Clock frequency drift algorithm monitoring sensing time service protection method and device terminal

Abstract

The invention relates to the technical field of time service safety, and provides a clock frequency drift algorithm monitoring sensing time service protection method and device terminal. According to the method, firstly, clock frequency signals meeting requirements are collected, and preprocessing such as discrete sampling, denoising and resampling is carried out; secondly, calculating and analyzing clock frequency drift features by using mean drift estimation and variance, and constructing feature vectors; and inputting the feature vector into a machine learning classifier trained by historical data, identifying abnormal frequency drift, and realizing perception of a time service error. When abnormity is detected, early warning information including a frequency value, a drift degree and time service error estimation is generated and sent to a monitoring terminal through a wireless communication module. And fusion and weight adjustment of monitoring data of a plurality of clock sources are also included, so that the accuracy of time service protection is improved. The device further comprises a power supply module and a shell, and the system is composed of a plurality of devices and a central monitoring server.

Description

Technical Field

[0001] The present invention relates to the technical field of timing security, and in particular to a monitoring and protection method for a timing signal clock frequency drift characteristic and a corresponding device and terminal equipment. Background Art

[0002] Clock frequency drift refers to the deviation between the actual frequency of the clock and the ideal frequency, which gradually accumulates over time. In the timing system, clock frequency drift may cause inaccurate timing, which will affect various applications that rely on precise time, such as communications, power systems, navigation, etc. For example, in the 5G power timing solution, a clock error of less than 1us needs to be achieved, and clock frequency drift may undermine this requirement. Timing protection is a series of measures to prevent or reduce the impact of various factors (including clock frequency drift) on the accuracy of timing. By checking the characteristics of clock frequency drift, the deviation of the clock can be discovered and corrected in time, thereby improving the accuracy and reliability of timing.

[0003] With the rapid advancement of science and technology, the Global Positioning System (GPS) and other satellite navigation systems have been widely used around the world, providing key time and location information for many fields. The popularity of this technology has gradually highlighted the security and accuracy of timing signals, becoming a hot topic in related research fields. Clock frequency drift, as a key factor affecting the accuracy of timing signals, has attracted widespread attention. This drift phenomenon may cause deviations in timing signals, which in turn poses a potential threat to systems and equipment that rely on precise time synchronization, affecting their normal operation and performance.

[0004] In the current technological context, the security and reliability of timing signals cannot be ignored. Many key areas, such as financial transactions, communication networks, transportation, and national defense security, are highly dependent on accurate timing signals to ensure the stable operation of their systems and equipment. Therefore, it is crucial to develop an effective monitoring and protection method to cope with the challenges brought by clock frequency drift.

[0005] The monitoring and protection technology for clock frequency drift has become a research hotspot. Many scholars and research institutions are committed to developing more advanced and accurate algorithms to improve the stability and security of timing signals. These studies not only involve in-depth analysis of clock frequency drift characteristics, but also include optimization and improvement of monitoring methods, as well as the development and application of related devices and terminal equipment.

[0006] The present invention proposes an innovative algorithm based on clock frequency drift feature inspection, which is applied to the monitoring perception timing protection method and device terminal. This method can achieve efficient identification and prevention of potential risks in the time synchronization process by accurately capturing subtle changes in clock frequency. In a complex network environment, especially in the face of increasingly severe time synchronization security challenges, the algorithm of the present invention significantly improves the security protection capability of the timing system with its high sensitivity and strong robustness.

[0007] Beneficial Effects

[0008] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0009] First, it can accurately identify the clock frequency drift characteristics in the timing signal. By adopting an advanced clock frequency drift feature detection algorithm, the present invention can accurately detect and identify the clock frequency drift phenomenon in the timing signal. The realization of this recognition capability is due to the algorithm's deep analysis and extraction of signal characteristics, thereby ensuring the accuracy and stability of the timing signal.

[0010] Second, it can timely sense and protect against timing signal anomalies, thereby improving the security and reliability of timing signals. Once clock frequency drift is detected, the system can respond quickly and take effective protective measures to prevent abnormal signals from interfering with or damaging systems and equipment that rely on precise time synchronization. This timely perception and protection mechanism significantly improves the security and robustness of the timing system.

[0011] Third, the present invention is applicable to a variety of systems and devices that rely on precise time synchronization, and has broad application prospects. Whether it is a communication system, a power system or a financial transaction system, precise time synchronization is crucial. The present invention provides a strong guarantee for the stable operation of these key systems by providing an efficient and accurate clock frequency drift detection and protection mechanism, and therefore has broad market application and promotion value. Summary of the invention

[0012] The present invention relates to the field of timing security technology, and in particular to a timing protection method and device terminal based on algorithm monitoring and sensing of clock frequency drift feature inspection. The basic principles, main methods and performance evaluation indicators of timing protection are summarized in detail below.

[0013] Specifically, the present invention includes a clock frequency drift algorithm monitoring perception timing protection method, comprising the following steps:

[0014] Step S001: Obtain a clock frequency signal through a sensor and perform discrete sampling on it to obtain a discrete sequence ;

[0015] Step S002: Use the mean drift estimation in the clock frequency drift algorithm to calculate the mean of the clock frequency signal within a certain time window. ,in is the number of sampling points in the time window, which is used to preliminarily evaluate the central trend of the clock frequency;

[0016] Step S003: Calculate the variance of the clock frequency signal , determine the degree of fluctuation of the clock frequency according to the size of the variance;

[0017] Step S004: Constructing frequency drift feature vector , which will serve as the basis for subsequent analysis;

[0018] Step S005: Transform the feature vector The input is sent to a machine learning-based classifier, which is trained using labeled historical clock frequency drift data. The loss function used in the training process is cross entropy loss. ,in is the number of categories, is the true label, In order to predict the probability, the abnormal frequency drift is identified through the output of the classifier, so as to realize the perception of the possible errors in timing.

[0019] Further, the following steps are included:

[0020] Step S001: Before calculating the mean and variance, the collected clock frequency signal is denoised by using a wavelet transform denoising method, by selecting an appropriate wavelet basis function and the number of decomposition levels , for the signal Perform wavelet decomposition ,in is the wavelet coefficient, The high-frequency noise components in the signal are removed by stretching and translating the wavelet basis function;

[0021] Step S002: resample the denoised signal so that the sampling frequency of the signal meets the requirements of subsequent analysis. The resampling formula is: ,in is the resampled signal, is the new sampling period.

[0022] Further, the following steps are included:

[0023] Step S001: After constructing the feature vector, the feature vector is normalized using the minimum-maximum normalization method. The formula is: ,in is the normalized eigenvalue, is the original eigenvalue, and are the minimum and maximum values ​​of the feature values, respectively, so that features of different dimensions have the same scale and improve the performance of the classifier;

[0024] Step S002: Regularly update the classifier parameters, using the stochastic gradient descent algorithm based on the newly acquired clock frequency drift data ,in are the classifier parameters, is the learning rate, The gradient of the loss function with respect to the parameters is used to fine-tune the classifier to adapt to the changes in the clock frequency drift characteristics.

[0025] Further, the following steps are included:

[0026] Step S001: When the classifier identifies abnormal frequency drift, it generates warning information, which includes the abnormal frequency value, the degree of drift and the possible timing error estimation;

[0027] Step S002: Send the warning information to the monitoring terminal through the wireless communication module. The wireless communication module adopts a specific modulation coding method. For binary phase shift keying (BPSK) modulation, the formula is: ,in To send a signal, is the signal amplitude, is the carrier angular frequency, It is a binary information sequence for signal transmission.

[0028] Further, the following steps are included:

[0029] Step S001: comprising the method steps described in any one of claims 1 to 4;

[0030] Step S002: Fusion the monitoring data of multiple clock sources. The fusion formula is: ,in is the fused monitoring data, For the The weight of the clock source, For the Monitoring data of clock sources, The number of clock sources is considered comprehensively to improve the accuracy of timing protection;

[0031] Step S003: Adjust the weight of each clock source according to the integrated monitoring data , the adjustment formula is ,in is the adjusted weight, It is the weight adjustment amount, which is determined according to the deviation between the fused data and the historical data to optimize the performance of the entire timing protection system.

[0032] Further, the following steps are included:

[0033] Step S001: Before fusing the monitoring data of multiple clock sources, the data of different clock sources are time-aligned, and a time synchronization algorithm based on the least squares method is used to minimize the error function. ,in For the The clock source Sampling point time, To ensure the consistency of the data of each clock source in time after alignment;

[0034] Step S002: Perform correlation analysis on the time-aligned clock source data and calculate the correlation coefficients between different clock sources. ,in and Respectively and The clock source The sampling point value, and is the corresponding mean, and the weight is adjusted according to the size of the correlation The initial value of .

[0035] Further, the following steps are included:

[0036] Step S001: According to the adjusted weight , recalculate the fused monitoring data ;

[0037] Step S002: Feedback the recalculated fusion data to each clock source to update the classifier parameters and adjust the warning threshold to further optimize the timing protection effect.

[0038] Furthermore, the present invention also includes a component using this method, including a clock frequency sensor, a signal processing unit, a machine learning classifier processing unit, and a wireless communication module, wherein the clock frequency sensor is used to collect clock frequency signals, the signal processing unit is used to perform operations such as denoising, resampling, and feature normalization, the machine learning classifier processing unit is used to run the classifier algorithm, and the wireless communication module is used to send an early warning signal.

[0039] Furthermore, the present invention also includes a device using the method, including the components of implementing technical solution 8, and also including a power module and a casing, the power module supplies power to each component, and the casing protects the internal components.

[0040] Furthermore, the present invention also includes a system using the method, including multiple devices implementing technical solution 9, each device is connected through a wired or wireless network to achieve distributed timing protection, and also includes a central monitoring server for receiving and processing early warning information and monitoring data sent by each device, so as to uniformly manage and regulate the entire timing protection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings of the present invention show in detail the implementation process of the method, device and terminal equipment for algorithm monitoring and sensing timing protection based on clock frequency drift feature inspection. The specific drawings are as follows:

[0042] Figure 1 This is a schematic diagram of the clock frequency signal monitoring process;

[0043] Figure 2 This is a schematic diagram of the signal preprocessing process;

[0044] Figure 3 This is a schematic diagram of the feature vector normalization and classifier update process;

[0045] Figure 4 This is a schematic diagram of the abnormal frequency drift warning process;

[0046] Figure 5 This is a schematic diagram of the multi-clock source data fusion and weight adjustment process;

[0047] Figure 6 This is a schematic diagram of the time alignment and correlation analysis process;

[0048] Figure 7 This is a schematic diagram of the weight adjustment and feedback optimization process;

[0049] Figure 8 It is a schematic diagram of the monitoring components;

[0050] Fig. 9 This is a schematic diagram of the timing protection device;

[0051] Fig.10 This is a schematic diagram of a distributed timing protection system; DETAILED DESCRIPTION

[0052] Implementation method 1, combined with Figure 1 This is a schematic diagram of the clock frequency signal monitoring process, which includes:

[0053] First, the clock frequency signal is obtained through the sensor, which is the starting point of the whole method. The selection of the sensor is very important, and the equipment that meets the relevant high-precision sensor standards needs to be selected.

[0054] 2. Clock Frequency Drift Characteristics Analysis

[0055] Mean shift estimation: Use the mean shift estimation in the clock frequency drift algorithm to calculate the mean of the clock frequency signal within a certain time window. .in is the number of sampling points in the time window, and the size of this time window is determined according to relevant standards. The central trend of the clock frequency can be preliminarily evaluated by the mean estimation.

[0056] Feature vector construction: Construct frequency drift feature vector , which complies with the industry's frequency drift characteristic vector specification standards. This characteristic vector will serve as an important basis for subsequent analysis.

[0057] 3. Classification and Anomaly Identification Based on Machine Learning

[0058] The feature vector Input into the machine learning based classifier. The selection of the classifier needs to meet the relevant standards of machine learning classifiers, such as support vector machine (SVM) or neural network (NN). The classifier is trained with a large amount of labeled historical clock frequency drift data that meets the historical data collection standards. The training adopts the cross entropy loss function. ,in is the number of categories, is the true label, The prediction probability is obtained by identifying abnormal frequency drift through the output of the classifier, thus realizing the perception of possible errors in timing.

[0059] Implementation method 2, Figure 2 This is a schematic diagram of the signal preprocessing process, which includes:

[0060] 1. Clock frequency signal acquisition and preprocessing

[0061] 1. Sensor selection

[0062] As described in Implementation 1, it is critical to select equipment that meets the relevant high-precision sensor standards.

[0063] 2. Discrete Sampling

[0064] Discrete sampling is performed on the collected clock frequency signal to obtain a discrete sequence The determination of sampling frequency requires comprehensive consideration of the characteristics of the signal and the needs of subsequent analysis.

[0065] 3. Denoising

[0066] Use wavelet transform denoising method. Select appropriate wavelet basis function according to the characteristics of the signal (such as one of the Daubechies wavelet series) and the number of decomposition levels . For signal Perform wavelet decomposition , remove high-frequency noise components.

[0067] 4. Resampling

[0068] Resample the denoised signal so that the sampling frequency meets the requirements of subsequent analysis. The resampling formula is ,in is a new sampling period, which is determined according to relevant standards of signal processing and analysis.

[0069] 2. Clock Frequency Drift Characteristics Analysis

[0070] 1. Mean Shift Estimation

[0071] Use the mean drift estimation in the clock frequency drift algorithm to calculate the mean of the clock frequency signal within a certain time window .in is the number of sampling points in the time window. The size of this time window is determined according to relevant standards.

[0072] 2. Variance calculation

[0073] Calculate the variance of a clock frequency signal ,The fluctuation degree of the clock frequency is judged according to the size of the variance, and the judgment is based on the relevant fluctuation degree evaluation criteria.

[0074] 3. Feature vector construction

[0075] Constructing frequency drift feature vector , which is used as the basis for subsequent analysis. This feature vector integrates the central trend and fluctuation information of the clock frequency, providing effective input data for the classifier based on machine learning.

[0076] 3. Classification and Anomaly Identification Based on Machine Learning

[0077] 1. Classifier selection and training

[0078] The selection of the classifier needs to meet the relevant standards of machine learning classifiers, such as support vector machines (SVM) or neural networks (NN). The classifier is trained with a large amount of labeled historical clock frequency drift data that meets the historical data collection standards. The training adopts the cross entropy loss function. ,in is the number of categories, is the true label, is the predicted probability.

[0079] 2. Eigenvector normalization

[0080] In the feature vector Before inputting into the classifier, the feature vector is normalized by using the minimum and maximum normalization method that meets the data normalization standard to improve the classifier performance.

[0081] 3. Classification and anomaly identification

[0082] The normalized feature vector is input into the classifier, and the abnormal frequency drift is identified through the output of the classifier, so as to realize the perception of possible errors in timing.

[0083] The above parts can process and analyze the clock frequency more finely, improve the timing protection system's monitoring and perception capabilities of clock frequency drift, and thus better ensure the accuracy and reliability of the timing system.

[0084] Implementation method 3, Figure 3 This is a schematic diagram of the feature vector normalization and classifier update process, which includes:

[0085] 1. Feature vector processing

[0086] 1. Feature vector normalization

[0087] In constructing the feature vector After that, normalization is a crucial step. Using the minimum-maximum normalization method, the formula is .

[0088] 2. Importance and impact of normalization

[0089] The normalized feature vector provides a more standardized data input form for the classifier. It enables the classifier to better focus on the relative relationship between features rather than being disturbed by the absolute size of the feature value. In machine learning algorithms, many classifiers are sensitive to the scale of the data.

[0090] 2. Classifier Update

[0091] 1. Update requirements based on new data

[0092] As time goes by and the system runs, the clock frequency drift data will be continuously updated. In order for the classifier to adapt to these changes, the classifier parameters need to be updated regularly. This is because the characteristics of clock frequency drift may change due to environmental factors, equipment aging, etc. If the classifier parameters are not updated in time, it may gradually lose the ability to recognize the newly emerging frequency drift patterns, resulting in misjudgment or missed judgment.

[0093] 2. Application of Stochastic Gradient Descent Algorithm

[0094] Using stochastic gradient descent algorithm To update the parameters of the classifier. Here represents the parameters of the classifier, is the learning rate, is the gradient of the loss function with respect to the parameters.

[0095] 3. Anomaly Identification and Timing Error Perception

[0096] 1. Mechanism of action of the classifier

[0097] The normalized feature vector is input into the classifier. The classifier is trained with the labeled historical clock frequency drift data. The loss function in the training process adopts the cross entropy loss. .

[0098] The classifier judges the input feature vector based on the patterns it has learned internally. For example, in a training data set containing a variety of different frequency drift patterns, the classifier has learned the characteristics of normal frequency drift patterns and several abnormal frequency drift patterns. When a new feature vector is input, the classifier compares it with the learned patterns to determine whether it belongs to the normal or abnormal category.

[0099] 2. Identification of abnormal frequency drift and correlation with timing error

[0100] If the classifier output shows that the category corresponding to a feature vector is an abnormal category, it means that the clock frequency is abnormal. This abnormal frequency drift may cause timing errors.

[0101] The content described in this implementation further improves the clock frequency drift monitoring and perception timing protection system through operations such as feature vector normalization and classifier update, improves the system's ability to identify clock frequency drift and perceive timing errors, thereby better ensuring the normal operation of the timing system.

[0102] Implementation method 4, Figure 4 This is a schematic diagram of the abnormal frequency drift warning process, which includes:

[0103] 1. Anomaly Identification and Warning Information Generation

[0104] 1. Abnormal frequency drift identification

[0105] When the classifier identifies an abnormal frequency drift, this is based on the processing of the feature vector in implementation schemes 1, 2, and 3 and the training and updating of the classifier.

[0106] 2. Generation of early warning information

[0107] Generate warning information, which includes abnormal frequency value, drift degree and possible timing error estimation. The abnormal frequency value comes directly from the classifier output, which reflects the degree to which the current clock frequency deviates from the normal one.

[0108] The determination of the drift degree can be based on the analysis of relevant parameters in the eigenvector, such as variance. If the variance is large, it means that the frequency fluctuates violently and the drift degree is high. For the estimation of the timing error that may be caused, it is necessary to combine the specific characteristics and requirements of the timing system.

[0109] 2. Early Warning Information Transmission

[0110] 1. Wireless communication module

[0111] The warning information is sent to the monitoring terminal through the wireless communication module. The wireless communication module adopts a specific modulation coding method. For binary phase shift keying (BPSK) modulation, the formula is: .here To send a signal, is the signal amplitude, is the carrier angular frequency, is a binary information sequence.

[0112] 2. Monitoring terminal reception and processing

[0113] After receiving the warning information, the monitoring terminal will process it and take corresponding measures according to the content of the warning information.

[0114] 3. Implementation of protective measures

[0115] 1. Micro-level protection

[0116] At the micro level, when abnormal frequency drift is detected, appropriate measures can be taken according to the specific situation, such as adjusting the frequency parameters of the local clock.

[0117] 2. Macro-level protection

[0118] At the macro level, there are situations where multiple clock sources are involved.

[0119] First, the monitoring data of multiple clock sources are fused. The fusion formula is: ,in is the fused monitoring data, For the The weight of the clock source, For the Monitoring data of clock sources, The number of clock sources. By integrating data, the situation of each clock source is comprehensively considered to improve the accuracy of timing protection.

[0120] Then, according to the integrated monitoring data, adjust the weight of each clock source ,in is the adjusted weight, It is the weight adjustment amount, which is determined according to the deviation between the fused data and the historical data to optimize the performance of the entire timing protection system.

[0121] The method described in this embodiment further improves the timing protection system, covering both the micro and macro levels, from anomaly identification to warning information generation and transmission, and then to the implementation of protection measures, thereby improving the timing system's ability to cope with clock frequency drift and ensuring the accuracy and reliability of timing.

[0122] Implementation method 5, Figure 5 This is a schematic diagram of the multi-clock source data fusion and weight adjustment process, which includes:

[0123] 1. Multi-clock source data fusion basis

[0124] 1. Includes preliminary method steps

[0125] This means that a series of processes from clock frequency signal acquisition, preprocessing, feature analysis, classification and anomaly identification to early warning information generation and transmission and preliminary protective measures are carried out on this basis. 2. The necessity of multi-clock source monitoring data fusion

[0126] The purpose of fusing the monitoring data of multiple clock sources is to comprehensively consider the situation of each clock source and improve the accuracy of timing protection. In practical applications, different clock sources may have different accuracy, stability and degree of environmental influence.

[0127] 1. Fusion formula and weight determination

[0128] The fusion formula is ,in is the fused monitoring data, For the The weight of the clock source, For the Monitoring data of clock sources, The number of clock sources. The determination of is based on the performance of the clock source, historical data, etc. For example, a stable and highly accurate atomic clock source may be given a higher weight, while a satellite navigation system clock source that is susceptible to interference may be given a relatively low weight.

[0129] 2. Data Fusion Example

[0130] Assume that there are three clock sources in a timing system, namely a high-precision atomic clock (referred to as clock source 1), a satellite navigation system (referred to as clock source 2), and an ordinary crystal oscillator (referred to as clock source 3). Monitoring data of atomic clocks The frequency is very stable, and the monitoring data of the satellite navigation system There is a certain fluctuation, the monitoring data of the crystal oscillator Atomic clocks are weighted based on their performance and historical data. , satellite navigation system weight , crystal oscillator weight The fused monitoring data .

[0131] 3. Weight adjustment based on fusion data

[0132] 1. Basis for weight adjustment

[0133] Adjust the weight of each clock source based on the integrated monitoring data ,in is the adjusted weight, is the weight adjustment amount, which is determined according to the deviation between the fused data and the historical data. If the fused data deviates greatly from the historical data, it means that a clock source may have a problem and its weight needs to be adjusted. For example, if the monitoring data of the satellite navigation system deviates greatly from the historical data after fusion, it may be due to factors such as weather that cause its performance to deteriorate, and its weight needs to be appropriately reduced.

[0134] 2. Weight adjustment example

[0135] Continuing with the above example, if after a period of time, it is found that the fusion data deviates greatly from the historical data, it is mainly due to the deviation of the monitoring data of the satellite navigation system. Assuming the weight adjustment amount , then the adjusted satellite navigation system weight , and the weight of other stable clock sources can be appropriately increased. For example, the weight of the atomic clock can be increased to , in order to optimize the performance of the entire timing protection system.

[0136] 4. Impact on the Timing System

[0137] 1. Improve timing accuracy

[0138] The accuracy of timing protection can be improved by fusion of multiple clock source data and weight adjustment. In the above example, by adjusting the weight, the role of stable clock source is more prominent, and the impact of unstable clock source on timing is reduced, thereby improving the accuracy of timing. For example, in a communication system with extremely high requirements for timing accuracy, this method can ensure the synchronization and accuracy of signal transmission and improve communication quality.

[0139] 2. Enhance system stability

[0140] At the same time, this method also enhances the stability of the timing system. Even if a clock source has a problem, the timing system can continue to operate normally by integrating the data of other clock sources and adjusting the weights. For example, in a distributed timing system, if a satellite navigation system as a clock source fails, by integrating the data of other clock sources and adjusting the weights, the timing system can still provide relatively accurate timing services and avoid system paralysis.

[0141] The method described in this embodiment further optimizes the timing protection system through multi-clock source data fusion and weight adjustment, improves the accuracy and stability of timing, and enables it to better adapt to complex application environments and meet different timing requirements.

[0142] Implementation method 6, Figure 6 This is a schematic diagram of the time alignment and correlation analysis process, which includes:

[0143] 1. The Importance of Data Preprocessing

[0144] 1. Prerequisites for multi-clock source data fusion

[0145] Before fusing the monitoring data of multiple clock sources, it is very important to perform time alignment and correlation analysis on the data of different clock sources. This is because different clock sources may have inconsistent data in time due to their own characteristics, working environment, data collection method, etc., and the correlation between different clock sources may also be different. If these preprocessing are not performed, direct data fusion may lead to inaccurate fusion results, which in turn affects the performance of the timing protection system.

[0146] 2. Time Alignment Process

[0147] 1. Time synchronization algorithm based on least squares method

[0148] The least squares-based time synchronization algorithm is used to align the data from different clock sources. The time consistency of the data from each clock source is ensured by minimizing the error function.

[0149] 2. Time Alignment Example

[0150] Consider an actual timing system, which includes a satellite navigation system clock source and a ground atomic clock clock source. Due to factors such as signal transmission delay, the sampling point time of the satellite navigation system deviates from the sampling point time of the ground atomic clock. The aligned time is calculated by the time synchronization algorithm based on the least squares method. Assume that the sampling point time of the satellite navigation system is originally , the sampling point time corresponding to the ground atomic clock is , after the time alignment algorithm, find the aligned time , which reduces the time deviation of the two clock sources at the sampling point and improves the consistency of the data.

[0151] 3. Correlation Analysis

[0152] 1. Correlation coefficient calculation

[0153] Perform correlation analysis on the clock source data after time alignment and calculate the correlation coefficient between different clock sources. The calculation formula for the correlation coefficient is: ,in) and are the kth sampling point values ​​of the ith and jth clock sources respectively, and is the corresponding mean.

[0154] 2. Correlation Analysis Example

[0155] Continuing with the satellite navigation system and ground atomic clock as an example, suppose that the correlation analysis is performed on the multiple sampling point data after their time alignment. The sampling point value of the satellite navigation system is , the sampling point value of the ground atomic clock is By calculating the correlation coefficient, if the obtained correlation coefficient is close to 1, it means that there is a strong correlation between the two clock sources, that is, their frequency change trends are similar; if the correlation coefficient is close to 0, it means that the correlation between the two clock sources is weak, that is, their frequency change trends are different. For example, if the calculated correlation coefficient is 0.8, it means that there is a strong correlation between the satellite navigation system and the ground atomic clock, and the corresponding weight adjustment strategy can be considered in the timing protection system.

[0156] 4. Weight Adjustment Strategy

[0157] 1. Adjust the initial weight value according to the relevance

[0158] Adjust weights based on relevance If the correlation between the two clock sources is strong, a higher initial weight value may be given; if the correlation is weak, a lower initial weight value may be given.

[0159] 2. Comprehensive consideration to adjust weights

[0160] The adjustment of weights is not only based on relevance, but also needs to take into account other factors, such as the performance of the clock source, historical data, etc.

[0161] The method described in this embodiment further optimizes the data fusion process of the timing protection system by performing time alignment and correlation analysis on data from different clock sources, as well as corresponding weight adjustment strategies, thereby improving the accuracy and stability of the timing protection system and enabling it to better adapt to complex timing environments.

[0162] Implementation method 7, Figure 7 Schematic diagram of weight adjustment and feedback optimization process, including:

[0163] 1. Fusion data update based on weight adjustment

[0164] 1. Recalculate the monitoring data after fusion

[0165] According to the adjusted weight , recalculate the fused monitoring data. The fusion formula is For example, suppose there are three clock sources, and their initial weights are , , The corresponding monitoring data is , , , then the fused monitoring data Later, for some reason, the weight was adjusted to , , , then recalculate the fused monitoring data .

[0166] 2. The significance of fusion data update

[0167] The purpose of recalculating the fusion data is to better reflect the combined impact of each clock source on timing under the current weight. As the clock source operates and the environment changes, its performance may change, resulting in the need to adjust the weight. By updating the fusion data, the timing protection system can more accurately utilize the information of each clock source and improve the accuracy and stability of timing.

[0168] 2. Feedback Optimization Process

[0169] 1. The role of feedback to the clock source

[0170] The recalculated fusion data is fed back to each clock source to update the classifier parameters and adjust the warning threshold. This is an important feedback mechanism that enables the entire timing protection system to continuously optimize its performance. For example, for a clock source, if the fusion data shows that its contribution to timing has changed, then its internal classifier parameters can be adjusted accordingly to enable it to better identify the frequency drift of the clock source. At the same time, the warning threshold can also be adjusted to make the warning more accurate and timely.

[0171] 2. Feedback Optimization Example

[0172] Consider a timing system with two clock sources, one is the satellite navigation system clock source (denoted as clock source 1), and the other is the ground atomic clock clock source (denoted as clock source 2). Initially, after the fusion data is calculated, the weights of the two clock sources are , As time goes by, the performance of satellite navigation systems degrades due to weather conditions, and the weights are adjusted to , . After the fusion data is recalculated, it is fed back to the two clock sources. For the satellite navigation system clock source, its internal classifier parameters may be adjusted according to the feedback data, such as increasing the sensitivity to frequency drift, because its importance in timing is relatively reduced. At the same time, the warning threshold may be adjusted accordingly so that an early warning is issued when a small frequency drift occurs to better monitor its performance changes. For the ground-based atomic clock source, its classifier parameters and warning threshold may also be adjusted according to the feedback data to better adapt to its new role in timing, that is, to assume more important timing tasks.

[0173] 3. Overall Impact on the Timing Protection System

[0174] 1. Improve timing accuracy

[0175] By recalculating the fused data and feeding it back to the clock source for optimization, the accuracy of the timing protection system is improved. In the above example, when the performance of the satellite navigation system deteriorates, by adjusting the weights and feedback optimization, the timing system can better utilize the stable performance of the ground atomic clock, reduce the impact of the instability of the satellite navigation system, and thus ensure the accuracy of the timing.

[0176] 2. Enhance system stability

[0177] At the same time, this feedback optimization mechanism also enhances the stability of the timing system. Even if a clock source has a problem, the timing system can continue to operate normally through continuous adjustment of weights and feedback optimization.

[0178] The method described in this embodiment further improves the timing protection system by recalculating fusion data and feedback optimization, improves the accuracy and stability of timing, and enables it to better adapt to complex timing environments and meet different timing requirements.

[0179] Implementation method 8, Figure 8 Schematic diagram of monitoring components, including:

[0180] 1. Component composition and function overview

[0181] 1. Clock frequency sensor

[0182] The clock frequency sensor is the basic component of the whole device, which is used to collect the clock frequency signal.

[0183] 2. Signal processing unit

[0184] The signal processing unit complies with the signal processing unit related standards and is responsible for performing operations such as denoising, resampling, and feature normalization.

[0185] 3. Machine Learning Classifier Processing Unit

[0186] The machine learning classifier processing unit complies with the machine learning classifier related standards and is used to run the classifier algorithm. It is trained with a large amount of labeled historical clock frequency drift data.

[0187] 4. Wireless communication module

[0188] The wireless communication module complies with the wireless communication module and is used to send early warning signals.

[0189] The components and their collaborative working mechanism described in this embodiment form the basis of an effective clock frequency drift monitoring, perception and timing protection system, and provide important support for ensuring the accuracy and reliability of timing.

[0190] Implementation method 9, Fig. 9 This is a schematic diagram of the timing protection device, which includes:

[0191] 1. Overview of device composition and functions

[0192] 1. Component Integration

[0193] The device includes the components, namely, a clock frequency sensor, a signal processing unit, a machine learning classifier processing unit and a wireless communication module. These components work together to form the core functional module of the device.

[0194] 2. Power module

[0195] It also includes a power module that complies with relevant standards for power modules, which is used to supply power to various components. The power module needs to provide stable power to ensure the normal operation of the device.

[0196] 3. Shell

[0197] As well as the housing that complies with the relevant standards for housings, the housing protects the internal components. It can prevent external factors such as dust, moisture and mechanical shock from affecting the internal components.

[0198] The device described in this embodiment demonstrates its important role in ensuring the accuracy and reliability of timing through its reasonable component composition, collaborative working mechanism, and application examples in different systems.

[0199] Embodiment 10, Fig.10 This is a schematic diagram of a distributed timing protection system, which includes:

[0200] 1. System composition and architecture

[0201] 1. Device integration

[0202] The system includes devices. These devices are connected via a wired or wireless network to achieve distributed timing protection. Each device has its own set of components, including a clock frequency sensor, a signal processing unit, a machine learning classifier processing unit, a wireless communication module, a power module and a housing.

[0203] 2. Central monitoring server

[0204] It also includes a central monitoring server that complies with the relevant standards of the central monitoring server. The central monitoring server is used to receive and process the warning information and monitoring data sent by each device, and to uniformly manage and control the entire timing protection system.

Claims

1. A clock frequency drift algorithm monitoring perception timing protection method, characterized in that: The following steps are involved: Step S001: Obtain a clock frequency signal through a sensor and perform discrete sampling on it to obtain a discrete sequence ; Step S002: Use the mean drift estimation in the clock frequency drift algorithm to calculate the mean of the clock frequency signal within a certain time window. ,in is the number of sampling points in the time window, which is used to preliminarily evaluate the central trend of the clock frequency; Step S003: Calculate the variance of the clock frequency signal , determine the degree of fluctuation of the clock frequency according to the size of the variance; Step S004: Constructing frequency drift feature vector , which will serve as the basis for subsequent analysis; Step S005: Transform the feature vector The input is sent to a machine learning-based classifier, which is trained using labeled historical clock frequency drift data. The loss function used in the training process is cross entropy loss. ,in is the number of categories, is the true label, In order to predict the probability, the abnormal frequency drift is identified through the output of the classifier, so as to realize the perception of the possible errors in timing.

2. The method according to claim 1, characterized in that The following steps are involved: Step S001: Before calculating the mean and variance, the collected clock frequency signal is denoised by using a wavelet transform denoising method, by selecting an appropriate wavelet basis function and the number of decomposition levels , for the signal Perform wavelet decomposition ,in is the wavelet coefficient, The high-frequency noise components in the signal are removed by stretching and translating the wavelet basis function; Step S002: resample the denoised signal so that the sampling frequency of the signal meets the requirements of subsequent analysis. The resampling formula is: ,in is the resampled signal, is the new sampling period.

3. The method according to claim 2, characterized in that The following steps are involved: Step S001: After constructing the feature vector, the feature vector is normalized using the minimum-maximum normalization method. The formula is: ,in is the normalized eigenvalue, is the original eigenvalue, and are the minimum and maximum values ​​of the feature values, respectively, so that features of different dimensions have the same scale and improve the performance of the classifier; Step S002: Regularly update the classifier parameters, using the stochastic gradient descent algorithm based on the newly acquired clock frequency drift data ,in are the classifier parameters, is the learning rate, The gradient of the loss function with respect to the parameters is used to fine-tune the classifier to adapt to the changes in the clock frequency drift characteristics.

4. The method according to claim 3, characterized in that The following steps are involved: Step S001: When the classifier identifies abnormal frequency drift, it generates warning information, which includes the abnormal frequency value, the drift degree and the possible timing error estimation; Step S002: Send the warning information to the monitoring terminal through the wireless communication module. The wireless communication module adopts a specific modulation coding method. For binary phase shift keying BPSK modulation, the formula is: ,in To send a signal, is the signal amplitude, is the carrier angular frequency, It is a binary information sequence for signal transmission.

5. The method according to claim 4, characterized in that The following steps are involved: Step S001: comprising the method steps described in any one of claims 1 to 4; Step S002: Fusion the monitoring data of multiple clock sources. The fusion formula is: ,in is the fused monitoring data, For the The weight of the clock source, For the Monitoring data of clock sources, The number of clock sources is considered comprehensively to improve the accuracy of timing protection; Step S003: Adjust the weight of each clock source according to the integrated monitoring data , the adjustment formula is ,in is the adjusted weight, It is the weight adjustment amount, which is determined according to the deviation between the fused data and the historical data to optimize the performance of the entire timing protection system.

6. The method according to claim 5, characterized in that The following steps are involved: Step S001: Before fusing the monitoring data of multiple clock sources, the data of different clock sources are time-aligned, and a time synchronization algorithm based on the least squares method is used to minimize the error function. ,in For the The clock source Sampling point time, To ensure the consistency of the data of each clock source in time after alignment; Step S002: Perform correlation analysis on the time-aligned clock source data and calculate the correlation coefficients between different clock sources. ,in and Respectively and The clock source The sampling point value, and is the corresponding mean, and the weight is adjusted according to the size of the correlation The initial value of .

7. The method according to claim 5, characterized in that The following steps are involved: Step S001: According to the adjusted weight , recalculate the fused monitoring data ; Step S002: Feedback the recalculated fusion data to each clock source to update the classifier parameters and adjust the warning threshold to further optimize the timing protection effect.

8. A component using the method according to any one of claims 1 to 7, characterized in that: It includes a clock frequency sensor, a signal processing unit, a machine learning classifier processing unit, and a wireless communication module, wherein the clock frequency sensor is used to collect clock frequency signals, the signal processing unit is used to perform denoising, resampling and feature normalization, the machine learning classifier processing unit is used to run the classifier algorithm, and the wireless communication module is used to send early warning signals.

9. A device using the method according to any one of claims 1 to 8, characterized in that: The device comprises the component as claimed in claim 8, and further comprises a power module and a casing, wherein the power module supplies power to each component and the casing protects the internal components.

10. A system using the method according to any one of claims 1 to 9, characterized in that: The invention comprises a plurality of devices as described in claim 9, wherein each device is connected to another via a wired or wireless network to realize distributed timing protection, and further comprises a central monitoring server for receiving and processing early warning information and monitoring data sent by each device to uniformly manage and regulate the entire timing protection system.

Images