain optical power data;

[0013] The wavelength analysis unit is configured to analyze the wavelength of the optical signal output by the basic light source to determine the wavelength value of the optical signal;

[0014] The trigger synchronization unit is configured to realize linkage control between the scanning light source and the polarization controller to form polarization state-wavelength coordinated scanning, thereby synchronously completing wavelength scanning and switching of multiple polarization states;

[0015] The integrated platform is configured to uniformly control and manage the scanning light source, the multi-channel scanning optical power meter, the wavelength analysis unit, the polarization controller and the trigger synchronization unit.

[0016] Further, the wavelength range of the scanning light source is configured to be continuously adjustable in the range of 1490 nm to 1630 nm; the set value of the scanning rate is greater than or equal to 100 nm / s; the set value of the scanning step is less than or equal to 1 pm; the scanning light source is configured to selectively carry an internal semiconductor laser amplifier light source or an external third-party light source; the internal semiconductor laser amplifier light source is configured to increase the power of the light signal output by the basic light source to 10 dBm and has a standardized optical interface.

[0017] Further, the multi-channel scanning optical power meter is configured to support ≥48 channels of synchronous sampling; when the scanning rate of the working mode is greater than or equal to 100 k points / s, the dynamic range is greater than 80 decibels; when the scanning rate of the working mode is 100 k points / s, the absolute value of the measurement noise is not greater than 0.005 decibels under the condition of an average time of 10 milliseconds; when the scanning rate of the working mode is greater than 10 k points / s and less than or equal to 100 k points / s, the dynamic range is greater than 60 decibels; when the scanning rate of the working mode is greater than 1 k points / s and less than or equal to 20 k points / s, the dynamic range is greater than 70 decibels;

[0018] The multi-channel scanning optical power meter is integrated with a coherent crosstalk suppression unit; the coherent crosstalk suppression unit comprises:

[0019] A local oscillation light source configured to output a local oscillation light signal with adjustable frequency;

[0020] A coherent mixer configured to mix the input light signal with the local oscillation light signal to generate a mixed electric signal carrying amplitude and phase information;

[0021] A crosstalk compensation processor configured to calculate a crosstalk compensation factor based on the mixed electric signal of the adjacent channel, and dynamically correct the optical power data of the current channel using the crosstalk compensation factor, so that the crosstalk error is ≤±0.01 decibels.

[0022] Further, the wavelength resolution accuracy of the wavelength resolution unit is configured to be <±10 pm; the wavelength range of the wavelength resolution unit is configured to be 1490-1630 nm; the maximum insertion loss of the wavelength resolution unit is configured to be ≤0.8 decibels; and the sampling rate of the wavelength resolution unit is ≤100 k points / s / channel.

[0023] Further, the polarization controller is configured to switch between -45° linear polarization state, 0° linear polarization state, +45° linear polarization state, +90° linear polarization state, left-handed circular polarization state and right-handed circular polarization state during wavelength scanning, and the adjustment time of the switching is not greater than 250 microseconds.

[0024] Further, the trigger synchronization unit is configured to support both electrical triggering and optical triggering modes, and is constructed to synchronize the scanning light source with the polarization controller through an FPGA clock, with a triggering error less than 1 microsecond; the trigger synchronization unit includes an AI prediction control subunit embedded with a lightweight neural network model,

[0025] The AI prediction control subunit is built-in with a scanning path prediction submodel, which is generated based on a historical polarization state-wavelength test dataset; the AI prediction control subunit is configured to skip non-critical wavelength points based on the scanning path prediction submodel, generate an adaptive scanning sequence, and control the scanning light source and the polarization controller to perform variable-step collaborative scanning based on the adaptive scanning sequence;

[0026] The lightweight neural network model is a hybrid model of a convolutional neural network and a long short-term memory network, training data is collected from no less than 1000 groups of historical test data, the historical test data includes wavelength scanning trajectories, polarization state switching timing, historical noise distribution, and test result confidence labels; the convolutional neural network is a three-layer one-dimensional structure, and the sizes of the convolution kernels are 11, 7, and 5 respectively, for extracting local time-frequency features; the long short-term memory network is a bidirectional 128-dimensional structure, for capturing time-dependent features based on local time-frequency features, and dynamically adjusting scanning step and key scanning areas according to the time-dependent features to generate an adaptive scanning sequence.

[0027] A data fusion system for a multi-channel synchronous optical wavelength scanning test module, comprising: a data acquisition module, a preprocessing module, a correlation analysis module, an error correction module, and a result output module;

[0028] The data acquisition module includes a universal data acquisition adapter configured to support serial ports, Ethernet ports, and optical fiber interfaces, and compatible with Modbus protocol, TCP / IP protocol, and SNMP protocol, for synchronously working with a multi-channel scanning optical power meter, a wavelength resolution unit, and a polarization controller of the multi-channel synchronous optical wavelength scanning test module, to collect polarization state-wavelength correlation data at a frequency matching the scanning rate of the scanning light source; the polarization state-wavelength correlation data includes optical power data, corresponding wavelength values, polarization state identifiers, and sampling time stamps of each channel;

[0029] The preprocessing module is used for preprocessing the polarization state-wavelength correlation data; the preprocessing includes but is not limited to: removing noise interference by using wavelet transform, filling in missing data values by using K-nearest neighbor interpolation method, and mapping the polarization state-wavelength correlation data to the interval [-1, 1] by using maximum and minimum value normalization;

[0030] The preprocessing module further comprises an intelligent data mapping unit, which automatically identifies semantics of the multi-source heterogeneous data by using a light-weight semantic analysis model based on MobileBERT, and unifies data units according to a preset unit conversion rule library;

[0031] The correlation analysis module is configured to perform correlation analysis on the preprocessed normalized matrix according to the constructed correspondence model of polarization state and wavelength; the correspondence model is constructed based on a Gaussian process regression model;

[0032] The error correction module is configured to correct errors in the polarization state-wavelength correlation data according to the result obtained by the correlation analysis module;

[0033] The result output module is configured to present the data after preprocessing, correlation analysis and error correction in the form of a chart and a data file; the chart comprises a polarization state-wavelength loss curve and a comparison chart of powers of each channel; and the data file comprises original data in CSV format and analysis results in JSON format.

[0034] Further, the error correction module comprises a quantum noise analysis subunit; the quantum noise analysis subunit is configured to generate a phase noise spectrum by performing fast Fourier transform on the output of the coherent mixer, and identify the noise source as thermal noise or quantum noise; when the proportion of quantum noise is greater than 30%, a measurement mode of balanced homodyne detection is started to reduce the background noise.

[0035] Further, the environmental monitoring and compensation module comprises:

[0036] The temperature and humidity sensor and the vibration sensor are configured to collect environmental parameters in real time;

[0037] The storage unit is configured to construct an error mapping model library to store the mapping relationship between the environmental parameters and the wavelength deviation / optical power drift;

[0038] The dynamic correction engine unit is configured to call the mapping relationship to compensate the optical power data and the wavelength value in the polarization state-wavelength correlation data after correction in real time.

[0039] Further, the scanning parameter adjustment control module comprises a cooperative interface configuration unit, a data anomaly prediction unit, a scanning parameter adjustment instruction generation unit and a feedback data receiving and processing unit.

[0040] The cooperative interface configuration unit is configured to configure a real-time cooperative interface between the data fusion system and the integrated platform of the multi-channel synchronous optical wavelength scanning test module; the real-time cooperative interface is configured to use the UDP protocol for data interaction.

[0041] Data anomaly prediction unit, for inputting the feature vector of the last 100 sets of polarization state-wavelength correlation data based on the random forest model; the feature vector includes but is not limited to loss value, fluctuation slope, correlation degree score; output the anomaly probability of future multiple wavelength points; the resolution of the anomaly probability is higher than 1%;

[0042] Scan parameter adjustment instruction generation unit, for generating scan parameter adjustment instruction when the anomaly probability is greater than 80%, and sending the scan parameter adjustment instruction to the integrated platform through the cooperative interface; the scan parameter adjustment instruction includes: reducing the scanning rate in the abnormal interval by 50%, reducing the scanning step to 1 / 10 of the original step, and synchronously increasing the sampling rate of the multi-channel scanning optical power meter by 50%;

[0043] Feedback data receiving and processing unit, for receiving the collection data of the scanning optical power meter after parameter adjustment through the real-time cooperative interface, if the collection data amount increases by 30%, and at the same time meets the following conditions:

[0044] (a) the data quality comprehensive score based on real-time calculation is higher than the preset threshold 0.7;

[0045] (b) input the current collection data and historical adjustment record by applying the reinforcement learning DQN algorithm model, and output the confidence of the parameter optimization strategy is greater than or equal to 90%;

[0046] (c) verify that the proportion of abnormal points in the data is less than 5% by the isolation forest anomaly detection algorithm;

[0047] Then execute the following operation: generate scan parameter recovery instruction, send to the integrated platform through the real-time cooperative interface, gradually restore the original scan parameter, and trigger the model online iteration training of the data anomaly prediction unit, and update the weight of the random forest model using the new collection data;

[0048] Otherwise, if any condition is not met, maintain the current adjustment parameter, generate an incremental data collection instruction, and input the feedback data to the data anomaly prediction unit to optimize the feature vector extraction algorithm in real time to improve the accuracy of future anomaly probability prediction; the incremental data collection instruction includes: increasing the scanning rate to the original level, and restoring the scanning step to the original size; the incremental data collection instruction is: increasing the sampling rate by 20% based on the adjusted sampling rate, while maintaining the current scanning rate and scanning step.

[0049] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0050] 1. The system can realize high-precision measurement; the scanning light source has a continuously adjustable wavelength range of 1490-1630 nm, fine setting of scanning rate and step, and built-in semiconductor laser amplifier light source to improve optical power, laying a foundation for subsequent accurate measurement; the multi-channel scanning optical power meter supports synchronous sampling of multiple channels, has excellent dynamic range and extremely low measurement noise at different scanning rates, and the integrated coherent crosstalk suppression unit can effectively reduce crosstalk error and further improve measurement accuracy; the high-precision wavelength resolution capability of the wavelength resolution unit and the characteristics of the polarization controller for quickly switching multiple polarization states can improve the accuracy when measuring various parameters of the optical signal.

[0051] 2. The trigger synchronization unit realizes linkage control between the scanning light source and the polarization controller, supports electrical trigger and optical trigger modes and FPGA clock synchronization, controls the trigger error within a very small range, ensures the synchronization of wavelength scanning and multiple polarization state switching, efficiently completes complex measurement tasks, and avoids measurement errors and data confusion caused by asynchronization; compared with the prior art, the measurement efficiency and data reliability are greatly improved.

[0052] 3. The data fusion system modules work closely together, from data acquisition to preprocessing, to correlation analysis, error correction and result output, forming an intelligent data processing flow; the preprocessing module uses multiple advanced algorithms to remove noise, fill in missing values and normalize data, and the intelligent data mapping unit can automatically identify the semantics of multi-source heterogeneous data and unify the units; the correlation analysis module constructs a corresponding relationship model based on the Gaussian process regression model, effectively analyzing the correlation between polarization state and wavelength; the error correction module can not only correct general errors, but also take corresponding measures for different noise sources through the quantum noise analysis subunit; the environment monitoring and compensation module collects environmental parameters in real time and compensates the data, and the scanning parameter adjustment control module dynamically adjusts the scanning parameters according to the data anomaly prediction, the entire data processing process embodies high intelligence and automation, and compared with traditional technology, it can process and analyze a large amount of measurement data more quickly and accurately.

[0053] 4. The universal data acquisition adapter of the data acquisition module supports multiple interfaces and protocols, can work synchronously with multiple components of the multi-channel synchronous optical wavelength scanning test module, ensures that the system can be flexibly adapted to different devices and working scenarios; the flexibility of components such as the multi-channel scanning optical power meter and the scanning light source in parameter configuration, such as the selection of built-in or external light source for the scanning light source and the dynamic range adjustment of the multi-channel scanning optical power meter in different working modes, enables the system to adapt to various complex optical signal measurement requirements; and the environmental monitoring and compensation module and the scanning parameter adjustment control module further enhance the adaptability of the system to environmental changes and data anomalies, compared with the prior art, have a wider application scenario and stronger adaptability.

[0054] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims.

[0055] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0056] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0057] Figure 1 It is a structural schematic diagram of a multi-channel synchronous optical wavelength scanning test module;

[0058] Figure 2 It is a structural schematic diagram of a coherent crosstalk suppression unit;

[0059] Figure 3 It is a structural schematic diagram of a data fusion system for a multi-channel synchronous optical wavelength scanning test module. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present application will be described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not constitute a limitation on the present application.

[0061] The present application provides a multi-channel synchronous optical wavelength scanning test module, as shown in Figure 1 , comprising:

[0062] a scanning light source, a multi-channel scanning optical power meter, a wavelength resolution unit, a polarization controller, a trigger synchronization unit and an integrated platform;

[0063] The scanning light source is configured to provide a basic light source;

[0064] The multi-channel scanning optical power meter is configured to measure optical power of multiple channels simultaneously to obtain optical power data;

[0065] The wavelength resolving unit is configured to resolve the wavelength of the optical signal output by the scanning light source to determine the wavelength value of the optical signal;

[0066] The trigger synchronization unit is configured to realize linkage control between the scanning light source and the polarization controller to form polarization state-wavelength collaborative scanning, so as to complete wavelength scanning and switching of multiple polarization states synchronously;

[0067] The integrated platform is configured to uniformly control and manage the scanning light source, the multi-channel scanning optical power meter, the wavelength resolving unit, the polarization controller and the trigger synchronization unit; specifically including: issuing wavelength scanning parameters to the scanning light source; issuing polarization state switching sequence to the polarization controller; issuing sampling trigger instruction and channel enable configuration to the multi-channel scanning optical power meter; generating a synchronous clock reference based on the FPGA clock signal output by the trigger synchronization unit; coordinating the wavelength stepping of the scanning light source and the state switching timing of the polarization controller to ensure that the trigger error is ≤1 μs; receiving and storing wavelength data from the wavelength resolving unit, optical power data from the multi-channel scanning optical power meter and state feedback data from the polarization controller in real time; monitoring the running state of each module and triggering shutdown or parameter reset when abnormal; providing a human-computer interaction interface.

[0068] The working principle of the above technical solution is: in order to realize a multi-channel synchronous optical wavelength scanning test module, the present application proposes to uniformly schedule and cooperatively manage each component module through an integrated platform. First, the scanning light source generates and outputs a basic optical signal, for example, the scanning light source can be set to continuously scan the wavelength in the L-band range of 1530 nanometers to 1580 nanometers with a step of 0.1 nanometer; the optical signal is delivered to the wavelength resolving unit, which performs real-time wavelength resolving processing on the input optical signal, for example, when the scanning light source outputs a 1550.0 nanometer optical signal, the wavelength resolving unit can accurately determine that the wavelength value of the current optical signal is 1550.0 nanometers, and immediately feedback this data to the integrated platform;

[0069] On the other hand, the optical signal enters the polarization controller, and under the action of the trigger synchronization unit, the scanning light source and the polarization controller form a linkage control mechanism; for example, the trigger synchronization unit can set, according to the preset control logic, that when the wavelength is 1550.0 nanometers, the polarization controller is driven to switch to a 0° linear polarization state; when the wavelength is scanned to 1550.1 nanometers, it is synchronously switched to a 45° linear polarization state; when the wavelength is 1550.2 nanometers, it is switched to a 90° linear polarization state; when the wavelength is 1550.3 nanometers, it is switched to a circular polarization state, and so on, so as to synchronously switch multiple polarization states during wavelength scanning, realize polarization state-wavelength collaborative scanning, and make the optical signal correspond to different polarization state combinations under different wavelengths;

[0070] After the polarization state modulation, the optical signal, for example, the optical signal carrying 1550.0 nanometer / 0° polarization, 1550.1 nanometer / 45° polarization, etc. information, is distributed to 8 test channels through a 1*8 optical splitter, and each test channel is connected with an optical coupler to be tested; a multi-channel scanning optical power meter (such as an 8-channel optical power meter) simultaneously measures the optical power of the 8 channels in real time, assuming that at a wavelength of 1550.0 nanometers and a 0° polarization state, the optical power of channel 1 is -10 dBm, the optical power of channel 2 is -10.2 dBm, and the optical power of channel 8 is -9.9 dBm; when the wavelength is scanned to 1550.1 nanometers and the polarization state is 45°, the optical power of each channel will have new readings, and these optical power data at different wavelengths and polarization states are acquired in real time and transmitted to the integrated platform;

[0071] The integrated platform centrally processes, stores and analyzes the received wavelength values (such as 1550.0 nanometers, 1550.1 nanometers, etc.), corresponding polarization state information (such as 0°, 45°, etc.), channel optical power data (such as -10 dBm, -10.2 dBm, etc.), and the working state information of each unit (such as whether the scanning light source is normal, whether the polarization controller is switched to the right position, etc.); for example, the integrated platform can draw a curve of the optical power of each test channel with respect to the wavelength, and mark the power difference at different polarization states, thereby completing the synchronous scanning test of the 8 optical couplers under different wavelengths and polarization states, quickly evaluating the key performance indicators such as wavelength dependence and polarization dependent loss (PDL) of these optical couplers, and realizing comprehensive and efficient evaluation of the performance of optical devices.

[0072] The above technical scheme has the beneficial effects that: by using the scheme provided in the embodiment, the efficiency and accuracy of optical wavelength scanning test can be effectively improved, and through the unified scheduling and collaborative management of the integrated platform, multi-channel synchronous test is realized, the test time is greatly shortened, and the test requirements of optical devices of different scales can be met.

[0073] In one embodiment, the wavelength range of the scanning light source is configured to be continuously adjustable in the range of 1490 nm to 1630 nm; the set value of the scanning rate is greater than or equal to 100 nm / s; the set value of the scanning step is less than or equal to 1 pm; the scanning light source is configured to selectively carry an embedded semiconductor laser amplifier light source or an external third-party light source; the embedded semiconductor laser amplifier light source is configured to increase the power of the light signal output by the basic light source to 10 dBm and has a standardized optical interface.

[0074] The working principle of the above technical solution is that by setting the wavelength range of the scanning light source to be continuously adjustable in the range of 1490 nm to 1630 nm, cooperating with a scanning rate greater than or equal to 100 nm / s and a scanning step less than or equal to 1 pm, the target in this specific wavelength interval can be scanned and analyzed at high speed and high precision; on the light source configuration, the embedded semiconductor laser amplifier light source can effectively increase the output light power of the basic light source to 10 dBm, and output through the standardized FC / APC optical interface to ensure stable transmission and good adaptation of the optical signal; at the same time, it also supports the adaptation of external third-party light sources to meet the diversified needs of light source characteristics in different application scenarios.

[0075] The beneficial effects of the above technical solution are that by using the scheme provided in this embodiment, the system can flexibly, efficiently and accurately complete the optical signal transmission and scanning detection task in the 1490 nm-1630 nm wavelength band.

[0076] In one embodiment, the multi-channel scanning optical power meter is configured to support ≥48 channels of synchronous sampling; when the scanning rate of the working mode is greater than or equal to 100 k points / s, the dynamic range is greater than 80 decibels; when the scanning rate of the working mode is 100 k points / s, the absolute value of the measurement noise is not greater than 0.005 decibels under the condition of 10 milliseconds of average time; when the scanning rate of the working mode is greater than 10 k points / s and less than or equal to 100 k points / s, the dynamic range is greater than 60 decibels; when the scanning rate of the working mode is greater than 1 k points / s and less than or equal to 20 k points / s, the dynamic range is greater than 70 decibels.

[0077] The multi-channel scanning optical power meter integrates a coherent crosstalk suppression unit; as shown in Figure 2 The coherent crosstalk suppression unit includes:

[0078] A local oscillation light source configured to output a local oscillation light signal with adjustable frequency;

[0079] A coherent mixer configured to mix the input light signal and the local oscillation light signal to generate a mixed electric signal carrying amplitude and phase information;

[0080] The crosstalk compensation processor is configured to calculate a crosstalk compensation factor based on mixed frequency signals of adjacent channels, and dynamically corrects the optical power data of the current channel by using the crosstalk compensation factor, so that the crosstalk error is ≤±0.01 decibel.

[0081] The working principle of the technical solution is as follows: the multi-channel scanning optical power meter realizes the synchronous sampling of ≥48 channels through a parallel data acquisition architecture, ensuring the real-time performance and consistency of the multi-channel signals; in different scanning rate modes, the system performance is dynamically adjusted by optimizing the gain control of the photoelectric conversion module and the signal processing bandwidth: when the scanning rate ≥100k points / second, a high-bandwidth low-noise front-end design is adopted, combined with a 10ms average time digital filtering algorithm, to realize a dynamic range of >80 decibels and a measurement noise of ≤0.005 decibels; when the rate is in the interval of 1k-20k points / second, the signal-to-noise ratio is enhanced by increasing the integration time to ensure a dynamic range of >70 decibels; and in the interval of 10k-100k points / second, the dynamic range of >60 decibels is maintained by balancing the bandwidth and gain.

[0082] The coherent crosstalk suppression unit aims to solve the problem of coherent crosstalk and realize high-precision multi-channel optical power synchronous measurement, and its work is divided into three steps: first, the local oscillation light source outputs a local oscillation light signal with adjustable frequency, which is mixed with the input optical signal in a coherent mixer to convert the optical domain information into mixed frequency signals carrying amplitude and phase for easy processing; for example, in a certain optical communication network, the input optical signal frequency is 193.1THz, the local oscillation light signal frequency is 193.2THz, and the mixed frequency signal is 100GHz; second, the crosstalk compensation processor collects the mixed frequency signals of adjacent channels in real time, and calculates the crosstalk compensation factor according to the phase difference and amplitude coupling model between channels; for example, in this network, it is found that the adjacent channel B has crosstalk to the current channel A, causing a 10° phase deviation and a 0.05 decibel amplitude increase in channel A, and the processor calculates the corresponding compensation factor; finally, the crosstalk compensation factor is used to dynamically phase cancel and amplitude correct the optical power data of the current channel, so that the crosstalk error is controlled within ≤±0.01 decibels; for example, after the correction of channel A, the crosstalk error is reduced from ±0.03 decibels to ±0.008 decibels, ensuring the accuracy and stability of the optical signal transmission of each channel in the optical communication network.

[0083] The beneficial effects of the technical solution are as follows: the scheme provided in the embodiment realizes the synchronous sampling of ≥48 channels, ensuring the real-time performance and consistency of the multi-channel signals; by dynamically adjusting the system performance, an optimal dynamic range and low measurement noise can be ensured in different scanning rate modes; the coherent crosstalk suppression unit effectively solves the problem of coherent crosstalk, controls the crosstalk error within ≤±0.01 decibels, and ensures the accuracy and stability of the optical signal transmission of each channel in the optical communication network.

[0084] In one embodiment, the wavelength resolution unit is configured to have a wavelength resolution accuracy of <±10 pm, a wavelength range of 1490-1630 nm, a maximum insertion loss of ≤0.8 dB, a sampling rate of ≤100 k points / s / channel, and a total sampling rate of ≥5 M points / s.

[0085] The working principle of the technical solution is that the wavelength resolution unit samples the input optical signal at a high speed within a wavelength range of 1490-1630 nm, separates and converts optical signals of different wavelengths into electrical signals through an internal high-precision splitting and detection mechanism, processes and analyzes the electrical signals using an advanced algorithm, achieves a wavelength resolution accuracy of <±10 pm, and accurately identifies the wavelength information of the optical signal. Meanwhile, by optimizing the optical path design and selecting low-loss optical elements, the maximum insertion loss is ensured to be ≤0.8 dB, the energy loss of the optical signal in the transmission and resolution process is reduced, and the effective detection and accurate resolution of the signal are ensured.

[0086] The beneficial effects of the technical solution are that the scheme provided in the embodiment can sample the input optical signal at a high speed within a wavelength range of 1490-1630 nm, achieve a wavelength resolution accuracy of <±10 pm through high-precision splitting and detection and advanced algorithm processing, accurately identify the wavelength information of the optical signal, and, by optimizing the optical path design and selecting low-loss elements, ensure a maximum insertion loss of ≤0.8 dB, effectively reduce the energy loss of the signal, and ensure the effective detection and accurate resolution of the signal.

[0087] In one embodiment, the polarization controller is configured to switch between -45° linear polarization state, 0° linear polarization state, +45° linear polarization state, +90° linear polarization state, left-handed circular polarization state, and right-handed circular polarization state during wavelength scanning, and the adjustment time of the switching is not greater than 250 microseconds.

[0088] The working principle of the technical solution is as follows: the polarization controller dynamically adjusts the polarization state of the incident light through the internal integrated electro-optic modulator (such as LiNbO3 phase modulator) or magneto-optic modulator array according to the preset control timing and driving voltage / current signal, and the principle is based on the polarization state superposition and phase delay effect of light: when linearly polarized light passes through a birefringent crystal, the ordinary light (o light) and the extraordinary light (e light) produce a phase difference, and by accurately controlling the phase difference (0°, 90°, 180°, 270°, etc.) and the rotation angle of the polarization direction, different target polarization states can be synthesized; for example, when the voltage is applied to make the phase difference of the two beams of light 90° and the amplitudes equal, a circularly polarized state can be output; by adjusting the angle between the fast axis of the wave plate and the polarization direction of the incident light (such as 0°, 45°, 90°), the linearly polarized state at different angles can be switched, and the adjustment time of the switching is not more than 250 microseconds, which depends on the electro-optic response speed of the modulator (usually nanoseconds) and the real-time signal processing capability of the control system, to ensure that the polarization state switching and wavelength change are synchronized during the rapid wavelength scanning (such as 1000 nanometers / second), and to avoid polarization crosstalk in spectral measurement; for example, in a fiber grating sensing system, when it is necessary to detect the reflection spectrum of the grating under different polarization states, the polarization controller completes the switching from 0° linear polarization to right circular polarization within 250 microseconds according to the set timing, at this time, the driving circuit applies a step voltage to the electro-optic modulator, so that the crystal birefringence effect is stable within 200 microseconds, and the output polarization state is verified by a polarization beam splitter, and the extinction ratio is switched from 30 decibels when the polarization is 0° to 28 decibels when the polarization is right circular, which ensures that the sequential measurement of six polarization states is completed within the wavelength scanning period (such as 1ms), and the switching interval between adjacent polarization states meets the requirement of the spectral resolution (0.1 nanometer) of the system.

[0089] The beneficial effects of the above technical solution are that the scheme provided in the embodiment can dynamically adjust the polarization state of the incident light, the switching adjustment time is ≤250 microseconds, which can be synchronized with the rapid wavelength scanning, and the polarization crosstalk in spectral measurement is avoided.

[0090] In one embodiment, the trigger synchronization unit is configured to support two trigger modes of electrical trigger and optical trigger, and is constructed to make the trigger error of the scanning light source and the polarization controller less than 1 microsecond through FPGA clock synchronization; the trigger synchronization unit includes an AI prediction control subunit embedded with a lightweight neural network model,

[0091] The AI prediction control subunit is built-in with a scanning path prediction submodel, and the scanning path prediction submodel is generated based on historical polarization state-wavelength test data sets; the AI prediction control subunit is configured to skip non-critical wavelength points based on the scanning path prediction submodel, generate an adaptive scanning sequence, and control the scanning light source and the polarization controller to perform variable-step collaborative scanning based on the adaptive scanning sequence;

[0092] The lightweight neural network model is a hybrid model of a convolutional neural network and a long short-term memory network. Training data is collected from no less than 1000 groups of historical test data, and the historical test data includes a wavelength scanning trajectory, a polarization state switching time sequence, a historical noise distribution, and a test result confidence label. The convolutional neural network is a three-layer one-dimensional structure, and the sizes of the convolution kernels are 11, 7 and 5 respectively, which are used to extract local time-frequency features. The long short-term memory network is a bidirectional 128-dimensional structure, which is used to capture time-dependent features based on the local time-frequency features, and dynamically adjust the scanning step and the key scanning area according to the time-dependent features to generate an adaptive scanning sequence.

[0093] The specific training method of the lightweight neural network model is as follows:

[0094] Firstly, the training data is preprocessed, and the wavelength scanning trajectory, the polarization state switching time sequence, the historical noise distribution and other data are normalized to make all data in the same order of magnitude range, so that the model can learn better. At the same time, the test result confidence label is encoded and converted into a format suitable for model training. Then, the preprocessed training data is divided into a training set and a validation set with a ratio of 8:2. The training set is used for parameter updating of the model, and the validation set is used to evaluate the performance of the model in the training process to prevent overfitting. In the training process, the data is input into the convolutional neural network in sequence. The first layer of the convolution kernel has a size of 11, and the input data is convolved through a sliding window to extract local time-frequency features in a larger range. Then, the second layer of the convolution kernel has a size of 7, which further refines feature extraction and focuses on more detailed local features. After that, the third layer of the convolution kernel has a size of 5, which makes the features more refined.

[0095] The feature map output by the convolutional neural network is then input into the bidirectional 128-dimensional long short-term memory network. The forward long short-term memory network processes from the beginning of the sequence, and the backward long short-term memory network processes from the end of the sequence. Both of them capture the time-dependent relationship in the data. The long short-term memory network analyzes the change pattern of the data over time based on the local time-frequency features extracted by the convolutional neural network, and then dynamically adjusts the scanning step and the key scanning area.

[0096] In the training process, the stochastic gradient descent algorithm is used as the optimizer, the learning rate is set to 0.001, the momentum factor is set to 0.9, and the cross-entropy loss function is selected as the loss function to measure the difference between the model prediction results and the true labels. Through continuous iterative training, the parameters of the model are adjusted to gradually reduce the loss function value. After training for a certain number of rounds (such as 10 rounds), the model performance is evaluated on the validation set, and the accuracy, recall rate and other indicators on the validation set are calculated.

[0097] The working principle of the technical solution is as follows: the trigger synchronization unit supports both electrical triggering and optical triggering modes, and uses FPGA clock synchronization technology to ensure that the trigger error between the scanning light source and the polarization controller is less than 1 microsecond, thereby laying a foundation for high-precision cooperative scanning; the core lies in the embedded AI prediction control subunit. The subunit uses a lightweight neural network model that combines convolutional neural networks (CNN) and long short-term memory networks (LSTM). First, the subunit receives at least 1000 sets of historical polarization state-wavelength test data sets including wavelength scanning trajectories, polarization state switching time sequences, historical noise distributions, and test result confidence labels, and uses them to train a scanning path prediction model. For example, if the historical data shows that the polarization state changes smoothly in the wavelength range of 400-500 nanometers and the test result confidence is generally high, while the polarization state fluctuates greatly in the wavelength range of 500-600 nanometers and the confidence is unstable, the model will learn these characteristics. During the training process, a three-layer one-dimensional CNN (with convolution kernel sizes of 11, 7, and 5, respectively) is responsible for extracting local time-frequency features such as polarization state change frequency and noise peak value in a specific wavelength range from the historical data. A bidirectional 128-dimensional LSTM captures the temporal dependence between different wavelength scanning orders based on these local features, such as how the polarization state switching at a certain wavelength point affects the scanning accuracy of subsequent points. Subsequently, based on the trained scanning path prediction model, the AI prediction control subunit can identify and skip non-critical wavelength points (such as the points in the 400-500 nanometer range mentioned above, which have minimal fluctuations and negligible impact on the overall results), and generate an adaptive scanning sequence. For example, small-step dense scanning is used in the key area of 500-600 nanometers, while large-step jump scanning is used in the non-critical area of 400-500 nanometers. Finally, the scanning light source and the polarization controller are controlled to perform variable-step cooperative scanning based on the adaptive scanning sequence, achieving efficient and high-precision polarization state-wavelength testing.

[0098] The technical solution has the following advantages: by using the scheme provided in the embodiment, the trigger error between the scanning light source and the polarization controller is ensured to be less than 1 microsecond, and a lightweight neural network model that combines CNN and bidirectional 128-dimensional LSTM is embedded to train a scanning path prediction model using at least 1000 sets of historical test data, thereby achieving intelligent skipping of non-critical wavelength points and generation of an adaptive scanning sequence. The scanning step can be dynamically adjusted according to the polarization state change characteristics and test result confidence. Small-step dense scanning is used in key areas (such as areas with large polarization state fluctuations and unstable confidence), and large-step jump scanning is used in non-critical areas (such as areas with smooth polarization state changes and high confidence), thereby balancing the efficiency and accuracy of the test.

[0099] A data fusion system for a multi-channel synchronous optical wavelength scanning test module, such as Figure 3As shown, it comprises: a data acquisition module, a preprocessing module, a correlation analysis module, an error correction module and a result output module;

[0100] The data acquisition module comprises a universal data acquisition adapter configured to support serial ports, Ethernet ports and optical fiber interfaces, and compatible with Modbus protocol, TCP / IP protocol and SNMP protocol, for synchronously working with the multi-channel scanning optical power meter, wavelength resolution unit and polarization controller of the multi-channel synchronous optical wavelength scanning test module, to collect polarization state-wavelength correlation data at a frequency matching the scanning rate of the scanning light source; the polarization state-wavelength correlation data comprises optical power data, corresponding wavelength values, polarization state identifiers and sampling time stamps of each channel;

[0101] The preprocessing module is used for preprocessing the polarization state-wavelength correlation data; the preprocessing includes but is not limited to: removing noise interference by using wavelet transform, filling in missing data values by using K-nearest neighbor interpolation method, and mapping the polarization state-wavelength correlation data to the interval [-1, 1] by using maximum and minimum value normalization;

[0102] The preprocessing module further comprises an intelligent data mapping unit, which automatically identifies the semantics of multi-source heterogeneous data by using a lightweight semantic parsing model based on MobileBERT, and unifies the data units according to a preset unit conversion rule library; the lightweight semantic parsing model based on MobileBERT compresses the model volume to 26% of the original BERT model through knowledge distillation technology, and realizes real-time identification of unit semantics on an embedded platform; the training method of the lightweight semantic parsing model based on MobileBERT is as follows: first, a large number of multi-source heterogeneous data samples containing different units are collected, which cover various data units related to optical wavelength scanning test, such as power units and wavelength units, etc.; then, the samples are divided into training set, validation set and test set according to a certain proportion; in the training process, the original BERT model is used as the teacher model, and MobileBERT is used as the student model; knowledge distillation technology is used to transfer the knowledge of the teacher model to the student model; in this process, the network parameters of MobileBERT are adjusted so that it can realize similar semantic recognition ability to the original BERT model under a smaller model volume;

[0103] The correlation analysis module is used for correlation analysis of the preprocessed normalized matrix according to the constructed corresponding relationship model of polarization state and wavelength; the corresponding relationship model is constructed based on a Gaussian process regression model; the Gaussian process regression model uses a square exponential kernel function;

[0104] The error correction module is used for correcting errors in the polarization state-wavelength correlation data according to the results obtained by the correlation analysis module;

[0105] The results output module presents the preprocessed, correlation-analyzed, and error-corrected data in the form of charts and data files. The charts include polarization-wavelength loss curves and power comparison charts for each channel. The data files include raw data in CSV format and analysis results in JSON format.

[0106] The working principle of the above technical solution is as follows: To realize a data fusion system used in a multi-channel synchronous optical wavelength scanning test module, this invention proposes a data acquisition module. This module, through a universal data acquisition adapter, synchronously acquires data from the multi-channel scanning optical power meter, wavelength resolution unit, and polarization controller of the multi-channel synchronous optical wavelength scanning test module at a frequency matched to the scanning rate of the scanning light source. This yields polarization-wavelength correlation data containing optical power data for each channel, corresponding wavelength values, polarization state identifiers, and sampling timestamps. The acquired data first enters a preprocessing module, which uses wavelet transform to denoise the original data and eliminate noise interference. Then, K-nearest neighbor interpolation is used to fill in any missing data values ​​to ensure the integrity of the data sequence. Finally, maximum and minimum value normalization is used to uniformly map all polarization-wavelength correlation data to the [-1,1] interval for subsequent processing. Simultaneously, the intelligent data mapping unit in the preprocessing module activates a lightweight semantic parsing model based on MobileBERT to automatically perform semantic recognition on multi-source heterogeneous data. The data meaning is clearly defined, and units from different sources are unified according to a pre-defined unit conversion rule library to avoid unit confusion affecting the analysis. After preprocessing, the data is sent to the correlation analysis module. This module calls a pre-built polarization state-wavelength correspondence model based on a Gaussian process regression model to perform deep correlation analysis on the preprocessed data, uncovering the inherent laws and quantitative relationships between polarization state and wavelength. Subsequently, the error correction module identifies and evaluates the systematic or random errors in the polarization state-wavelength correlation data based on the analysis results from the correlation analysis module, and makes targeted corrections to improve the accuracy of the data. Finally, the data after the above series of processing is sent to the results output module. The results output module presents the processed data in two forms: charts and data files. The charts include polarization state-wavelength loss curves that visually show the change of polarization state with wavelength, as well as power comparison charts for different channels to compare their performance. The data files contain raw data records in CSV format and final analysis results in JSON format for users to view, archive, or use for further purposes.

[0107] For example, in the test of a certain optical fiber communication device, the scanning light source is scanned from 1550 nm to 1600 nm at a rate of 100 nm / s, and the data acquisition module synchronously acquires data of 3 channels at a frequency of 1 kHz; the preprocessing module first denoises the optical power data, fills in the missing values of the power data at 2 time points caused by weak signal at a certain time, and then normalizes the power data of all channels; the intelligent data mapping unit identifies that the wavelength data of one of the channels is in microns, and automatically converts it to nanometers, which is unified with other channels; the correlation analysis module finds that, near the wavelength of 1550 nm, the 0° polarization state and the optical power present a significant quadratic function relationship through the Gaussian process regression model analysis; the error correction module accordingly fine-tunes the power data near the wavelength, and corrects the about 0.2 dB error introduced by the system nonlinearity; the result output module finally generates clear polarization state-wavelength loss curves of the three channels, displays that the loss of channel 2 is the lowest at 1550 nm, and provides a CSV raw data and a JSON format analysis report containing regression coefficients, error ranges, etc.

[0108] The above technical solution has the beneficial effects that: by using the scheme provided in the embodiment, the comprehensiveness of data acquisition, the accuracy and compatibility of preprocessing, the depth of correlation analysis and the accuracy of data are effectively improved, high-quality data support and intuitive analysis results can be provided for the test of optical fiber communication devices and the like, and the user can conveniently view, archive and further utilize.

[0109] In one embodiment, the error correction module includes a quantum noise analysis subunit; the quantum noise analysis subunit is configured to generate a phase noise spectrum based on the output of the coherent mixer after fast Fourier transform conversion, and identify that the noise source is thermal noise or quantum noise; when the proportion of quantum noise is > 30%, a measurement mode of balanced homodyne detection is started to reduce the background noise; the step of fast Fourier transform conversion is as follows: first, discrete sampling is performed on the time domain signal output by the coherent mixer to obtain a series of discrete time series data; then, the discrete data is substituted into the algorithm formula of fast Fourier transform (FFT), and the signal is analyzed in the frequency domain; in the operation process, the time domain signal is converted into a frequency domain signal according to the operation rule of FFT, wherein the frequency resolution depends on the number of sampling points and the sampling frequency; after the fast Fourier transform conversion, the distribution of different frequency components can be clearly seen, so as to generate the phase noise spectrum.

[0110] The working principle of the technical solution is that the error correction module realizes accurate identification and targeted processing of system noise through the quantization noise analysis subunit. The subunit first obtains the phase noise spectrum generated by the fast Fourier transform conversion of the coherent mixer output. This noise spectrum contains various noise information of the system in a specific working state. Through feature analysis of the noise spectrum, it can effectively distinguish whether the main source of noise is thermal noise or quantum noise. Thermal noise and quantum noise have identifiable differences in power density distribution and frequency characteristics of the noise spectrum. For example, quantum noise may exhibit unique shot noise characteristics or vacuum fluctuation characteristics at a specific frequency band, while thermal noise usually has a wider frequency spectrum distribution and temperature-dependent power characteristics. When the quantum noise analysis subunit determines that the proportion of quantum noise in the total noise exceeds 30%, it means that quantum noise has become one of the dominant factors affecting system performance. At this time, the error correction module will start the measurement mode of balanced homodyne detection. The balanced homodyne detection technology mixes two phase-quadrature local oscillation lights with signal light, and then detects the mixed light signals with two photodetectors respectively. Then, the outputs of the two detectors are differentially processed. This differential processing can effectively suppress common-mode noise, especially the vacuum fluctuation component in quantum noise, thereby significantly reducing the system's noise floor and improving the measurement accuracy and signal-to-noise ratio.

[0111] For example, in a quantum communication system, when the phase noise spectrum generated by the fast Fourier transform conversion of the coherent mixer output shows that the proportion of quantum noise is 35%, the quantum noise analysis subunit identifies this situation, and the error correction module immediately switches to the balanced homodyne detection mode. Through the above differential detection mechanism, the noise floor can be reduced by about 15-20 decibels, ensuring accurate demodulation and information extraction of subsequent quantum states.

[0112] The beneficial effects of the above technical solution are that the scheme provided in this embodiment can accurately identify the source of system noise, and when the proportion of quantum noise exceeds 30%, it can effectively suppress common-mode noise by starting the measurement mode of balanced homodyne detection, especially the vacuum fluctuation component in quantum noise, thereby significantly reducing the system's noise floor and improving the measurement accuracy and signal-to-noise ratio.

[0113] In one embodiment, the environmental monitoring and compensation module includes:

[0114] The temperature and humidity sensor and the vibration sensor are used to collect environmental parameters in real time. The temperature sensor has an accuracy of ±0.1℃ (-10℃-60℃), the humidity sensor has an accuracy of ±1%RH (10%-90%RH), and the vibration sensor has a measurement range of ±5g and a frequency response of 0-500Hz.

[0115] The storage unit is used to construct an error mapping model library for storing the mapping relationship between the environmental parameters and the wavelength deviation / optical power drift; the error mapping model is based on a trained BP neural network, the input layer of which is temperature, humidity and vibration acceleration, the hidden layer of which is 3 layers (64 / 32 / 16 neurons), the activation function of which is ReLU, and the output layer of which is the wavelength deviation value and the optical power drift value, and the training data is 1000 groups of corresponding samples of environmental parameters and drift values, and the fitting error is less than 0.5 pm / 0.005 decibels;

[0116] The dynamic correction engine unit is used to call the mapping relationship to compensate the optical power data and wavelength values in the corrected polarization state-wavelength associated data in real time.

[0117] The working principle of the above technical solution is as follows: first, the temperature and humidity sensor and the vibration sensor in the sensor assembly continuously and in real time collect the temperature and humidity parameters of the current environment and the vibration parameters during the operation of the equipment, which together constitute the key environmental parameters affecting the characteristics of the optical signal in the optical communication system; then, the error mapping model library is pre-constructed and stored in the storage unit, which includes the specific mapping relationship between different environmental parameters such as temperature, humidity and vibration and the wavelength deviation value and the optical power drift of the optical signal; when the system is running, the dynamic correction engine unit will receive the current environmental parameter data from the sensor assembly in real time, and accurately call the corresponding mapping relationship between the environmental parameters and the wavelength deviation / optical power drift from the error mapping model library of the storage unit according to the data; then, the dynamic correction engine unit uses the called mapping relationship to further compensate the optical power data and wavelength values in the polarization state-wavelength associated data in real time;

[0118] For example, when the temperature and humidity sensor detects that the environmental temperature suddenly rises by 5℃, and the vibration sensor detects that the vibration amplitude increases by 0.2mm, the dynamic correction engine unit will immediately find the wavelength deviation value (assuming +0.03 nanometers) and the optical power drift value (assuming -0.5dBm) corresponding to the temperature rise of 5℃ and the vibration amplitude increase of 0.2mm from the error mapping model library, and then apply these deviation values and drift values to the current wavelength value and optical power data respectively for real-time adjustment, so as to ensure that the optical power and wavelength information in the finally output polarization state-wavelength associated data can accurately reflect the actual optical signal characteristics and eliminate the interference caused by environmental factors.

[0119] The beneficial effects of the above technical solution are as follows: by using the scheme provided in the embodiment, the environmental parameters are collected in real time by the temperature and humidity sensor and the vibration sensor, the BP neural network error mapping model library trained based on 1000 samples, and the dynamic correction engine can call the model to compensate the optical power and wavelength values in the polarization state-wavelength associated data in real time, which can effectively eliminate environmental interference and ensure that the output data accurately reflects the actual optical signal characteristics.

[0120] In one embodiment, a scan parameter adjustment control module is further included; the scan parameter adjustment control module comprises a cooperative interface configuration unit, a data anomaly prediction unit, a scan parameter adjustment instruction generation unit, and a feedback data receiving and processing unit;

[0121] The cooperative interface configuration unit is configured to configure a real-time cooperative interface between the data fusion system and the integrated platform of the multi-channel synchronous optical wavelength scanning test module, and the real-time cooperative interface is configured to adopt a UDP protocol for data interaction.

[0122] The data anomaly prediction unit is configured to input a feature vector of the last 100 groups of polarization state-wavelength correlation data based on a random forest model; the feature vector includes but is not limited to a loss value, a fluctuation slope, and a correlation score; and the data anomaly prediction unit is configured to output an anomaly probability of a plurality of future wavelength points; the resolution of the anomaly probability is higher than 1%.

[0123] The scan parameter adjustment instruction generation unit is configured to generate a scan parameter adjustment instruction when the anomaly probability is greater than 80%, and send the scan parameter adjustment instruction to the integrated platform through the cooperative interface; the scan parameter adjustment instruction includes reducing the scanning rate in the anomaly interval by 50%, reducing the scanning step to 1 / 10 of the original step, and synchronously increasing the sampling rate of the multi-channel scanning optical power meter by 50%.

[0124] The feedback data receiving and processing unit is configured to receive the collected data of the scanning optical power meter after the parameter adjustment through the real-time cooperative interface, and if the amount of collected data increases by 30% and the following conditions are met simultaneously:

[0125] (a) The data quality comprehensive score calculated in real time is higher than a preset threshold value 0.7; wherein, the calculation formula of the quality comprehensive score is:

[0126]

[0127] wherein, represents the data quality comprehensive score, represents the data signal-to-noise ratio after the current parameter adjustment, represents the reference signal-to-noise ratio before the parameter adjustment, , , is a weight coefficient, and + + = 1, is the standard deviation of the current data fluctuation slope, is the standard deviation of the reference fluctuation slope before the parameter adjustment, represents a moving window sequence of the correlation scores of 20 consecutive polarization state-wavelengths, is the variance of the moving window sequence of the correlation scores of 20 consecutive polarization state-wavelengths. is a threshold value, and the value is 0.1; the correlation score is calculated by the Pearson correlation coefficient of the polarization state-wavelength data;

[0128] (b) applying the reinforcement learning DQN algorithm model, inputting the current collected data and the historical adjustment record, and outputting the confidence of the parameter optimization strategy ≥ 90%;

[0129] (c) verifying that the proportion of abnormal points in the data is <5% by the Isolation Forest anomaly detection algorithm;

[0130] then the following operations are performed: generating a scanning parameter recovery instruction, sending it to the integrated platform through the real-time collaborative interface, gradually restoring the original scanning parameters, and triggering online iterative training of the data anomaly prediction unit model, and updating the weight of the random forest model using the newly collected data;

[0131] Otherwise, if any condition is not met, the current adjustment parameter is maintained, an incremental data collection instruction is generated, and the feedback data is input to the data anomaly prediction unit to optimize the feature vector extraction algorithm in real time to improve the accuracy of future anomaly probability prediction; the incremental data collection instruction includes: increasing the scanning rate to the original level, and restoring the scanning step to the original size; the incremental data collection instruction is: increasing the sampling rate by 20% based on the adjusted basis, while maintaining the current scanning rate and scanning step.

[0132] The working principle of the above technical solution is: the data interaction between the data fusion system and the multi-channel synchronous optical wavelength scanning test module integrated platform is realized by configuring the UDP protocol real-time collaborative interface; the random forest model is used to input the last 100 groups of polarization state-wavelength correlation data feature vectors containing loss value, fluctuation slope, correlation score, etc., to predict the anomaly probability of the future 5 wavelength points (resolution 0.01); when the anomaly probability > 80%, a scanning parameter adjustment instruction (abnormal interval scanning rate reduced by 50%, scanning step reduced to 1 / 10 of the original, sampling rate increased by 50%) is generated and sent through the collaborative interface; receiving the collected data after parameter adjustment, if the data volume increases by 30% and at the same time meets: the data quality comprehensive score is higher than 0.7, the DQN algorithm output parameter optimization strategy confidence ≥ 90%, and the proportion of abnormal points detected by the Isolation Forest is <5%, a scanning parameter recovery instruction (gradually restoring the original parameters) is generated and the random forest model is triggered for online iterative training; if any condition is not met, the current adjustment parameter is maintained, an incremental data collection instruction (sampling rate increased by 20%) is generated, and the feedback data is input to the data anomaly prediction unit to optimize the feature vector extraction algorithm;

[0133] For example, a certain communication equipment manufacturer conducts a fiber polarization state-wavelength correlation test, and the data fusion system interacts with the multi-channel synchronous optical wavelength scanning test module in real time through the UDP protocol. First, the UDP protocol cooperation interface is configured, and the polarization state-wavelength data feature vector containing the loss value, fluctuation slope, correlation degree score, etc. is sent to the fusion system in the form of a UDP message every time the test module completes a group of scans, with a transmission delay of less than 5 ms. Abnormal prediction: the system accumulates the last 100 groups of feature vectors input into the random forest model to predict the abnormal probability of the next 5 wavelength points, of which 3 wavelength points have an abnormal probability greater than 80%, triggering parameter adjustment. Generate adjustment instructions, reduce the scanning rate from 100 nm / s to 50 nm / s, reduce the scanning step from 0.1 nm to 0.01 nm, and increase the sampling rate from 1 kHz to 1.5 kHz. The instructions are issued through the UDP interface, and the module completes the switching within 200 ms. If the data volume increases by 42% after parameter adjustment, and the data quality comprehensive score is 0.82 (> 0.7), the DQN algorithm outputs a parameter optimization strategy confidence of 93% (≥ 90%), and the isolated forest detects an abnormal point ratio of 3.2% (< 5%), which meets the conditions, then generate a scan parameter recovery instruction to gradually restore the original parameters in 3 steps, with an interval of 5 s, and trigger online iterative training of the random forest model.

[0134] The beneficial effects of the above technical solutions are: the scheme provided by the embodiment can realize real-time monitoring of data in multi-channel synchronous optical wavelength scanning test, early prediction of abnormalities and timely dynamic adjustment of scanning parameters; low-delay data interaction is realized through the UDP protocol, and the random forest model is used to accurately predict the data abnormal probability. When the abnormal probability is high, the scanning rate is reduced, the scanning step is reduced, and the sampling rate is increased to optimize data acquisition. After adjustment, the parameter adjustment effect is verified through data volume growth, data quality comprehensive score, DQN algorithm confidence, and isolated forest anomaly detection, etc. to ensure that high-quality data is obtained. When effective, restore the parameters and iteratively optimize the model. When invalid, further increase the sampling rate and optimize the prediction algorithm, thereby significantly improving the system stability and data acquisition accuracy, meeting the strict requirements of complex optical test scenarios for data precision and reliability.

[0135] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include these modifications and variations.

Claims

1. A multi-channel simultaneous optical wavelength swept test module, characterized by, The application relates to a scanning light source, a multi-channel scanning light power meter, a wavelength resolution unit, a polarization controller, a trigger synchronization unit and an integrated platform. The scanning light source is configured to provide a basic light source. The multi-channel scanning light power meter is configured to simultaneously measure the light power of multiple channels to obtain light power data. The wavelength resolution unit is configured to resolve the wavelength of the light signal output by the basic light source to determine the wavelength value of the light signal. The trigger synchronization unit is configured to realize linkage control between the scanning light source and the polarization controller to form polarization state-wavelength coordinated scanning, thereby synchronously completing wavelength scanning and switching of multiple polarization states. The integrated platform is configured to uniformly control and manage the scanning light source, the multi-channel scanning light power meter, the wavelength resolution unit, the polarization controller and the trigger synchronization unit. The multi-channel scanning light power meter is configured to support >=48 channels of synchronous sampling. The multi-channel scanning light power meter is integrated with a coherent crosstalk suppression unit. The coherent crosstalk suppression unit comprises: A local oscillation light source configured to output a local oscillation light signal with adjustable frequency. A coherent mixer configured to mix the input light signal with the local oscillation light signal to generate a mixed electric signal carrying amplitude and phase information. A crosstalk compensation processor configured to calculate a crosstalk compensation factor based on the mixed electric signal of the adjacent channel, and dynamically correct the light power data of the current channel by using the crosstalk compensation factor, so that the crosstalk error is <=+ / -0.01 decibel. The wavelength range of the scanning light source is continuously adjustable in the range of 1490-1630 nanometers; the set value of the scanning rate is greater than or equal to 100 nanometers / second; the set value of the scanning step is less than or equal to 1 picometer; the scanning light source is configured to selectively carry an embedded semiconductor laser amplifier light source or an external third-party light source; the embedded semiconductor laser amplifier light source is configured to: increase the power of the light signal output by the basic light source to 10dBm, and has a standardized optical interface.

2. The multi-channel synchronous optical wavelength scanning test module of claim 1, wherein, The wavelength resolution accuracy of the wavelength resolution unit is configured to be <+ / -10pm; the wavelength range of the wavelength resolution unit is configured to be 1490-1630 nanometers; the maximum insertion loss of the wavelength resolution unit is configured to be <=0.8 decibels; and the sampling rate of the wavelength resolution unit is <=100k points / second / channel.

3. The multi-channel synchronous optical wavelength scanning test module of claim 1, wherein, The polarization controller is configured to switch between -45° linear polarization state, 0° linear polarization state, +45° linear polarization state, +90° linear polarization state, left-handed circular polarization state and right-handed circular polarization state during wavelength scanning, and the adjustment time of switching is not greater than 250 microseconds.

4. The multi-channel synchronous optical wavelength scanning test module of claim 1, wherein, ​ 5. The multi-channel simultaneous optical wavelength swept test module of claim 1, wherein, The trigger synchronization unit is configured to support both electrical triggering and optical triggering modes, and is constructed to synchronize the scanning light source with the trigger error of the polarization controller being less than 1 microsecond through an FPGA clock; the trigger synchronization unit comprises an AI prediction control subunit embedded with a lightweight neural network model, the AI prediction control subunit is built-in a scanning path prediction submodel, and the scanning path prediction submodel is generated based on historical polarization state-wavelength test data sets; the AI prediction control subunit is configured to skip non-critical wavelength points based on the scanning path prediction submodel, generate an adaptive scanning sequence, and control the scanning light source and the polarization controller to perform variable-step collaborative scanning based on the adaptive scanning sequence; The lightweight neural network model is a hybrid model of a convolutional neural network and a long short-term memory network, training data is collected from no less than 1000 groups of historical test data, the historical test data includes wavelength scanning trajectories, polarization state switching timing, historical noise distribution and test result confidence labels; the convolutional neural network is a three-layer one-dimensional structure, and the sizes of the convolution kernels are 11, 7 and 5 respectively, for extracting local time-frequency features; the long short-term memory network is a bidirectional 128-dimensional structure, for capturing time-dependent features based on local time-frequency features, and dynamically adjusting scanning steps and key scanning areas according to the time-dependent features to generate an adaptive scanning sequence.

6. A data fusion system for use with a multi-channel simultaneous optical wavelength swept test module as claimed in any one of claims 1 to 5, characterized in that, Comprise: a data acquisition module, a preprocessing module, a correlation analysis module, an error correction module and a result output module; The data acquisition module comprises a universal data acquisition adapter, the universal data acquisition adapter is configured to support serial ports, Ethernet ports and optical fiber interfaces, and is compatible with Modbus protocols, TCP / IP protocols and SNMP protocols, and is used to work synchronously with a multi-channel scanning optical power meter, a wavelength analysis unit and a polarization controller of a multi-channel synchronous optical wavelength scanning test module, to collect polarization state-wavelength correlation data at a frequency matching the scanning rate of the scanning light source; the polarization state-wavelength correlation data includes optical power data, corresponding wavelength values, polarization state identifiers and sampling time stamps of each channel; The preprocessing module is used for preprocessing the polarization state-wavelength correlation data; The preprocessing includes: removing noise interference by using wavelet transform, filling in missing data by using K nearest neighbor interpolation method, and mapping the polarization state-wavelength correlation data to the interval [-1, 1] by using maximum and minimum value normalization; The preprocessing module further comprises an intelligent data mapping unit, the intelligent data mapping unit automatically identifies the semantics of multi-source heterogeneous data by using a lightweight semantic analysis model based on MobileBERT, and unifies the data units according to a preset unit conversion rule library; The correlation analysis module is used for correlating the normalized matrix after preprocessing according to a constructed polarization state and wavelength correspondence model; the correspondence model is constructed based on a Gaussian process regression model; The error correction module is used for correcting errors in the polarization state-wavelength correlation data according to the results obtained by the correlation analysis module. The result output module is configured to present the data after preprocessing, correlation analysis and error correction in the form of charts and data files. The charts include polarization state-wavelength loss curves and channel power comparison charts. The data files include original data in CSV format and analysis results in JSON format.

7. The data fusion system of claim 6, wherein, The error correction module includes a quantization noise analysis subunit configured to generate a phase noise spectrum based on the output of the coherent mixer after fast Fourier transform conversion, and identify the noise source as thermal noise or quantum noise. When the proportion of quantum noise is greater than 30%, the measurement mode of balanced homodyne detection is started to reduce the noise floor.

8. The data fusion system of claim 6, wherein, The environmental monitoring and compensation module includes: A temperature and humidity sensor and a vibration sensor are configured to collect environmental parameters in real time. A storage unit is configured to build an error mapping model library and store the mapping relationship between environmental parameters and wavelength deviation / optical power drift. A dynamic correction engine unit is configured to call the mapping relationship to compensate the optical power data and wavelength values in the corrected polarization state-wavelength correlation data in real time.

9. The data fusion system of claim 6, wherein, The scanning parameter adjustment control module includes a cooperative interface configuration unit, a data anomaly prediction unit, a scanning parameter adjustment instruction generation unit and a feedback data receiving and processing unit. The cooperative interface configuration unit is configured to configure a real-time cooperative interface between the data fusion system and the integrated platform of the multi-channel synchronous optical wavelength scanning test module, and the real-time cooperative interface is configured to use the UDP protocol for data interaction. The data anomaly prediction unit is configured to input the feature vectors of the last 100 groups of polarization state-wavelength correlation data based on a random forest model. The feature vectors include loss values, fluctuation slopes and correlation scores. The output is the abnormal probability of future wavelength points. The resolution of the abnormal probability is higher than 1%. The scanning parameter adjustment instruction generation unit is configured to generate a scanning parameter adjustment instruction when the abnormal probability is greater than 80%, and send the scanning parameter adjustment instruction to the integrated platform through the cooperative interface. The scanning parameter adjustment instruction includes reducing the scanning rate in the abnormal interval by 50%, reducing the scanning step to 1 / 10 of the original step, and synchronously increasing the sampling rate of the multi-channel scanning optical power meter by 50%. The feedback data receiving and processing unit is configured to receive the collected data of the scanning optical power meter after parameter adjustment through the real-time cooperative interface. If the amount of collected data increases by 30% and the following conditions are met simultaneously: (a) The data quality comprehensive score calculated in real time is higher than the preset threshold of 0.7; (b) The confidence of the parameter optimization strategy output by the reinforcement learning DQN algorithm model is greater than or equal to 90% by inputting the current collected data and historical adjustment records; (c) The proportion of abnormal points in the data verified by the isolation forest anomaly detection algorithm is less than 5%; the following operations are performed: a scanning parameter recovery instruction is generated and sent to the integrated platform through the real-time cooperative interface to gradually restore the original scanning parameters, and the model online iterative training of the data anomaly prediction unit is triggered to update the weight of the random forest model with new collected data. Otherwise, if any condition is not met, the current adjustment parameter is maintained, an incremental data acquisition instruction is generated, and the feedback data is input to the data anomaly prediction unit to optimize the feature vector extraction algorithm in real time to improve the accuracy of future anomaly probability prediction; the incremental data acquisition instruction includes: increasing the scanning rate to the original level, and restoring the scanning step to the original size; the incremental data acquisition instruction is: increasing the sampling rate by 20% based on the adjusted basis, while maintaining the current scanning rate and scanning step.

Images