A method for controlling smart locks of power facilities based on multi-dimensional authority data processing

Abstract

The present application relates to a method for controlling smart locks of power facilities based on multi-dimensional authority data processing, comprising: obtaining a curve showing the change of magnetic induction signal strength and waveform over time when the smart lock is in operation, performing preliminary filtering on the data by analyzing the waveform distortion and intensity attenuation characteristics, and obtaining a change trend sequence; extracting the waveform distortion increment and signal strength decrease rate caused by wear of the smart lock body through the change trend sequence, and determining a grading index of the degree of wear of the smart lock body; combining the lock core gap change with the historical door body matching status data, calculating the dynamic mapping relationship between the magnetic induction signal strength and the gap change, and dynamically adjusting the magnetic induction threshold range; judging in real time whether the magnetic induction signal strength meets the target locking judgment standard, dynamically updating the lock control parameters when the lock body is in operation, and obtaining optimized lock control status data.

Description

Technical Field

[0001] The present application relates to the field of electrical digital data technology, and in particular to a method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing. Background Art

[0002] Smart lock management and control in power facilities is a critical area for ensuring the safe and efficient operation of energy infrastructure. Its importance is reflected in its direct impact on equipment safety, operational stability, and maintenance efficiency. As power facilities become increasingly intelligent, smart locks, as core components, must not only ensure high security but also adapt to the complex environments of long-term operation. Currently, smart lock management and control technologies based on multi-dimensional permission data are widely used, but existing methods have significant limitations in addressing the performance degradation of lock bodies over long-term use. Most solutions rely on static thresholds or fixed decision logic, which struggles to adapt to the dynamic changes caused by mechanical wear, leading to misjudgments and potential safety hazards. This is particularly true when the clearance between the lock cylinder and the door increases, as traditional methods are unable to effectively capture the subtle signs of early performance degradation. A key challenge in this area lies in accurately sensing and predicting the performance changes caused by lock wear. Increased clearance between the lock cylinder and the door increases the distortion of the magnetic induction signal waveform, blurring the determination of the closed state and altering the pattern of key removal resistance. These technical factors directly impact the reliability and security of locks. Failure to dynamically adjust relevant parameters in a timely manner can lead to lock failure and increased maintenance costs.

[0003] Existing technologies often lack adaptive prediction mechanisms when dealing with these dynamic changes, making it difficult to identify wear characteristics early and take effective intervention. Therefore, how to build a predictive maintenance algorithm that can dynamically adjust magnetic induction threshold parameters and lock determination tolerance based on changes in magnetic induction signal strength has become a key issue in extending equipment service life while ensuring safety. Summary of the Invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the purpose of this application is to provide a method for controlling smart locks of power facilities based on multi-dimensional authority data processing.

[0005] The present application discloses a method for controlling smart locks for electric power facilities based on multi-dimensional authority data processing, comprising:

[0006] Obtain the magnetic induction signal sequence when the smart lock is running and then denoise it to generate a denoised signal sequence;

[0007] Calculate and extract the attenuation parameters of the hyperbolic attenuation model parameters of the signal amplitude in the original magnetic induction signal sequence for the denoised signal sequence, and fit the change trend sequence;

[0008] Taking the trend sequence as input, the difference between the current waveform distortion and the historical baseline value is calculated as the distortion increment and decomposed. A support vector regression model is constructed to output a quantitative wear score of the lock body wear degree, and the wear index is divided according to the score.

[0009] After the wear index is divided, a gap change curve is generated based on the judgment result of whether the predicted value of the lock core gap exceeds the preset threshold;

[0010] The predicted value of the lock core gap is used as a parameter to establish a door body matching status database; a gap-magnetic induction intensity mapping model is constructed, the signal intensity compensation coefficient is output and segmented optimization is performed, and the interval boundaries of the segmented optimization are dynamically adjusted;

[0011] When the high-zone threshold value is continuously exceeded during segmented optimization, the tolerance increment of the gap between the lock cylinder and the door body for the next 7 days is output; the magnetic induction fluctuation compensation amount is calculated based on the tolerance increment; and a target judgment standard curve is generated;

[0012] Based on the target judgment standard of the target judgment standard curve, the magnetic induction signal strength level is divided into levels, and the torque output of the smart lock motor driving the lock cylinder or lock tongue is dynamically adjusted;

[0013] Based on the lock control status data of the torque output, a deep neural network misjudgment probability model is constructed to output a misjudgment score. The weight factor of the wear parameter is dynamically adjusted according to the risk level until the torque fluctuation of the smart lock motor driving the lock cylinder or lock tongue meets the preset parameters.

[0014] The advantage of the method for controlling smart locks of power facilities based on multi-dimensional authority data processing described in the present application is that, by acquiring the magnetic signal strength and waveform change curve, analyzing the waveform distortion and intensity attenuation characteristics, extracting the lock body wear index, establishing a wear prediction model, combining historical door body matching status data, calculating the dynamic mapping relationship between the magnetic signal and the gap change, adjusting the magnetic threshold range, identifying the lock core tolerance range according to the lock body wear degree and the lock core gap change prediction value, obtaining the target locking judgment standard, judging in real time whether the magnetic signal meets the judgment standard, dynamically updating the lock control parameters, evaluating the probability of misjudgment, judging the safety hazard risk, adjusting the threshold weight factor when necessary, and re-optimizing the control strategy, the present invention can effectively predict the wear status of the smart lock body, dynamically optimize the lock control parameters, improve the locking accuracy and safety, and extend the service life of the smart lock. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is the process of a method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing described in this application Figure 1 ;

[0016] Figure 2This is the process of a method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing described in this application Figure 2 . DETAILED DESCRIPTION

[0017] like Figure 1-Figure 2 As shown, the method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing described in this application includes the following steps:

[0018] S101, obtaining a curve showing the change in magnetic induction signal strength and waveform over time when the smart lock is in operation, preprocessing the signal data, and obtaining a change trend sequence;

[0019] S102. Extracting changes in magnetic induction signal characteristics caused by wear of the smart lock body based on the change trend sequence, quantifying waveform distortion increment and signal strength attenuation rate, and determining a grading index for the degree of wear of the lock body;

[0020] S103. Build a prediction model using historical smart lock body wear data, combine wear degree classification indicators and change trend sequences, predict the wear state of the smart lock body within a preset time, and generate a lock core gap change curve through fitting analysis;

[0021] S104. Combining the lock core gap change and historical door-body coordination status data, a dynamic mapping relationship between magnetic induction signal strength and gap parameters is constructed, and the magnetic induction threshold range is dynamically optimized to improve the determination accuracy of the smart lock;

[0022] S105. Based on the lock body wear degree classification index and the lock core gap prediction value, when the adjusted magnetic induction threshold exceeds the preset upper limit, identify the lock core tolerance range and generate a target locking judgment standard to ensure reliable operation of the smart lock;

[0023] S106, real-time monitoring of the magnetic signal strength to see if it meets the target locking criteria, and dynamic adjustment of the lock control parameters to optimize the operation status of the smart lock;

[0024] S107. Combine the optimized lock control status data and change trend sequence to evaluate the probability of misjudgment of the smart lock operation status and determine whether the safety risk is lower than the preset threshold. If the risk exceeds the threshold, adjust the weight factor of the magnetic induction threshold and the judgment tolerance, and re-optimize the lock control parameters to ensure the safe operation of the smart lock.

[0025] like Figure 1-Figure 2 As shown, in step S101, a curve showing the change in magnetic induction signal strength and waveform over time when the smart lock is running is obtained, and the signal data is preprocessed to obtain a change trend sequence.

[0026] Furthermore, in step S101, in an embodiment of the present application, when the control function of the smart lock is triggered, the system first collects magnetic induction signal data during operation through a magnetic induction sensor, and pre-processes the data to obtain a change trend sequence; wherein the above-mentioned system may be a smart lock control system, which may be controlled by a computer or an MCU chip, and is connected to the smart lock signal, and the signal connection may be a direct connection or an indirect connection;

[0027] Collect the original magnetic induction signal data sequence from the magnetic induction sensor of the smart lock. The original magnetic induction signal data sequence is the magnetic induction signal data sequence collected from the magnetic induction sensor and not processed, and record the signal amplitude, sampling time and frequency parameters;

[0028] In the embodiment of the present application, the magnetic induction sensor continuously collects the raw magnetic induction signal data sequence during the operation of the smart lock at a sampling interval of 10 milliseconds, and records the signal amplitude, sampling time and frequency parameters of each data point. For example, during the closing process of the lock, the signal amplitude may gradually decrease from 2.8 millitesla to 0.6 millitesla, which lasts for about 300 milliseconds. The embodiment of the present application does not impose too many restrictions on the specific sampling parameters, which can be adjusted according to the actual scenario;

[0029] The original magnetic induction signal data sequence is filtered using a digital filter to generate a denoised signal sequence;

[0030] To remove noise interference, the embodiment of the present application uses a Butterworth digital filter to perform third-order high-pass filtering and second-order low-pass filtering on the original signal to filter out high-frequency noise above 500 Hz and low-frequency interference below 5 Hz, thereby generating a filtered signal sequence;

[0031] High-pass filtering can remove baseline offsets caused by geomagnetic field fluctuations or temperature drift, and low-pass filtering can filter out sensor thermal noise or external electromagnetic interference;

[0032] Perform spectrum analysis on the denoised signal sequence to extract waveform distortion and signal attenuation characteristics;

[0033] In the embodiment of the present application, the denoised signal sequence is subjected to spectrum analysis through a 512-point discrete Fourier transform, the amplitude ratio of the fundamental component to the harmonic component is calculated, the waveform distortion index is obtained, and the attenuation curve of the signal amplitude over time is recorded. Under normal circumstances, the fundamental frequency of the smart lock is about 50 Hz, the second harmonic component does not exceed 15% of the fundamental wave, and the third harmonic does not exceed 8%;

[0034] A signal strength attenuation model is established based on the waveform distortion and attenuation curve, and a change trend curve is fitted;

[0035] Based on the waveform distortion and attenuation curve, the embodiment of the present application sets the signal amplitude determination threshold interval to ±20% of the nominal value, for example, 2.0 to 3.0 millitesla. A hyperbola fitting method is used to establish a signal strength attenuation model, including an initial strength coefficient of 2.5 millitesla and a decay rate coefficient of 0.008 / millisecond. Subsequently, the signal sequence is segmented into 100-millisecond time windows, the signal-to-noise ratio of each segment is calculated, and signal segments with a signal-to-noise ratio greater than 15 decibels are selected. Peak and valley data points are extracted (peak value of approximately 2.8 millitesla, valley value of approximately 2.2 millitesla), and a trend curve including timestamps, magnetic induction intensity, and waveform distortion values is fitted using a cubic spline interpolation method.

[0036] In the embodiment of the present application, the trend curve generated by the above steps can accurately characterize the magnetic field characteristics of the smart lock under different working conditions, providing a reliable data basis for wear prediction and parameter optimization. The lock control parameters can be dynamically adjusted according to the trend curve to ensure locking accuracy and security, significantly improving the reliability and service life of the smart lock.

[0037] In the embodiment of the present application, the design of the lock determines the lock state (unlocked or locked) based on the in and out movement of the lock tongue. The magnetic induction signal is collected by the magnetic induction sensor (Hall sensor) built into the smart lock. Its principle is based on the magnetic field change caused by the mechanical movement of the lock cylinder. When the lock cylinder rotates or the lock tongue moves, the relative position between the internal magnet and the sensor changes, causing the magnetic induction intensity (unit: millitesla) and waveform output by the sensor to change accordingly.

[0038] The deadbolt is the part of the smart lock that mates directly with the latch on the door frame and is responsible for the physical locking and unlocking action. When the lock is closed, the deadbolt inserts into the latch on the door frame; when the lock is open, the deadbolt retracts into the lock body.

[0039] A permanent magnet is mounted on the lock tongue or a component that interacts with it. As the lock tongue moves in and out, the position of the permanent magnet relative to the magnetic sensor changes. A magnetic sensor (Hall sensor) is fixed within the lock body, located at a position capable of detecting changes in the magnetic field caused by the permanent magnet. The magnetic sensor is mounted within the lock body at a position that corresponds to the position where it can detect changes in the magnetic field when the lock tongue is fully extended or retracted. For example, when the lock tongue is fully extended, the magnetic sensor detects the strongest magnetic field strength; when the lock tongue is fully retracted, it detects the weakest magnetic field strength. The sensor determines the lock's status (for example, whether the lock is fully closed) by monitoring changes in magnetic field strength. In one approach, the sensor is mounted within the lock body at a fixed point aligned with the lock tongue's path of motion, allowing it to dynamically monitor changes in magnetic field strength as the lock tongue moves in and out.

[0040] The lock cylinder is directly or indirectly connected to the deadbolt through internal mechanical structures (e.g., gears, linkages, etc.). This means that when the lock cylinder is rotated, it causes the deadbolt to undergo a corresponding linear movement (extend or retract). For example, when the correct key is inserted and turned, the mechanism within the lock cylinder triggers the deadbolt to retract from the striker on the door frame, thereby opening the door.

[0041] The state of the deadbolt (extended or retracted) determines whether the door is locked. The lock cylinder provides a method to control the position of the deadbolt so that it can only be changed under certain conditions (such as correctly entering a password, using the correct key, or passing electronic authentication) to open or close the door.

[0042] The movement of the lock tongue can be used as part of the feedback signal. For example, a magnetic induction sensor can detect the position change of a permanent magnet connected to the lock tongue to determine whether the lock tongue is fully retracted or extended, thereby confirming the lock status.

[0043] When locking, the lock tongue moves outward and inserts into the lock catch on the door frame. During this process, the permanent magnet attached to the lock tongue gradually approaches the magnetic induction sensor, causing the gap between them to decrease. As the distance decreases, the magnetic field strength detected by the magnetic induction sensor increases.

[0044] Conversely, when unlocking, the lock tongue retracts inward, away from the lock catch on the door frame, and at the same time drives the permanent magnet away from the magnetic induction sensor, causing the gap between the two to increase. At this time, the magnetic field strength detected by the magnetic induction sensor decreases.

[0045] In addition, long-term mechanical wear can cause the lock cylinder gap to expand, increasing signal waveform distortion (the proportion of harmonic components) and causing the signal strength to exhibit nonlinear attenuation characteristics. The lock cylinder gap refers to the distance between the lock cylinder and the door body.

[0046] In the embodiment of the present application, the input source of the magnetic induction signal is: the original magnetic induction signal is collected in real time by the magnetic induction sensor at a sampling interval of 10 milliseconds. The original signal sequence includes magnetic induction intensity amplitude, timestamp, and frequency parameters. For example, during the closing phase, the signal amplitude drops from 2.8 millitesla to 0.6 millitesla for about 300 milliseconds.

[0047] The original magnetic induction signal is denoised to obtain the denoised signal sequence as follows:

[0048] Filtering and denoising uses a Butterworth digital filter to perform third-order high-pass filtering (to filter out low-frequency drift <5Hz and temperature drift) and second-order low-pass filtering (to filter out high-frequency noise >500Hz and electromagnetic interference) to generate a denoised signal sequence.

[0049] Alternatively, spectrum analysis can be performed on the denoised signal sequence using a 512-point discrete Fourier transform (DFT), calculating the amplitude ratio of the fundamental and harmonic components and extracting the waveform distortion (the third harmonic should not exceed 8% of the fundamental).

[0050] Attenuation modeling was based on the signal amplitude attenuation curve. A hyperbola fitting method was used to establish a signal intensity attenuation model. The parameters included the initial intensity coefficient (2.5 mTesla) and the attenuation rate coefficient (0.008 / ms).

[0051] Trend fitting was performed by calculating the signal-to-noise ratio (SNR>15dB) in 100-millisecond time windows. Peak and valley data points (peak value 2.8mT, valley value 2.2mT) were extracted, and a trend curve was generated using cubic spline interpolation, including timestamps, magnetic induction intensity, and waveform distortion values.

[0052] Output data: A set of structured data that represents the time evolution of the magnetic induction signal, which is used for subsequent wear analysis and parameter optimization.

[0053] like Figure 1-Figure 2 As shown, in step S102, the change in magnetic induction signal characteristics caused by the wear of the smart lock body is extracted based on the change trend sequence, the waveform distortion increment and the signal strength attenuation rate are quantified, and the grading index of the degree of lock body wear is determined.

[0054] Furthermore, in step S102, in the embodiment of the present application, by analyzing the change trend sequence, the magnetic induction signal characteristics reflecting the wear state of the smart lock body are extracted, the waveform distortion increment and the signal strength attenuation rate are quantified, and a grading index of the degree of lock body wear is generated based on multi-dimensional feature analysis to provide a basis for parameter optimization;

[0055] Obtain the signal sequence of the lock body during operation from the magnetic induction sensor and calculate the waveform distortion increment parameter;

[0056] In an embodiment of the present application, the magnetic induction sensor of the smart lock collects a continuous magnetic induction signal sequence during the operation of the lock body at a sampling frequency of 100 Hz, records the signal amplitude, sampling time and frequency information, and the system uses the waveform distortion average of the historical operation data as a benchmark, wherein the amplitude ratio of the fundamental wave to the third harmonic is calculated as the waveform distortion. For example, the benchmark value is set to 5%, and the waveform distortion increment parameter is calculated by comparing the difference between the waveform distortion of the current signal sequence and the benchmark value. For example, in a state of slight wear, the increment usually does not exceed 10%; in moderate wear, the increment is between 15% and 25%; in severe wear, the increment may exceed 30%. The embodiment of the present application does not impose too many restrictions on the specific setting of the benchmark value, and it can be adjusted according to the lock body type and usage scenario;

[0057] Combined with the lock body rotation angle data, the signal strength change curve is fitted and the signal strength attenuation rate is calculated;

[0058] In the embodiment of the present application, a lock body angle sensor is used to record the angle data of the lock cylinder during rotation. The sampling frequency is 100 Hz and the rotation angle range is 0 to 180 degrees. The amplitude data of the magnetic induction signal at different angles is extracted. A continuous curve of signal strength versus angle is fitted using the cubic spline interpolation method. The system calculates the slope of the curve at each angle point to obtain the signal strength attenuation rate parameter. In the new lock body state, the signal strength changes smoothly, and the absolute value of the slope usually does not exceed 0.02 millitesla / degree. With increased wear, the absolute value of the slope may increase to 0.05 millitesla / degree or more, reflecting the instability of the lock body operation.

[0059] The frequency domain features of waveform distortion increment and signal strength attenuation rate are extracted through wavelet decomposition;

[0060] To further analyze signal characteristics, the embodiment of the present application performs four-layer wavelet decomposition on the waveform distortion increment parameter and the signal strength attenuation rate parameter, respectively. Using the db4 wavelet basis function, the signal is decomposed into five frequency bands: 0-2 Hz, 2-4 Hz, 4-8 Hz, 8-16 Hz, and 16-32 Hz. The system calculates the energy distribution characteristics of each frequency band and generates a feature set containing low-frequency and high-frequency coefficients. In the new lock body state, the energy proportion of the low-frequency band (0-4 Hz) typically exceeds 80%, reflecting a stable magnetic field distribution. As wear increases, the energy proportion of the high-frequency band (8-32 Hz) increases, indicating more unstable vibrations during operation.

[0061] Construct a multidimensional feature vector and quantify the degree of wear to generate a grading index;

[0062] Based on the energy distribution characteristics extracted by wavelet decomposition, the embodiment of the present application constructs an eight-dimensional wavelet feature, which includes the energy proportion of the waveform distortion increment and the signal strength attenuation rate in each frequency band. The system inputs the 8-dimensional wavelet feature and the historical wear data into the support vector regressor based on the radial basis kernel function for processing. The kernel function parameter σ is set to 0.8, and the relaxation factor C is set to 10. The regressor outputs a wear quantification score value of the wear degree of the lock body, ranging from 0 to 1. The system divides the wear level according to the wear quantification score value of the wear degree of the lock body: 0 to 0.3 is slight wear, 0.3 to 0.6 is moderate wear, and 0.6 to 1.0 is severe wear. For example, when the score value is 0.4, it is judged to be moderate wear. The grading index is determined by experimental data statistics to ensure the reliability and consistency of the judgment;

[0063] Continuously monitor wear classification indicators and analyze wear trends;

[0064] In the embodiment of the present application, the system continuously monitors the lock body wear grading index and records the time point when the wear quantitative score of the lock body wear degree exceeds the threshold of each interval, for example, the time when the wear degree upgrades from mild wear to moderate wear. The exponential smoothing method is used to calculate the rate of change of the wear degree, and the smoothing coefficient is set to 0.3. If the wear degree increases rapidly from mild to moderate state, or from moderate state to severe state within 30 days, it indicates that the lock body wear rate is abnormal and further maintenance warnings need to be triggered. The system generates a trend sequence containing timestamps and wear levels to provide data support for dynamic adjustment of lock control parameters.

[0065] In an embodiment of the present application, by extracting waveform distortion increments and signal strength attenuation rates, combined with wavelet decomposition and support vector regression analysis, the system can accurately quantify the degree of lock body wear and generate a grading index. Compared with the traditional static threshold method, the embodiment of the present application significantly improves the sensitivity and accuracy of wear detection through multi-dimensional feature analysis and dynamic trend monitoring. For example, in actual tests, when the lock body is in a state of moderate wear, the system can identify potential instability in advance through the increase in the proportion of high-frequency energy, thereby providing a reliable basis for the optimization of lock control parameters. The embodiment of the present application does not impose too many restrictions on the specific dimensions of the wavelet features or the parameter settings of the regressor. It can be optimized and adjusted by technical personnel according to actual scenarios to adapt to different types of smart lock bodies and operating environments. The wear grading index and trend sequence generated by the system lay the foundation for subsequent predictive maintenance and dynamic parameter adjustment, effectively extending the service life of the smart lock and reducing safety hazards.

[0066] like Figure 1-Figure 2 As shown, in step S103, a prediction model is constructed using historical smart lock body wear data, and the wear state of the smart lock body within a preset time is predicted by combining the wear degree classification index and the change trend sequence, and a lock core gap change curve is generated through fitting analysis.

[0067] Furthermore, in step S103, in the embodiment of the present application, by analyzing the historical smart lock body wear data, an adaptive prediction model is constructed, the wear degree classification index and the change trend sequence are input, the wear state of the lock body in the future period of time is predicted, and the lock cylinder gap change curve is generated by the fitting method, providing a basis for dynamically optimizing the lock control parameters;

[0068] Extract wear characteristics from historical data and group them to build a wear trend prediction model;

[0069] In an embodiment of the present application, the system extracts a data set containing wear degree classification indicators and change trend sequences from the historical operation records of the smart lock body, and groups them according to the daily rotation frequency of the lock body. For example, the daily use frequency of the lock body is divided into three groups: 10-20 times, 20-40 times, and 40-50 times. Each group of data covers 30 consecutive days of operation records. A long short-term memory neural network is used to construct a wear trend predictor. The network has a three-layer structure: the input layer receives three characteristic parameters: the wear degree of the lock body, the rate of change of the wear degree of the lock body, and the lock body usage cycle; the hidden layer contains 128 neurons, and dynamically controls the retention and update of information through the forget gate and the input gate; the output layer generates the predicted value of the lock cylinder gap for the next 7 days. The training data contains 10,000 sets of historical records, of which 80% are used for model training and 20% are used for verification to ensure the generalization ability of the predictor. The embodiment of the present application does not impose too many restrictions on the grouping method or network structure, and can be adjusted according to actual scenarios.

[0070] Analyze wear trends through polynomial fitting and quantify wear acceleration characteristics;

[0071] For the trend sequence of each set of lock cylinder clearance prediction values, the embodiment of the present application adopts a fourth-order polynomial fitting method to generate a fitting curve reflecting the wear acceleration characteristics. The polynomial coefficient characterizes the change law of the wear rate. Under mild wear conditions, the absolute value of the quadratic term coefficient is usually less than 0.01, indicating that the wear process is stable; when moderate wear occurs, the quadratic term coefficient increases to 0.02 to 0.05, indicating that the wear rate is accelerating; when severe wear occurs, the quadratic term coefficient exceeds 0.05, reflecting a significant acceleration trend. The quadratic term system calculates the duration distribution of the wear state in the three intervals of mild, moderate and severe. For example, mild wear usually lasts for more than 90 days, moderate wear lasts for 30-60 days, and severe wear lasts for less than 30 days, providing a reference for subsequent predictions.

[0072] Use sliding time window to predict wear status and generate prediction sequence;

[0073] In an embodiment of the present application, based on the lock cylinder clearance prediction value output by the wear trend predictor, the system uses a sliding time window method to predict the short-term wear status. The time window is set to 30 days, the sampling interval is 1 day, and the predictor updates the prediction result once a day to generate a sequence containing a timestamp and the predicted wear status. The system calculates the lock cylinder clearance prediction error by comparing the deviation between the lock cylinder clearance prediction value and the actual measurement value using an exponential weighting method. The weight coefficient is set to 0.7 to emphasize the contribution of recent data. If the prediction deviation exceeds 10% for 5 consecutive days, the system triggers the online update mechanism and uses the back propagation algorithm to fine-tune the network weights at a learning rate of 0.01 to improve the prediction accuracy.

[0074] Establish a lock cylinder gap change curve and set up an early warning mechanism;

[0075] Based on the sequence of predicted values of the lock cylinder clearance, the embodiment of the present application constructs a continuous change curve of the lock cylinder clearance through the cubic spline interpolation method. The sampling point contains the timestamp, the predicted value of the lock cylinder clearance and the corresponding wear degree. In the new lock body state, the lock cylinder clearance is about 0.1 mm; when it is slightly worn, the clearance increase rate does not exceed 0.01 mm / month; when it is moderately worn, it increases to 0.02 to 0.03 mm / month; when it is severely worn, it can reach 0.05 mm / month. The system sets a clearance warning threshold of 0.3 mm. When the predicted clearance value exceeds this threshold, the trigger time point is recorded, and based on the clearance data of the last 30 days, the linear extrapolation method is used to estimate the time to reach the critical clearance of 0.5 mm. Actual measurements show that the time interval from warning to critical state is usually 60 to 90 days, providing a sufficient response window for maintenance intervention;

[0076] In an embodiment of the present application, by constructing a long short-term memory neural network predictor and combining it with polynomial fitting and sliding window methods, the system can accurately predict the wear status of the smart lock body and the change trend of the lock core gap. Compared with the traditional static model, the embodiment of the present application significantly improves the sensitivity and reliability of the prediction by dynamically analyzing multi-dimensional features. For example, in high-frequency usage scenarios, the system can identify the trend of accelerated wear in advance through the rapid increase of the quadratic term coefficient, thereby triggering an early warning mechanism. The embodiment of the present application does not impose too many restrictions on the time window length or fitting method, and can be optimized and adjusted by technical personnel according to the lock body characteristics and operating environment. The generated gap change curve provides accurate data support for subsequent threshold adjustments and maintenance decisions, effectively extending the service life of the smart lock and reducing the safety risks caused by wear.

[0077] like Figure 1-Figure 2 As shown, in step S104, the lock core gap change and the historical door body matching status data are combined to construct a dynamic mapping relationship between the magnetic induction signal strength and the gap parameter, and the magnetic induction threshold range is dynamically optimized to improve the judgment accuracy of the smart lock.

[0078] Furthermore, in step S104, in the embodiment of the present application, by analyzing the gap change in the lock core gap change curve and the historical door body matching state data, the system constructs a dynamic mapping model between the magnetic induction signal strength and various gap parameters, and dynamically adjusts the magnetic induction threshold range to ensure the stability and reliability of the smart lock in complex environments;

[0079] Extract and fit the door body coordination status data to generate a time series change curve;

[0080] In an embodiment of the present application, the system extracts fitting status data including door deformation, door frame displacement, and door lock fitting clearance from historical records and groups them at a sampling frequency of once per hour. Under normal circumstances, door deformation remains within 0.2 mm, door frame displacement does not exceed 0.3 mm, and door lock fitting clearance is between 1.0 and 1.5 mm. A cubic spline function is used to fit these data to generate a continuous curve reflecting the change in door fitting status over time. This curve can capture the periodic deformation caused by the temperature difference between day and night, such as the deformation of the door body due to daytime expansion reaching 0.15 mm and falling back to 0.05 mm at night, as well as the slow displacement trend of the door frame due to building settlement, with an average monthly increase of approximately 0.02 mm. The embodiment of the present application does not strictly limit the sampling frequency or fitting method, and can be adjusted according to actual scenarios;

[0081] Extract the stability trend of the door structure through sliding average filtering;

[0082] For the time series change curve of the door body matching state, the embodiment of the present application uses a 10-point sliding average filter to remove short-term disturbances caused by temperature fluctuations or external vibrations, extract the long-term change trend of the door body deformation and door frame displacement, and generate a door body structural stability index based on the filtered data, reflecting the deformation characteristics of the door body and door frame in long-term operation. For example, when the door panel thickness is 40 mm, the door frame is fixed with expansion bolts, and the lock body installation depth is 25 mm, the stability index is high, indicating that the structural deformation has little effect on the magnetic induction signal. If the stability index decreases, it may indicate that the door body structure is loose or abnormally deformed, and its impact on the magnetic induction signal needs to be further analyzed.

[0083] Construct a mapping model between gap and magnetic signal intensity to generate dynamic fit parameters;

[0084] Based on the door structure stability index, the embodiment of the present application uses a random forest regression algorithm to establish a mapping function between the lock core gap and the magnetic induction signal strength. The input features include door structure parameters, such as a door panel thickness weight coefficient of 0.8, a door frame fixing strength weight coefficient of 0.9, and a lock body installation depth weight coefficient of 1.0. The random forest model uses 100 decision trees to ensemble predictions and output dynamic parameters of the door lock matching to characterize the nonlinear effect of the gap change on the magnetic induction signal. For example, when the door body deformation is less than 0.1 mm, the signal strength attenuation is less than 5%; when the deformation reaches 0.2 mm, the attenuation increases to 15% to 20%; when it exceeds 0.3 mm, the attenuation is significantly increased. The model is trained with a large amount of historical data to ensure the accuracy of the mapping relationship.

[0085] Fit the signal intensity compensation curve and optimize the threshold range;

[0086] For the door lock with dynamic parameter sequence, the embodiment of the present application performs spectrum analysis to separate the low-frequency drift component and the high-frequency fluctuation component, and uses a second-order polynomial to fit the compensation curve of the magnetic induction signal strength through the least squares method to generate a reference threshold range for the magnetic induction signal. The system divides the threshold interval into three levels: low, medium, and high, corresponding to 70% to 85%, 85% to 115%, and 115% to 130% of the nominal signal, respectively. The signal fluctuation variance is calculated in each interval. Under normal conditions, the variance is less than 3% of the mean, while the variance can increase to more than 10% when the magnetic field is interfered with. Based on the variance characteristics, the system uses an adaptive weighting method to segment the signal to reduce the impact of the interference interval on the threshold adjustment. Actual measurements show that the signal fluctuation amplitude after adjustment is reduced by about 40%, the false alarm rate is reduced from 12% to 3%, and the recognition accuracy is improved to more than 95%;

[0087] Optimize signal judgment criteria through deep neural networks and Kalman filtering;

[0088] In an embodiment of the present application, the system extracts the lock core gap, door lock gap and door frame gap from the historical door body matching state data, maps the data to the range of 0 to 1 using maximum and minimum value normalization, and trains the mapping function of the gap parameter and the magnetic signal strength through a three-layer deep neural network. The network structure includes hidden layers of 128, 64 and 32 nodes, and is trained using the random gradient descent method with a batch size of 64 and a learning rate of 0.001. The training data contains 10,000 sets of records, 80% of which are used for training and 20% for verification. Combined with the lock core gap change trend output by the wear predictor and the real-time gap data collected by the door body displacement sensor, the system generates a theoretical change curve of the magnetic signal strength, and uses a 60-minute sliding window to segment the curve, and calculates the mean and Standard deviation: The measurement noise covariance of the Kalman filter is set to the square of the standard deviation, and the state noise covariance is set to one tenth of it. The upper and lower limits of the fluctuation range of the filtered signal are expanded and reduced by 10% respectively to form a dynamic threshold interval. According to the width of the fluctuation range, the system sets a three-level sampling frequency: when the fluctuation is less than one-quarter of the interval width, 10 seconds of low-frequency sampling is used; when it is between one-quarter and one-half, 5 seconds of medium-frequency sampling is used; when it is greater than one-half, 2 seconds of high-frequency sampling is used. Finally, combined with the door body matching state parameters, the weighted average method is used to calculate the judgment threshold, where the lock core gap weight is 0.5, the door lock gap weight is 0.3, and the door frame matching gap weight is 0.2. The threshold stability index for 24 consecutive hours shows that the standard deviation is lower than 3% of the mean, verifying the optimization effect.

[0089] In an embodiment of the present application, by integrating random forest regression, deep neural network and Kalman filtering, the system constructs a precise dynamic mapping model of magnetic induction signal strength and gap parameters, which significantly improves the adaptability and stability of threshold adjustment. Compared with the traditional fixed threshold method, the present application can effectively cope with the influence of door deformation and environmental interference. For example, when the abnormal deformation of the door body causes the signal standard deviation to increase to 15%, the system quickly adjusts the threshold through high-frequency sampling and filtering processing to maintain the judgment accuracy. The embodiment of the present application does not strictly limit the neural network structure or filtering parameters, and can be optimized by technical personnel according to the actual operating environment. The optimized threshold range and sampling strategy provide strong support for the reliable operation of the smart lock, significantly reduce the risk of misjudgment and extend the service life of the equipment.

[0090] like Figure 1-Figure 2 As shown, in step S105, based on the lock body wear degree grading index and the lock core gap prediction value, when the adjusted magnetic induction threshold exceeds the preset upper limit, the lock core tolerance range is identified and the target locking judgment standard is generated to ensure the reliable operation of the smart lock.

[0091] Furthermore, in step S105, in the embodiment of the present application, when the adjusted magnetic induction threshold exceeds the preset upper limit, the system dynamically identifies the lock core tolerance range based on the lock body wear degree and lock core clearance prediction data, generates a target locking judgment standard, and improves the judgment accuracy and stability of the smart lock in complex operating environments;

[0092] Obtain and smooth the magnetic induction threshold data sequence and determine the threshold over-limit situation;

[0093] In this application embodiment, the system obtains the adjusted magnetic induction threshold data sequence at a sampling frequency of 10 times per second through the magnetic induction signal acquisition module, and records the signal amplitude and timestamp. Since the original data is affected by environmental interference and electronic noise, the fluctuation amplitude may reach 15% of the nominal value. To improve data quality, the system uses a Kalman filter for smoothing, setting the measurement noise variance to one-tenth of the signal variance, for example, 0.02, and the state noise variance to one-tenth of the signal variance. After filtering, the signal fluctuation amplitude is reduced to less than 5%. The system compares the filtered threshold with a preset upper limit, for example, 2.5 millitesla. If the threshold exceeds the upper limit continuously, the exceeding time point is recorded and the subsequent tolerance adjustment process is triggered. The embodiment of the present application does not strictly limit the filtering parameters or the upper limit of the threshold, and can be optimized according to the actual operating environment.

[0094] Use deep neural networks to predict lock cylinder tolerance trends;

[0095] For the three levels of lock body wear grading indicators (mild, moderate, and severe), as well as the lock cylinder gap prediction value sequence, the embodiment of the present application constructs a four-layer deep neural network model, including an input layer, two hidden layers, and an output layer. The input layer receives the wear grading index and the gap prediction value, the hidden layer contains 64 and 32 neurons respectively, and uses the ReLU activation function. The output layer generates the lock cylinder tolerance change trend. The network training data set contains 1,000 sets of historical records, of which 40% are mild wear, 35% are moderate wear, and 25% are severe wear. The training adopts the stochastic gradient descent algorithm with a batch size of 32 and a learning rate of 0.001. After 50 rounds of training, the prediction accuracy reaches 90%. For example, in the moderate wear state, the network can predict that the lock cylinder tolerance will increase by 0.1 mm in the next 7 days, providing an accurate basis for threshold adjustment.

[0096] Calculate the magnetic induction fluctuation compensation amount and generate tolerance compensation parameters;

[0097] In the embodiment of the present application, the system uses an exponential weighting method to calculate the compensation amount of the magnetic induction signal fluctuation amplitude based on the tolerance increment and the signal strength compensation coefficient, and sets differentiated signal strength compensation coefficients according to the degree of wear: when the wear is slight, the signal strength compensation coefficient is 0.8 to reduce unnecessary compensation amplitude; when the wear is moderate, the signal strength compensation coefficient is increased to 1.2 to adapt to the trend of accelerated gap change; when the wear is severe, the signal strength compensation coefficient is 1.5 to expand the tolerance range to ensure the stability of the judgment. The tolerance values after compensation are 0.2 mm for slight wear, 0.3 mm for moderate wear, and 0.5 mm for severe wear. The system generates a tolerance compensation parameter sequence, records the compensation effect of each wear interval, and provides data support for the threshold segmentation calculation. This method effectively balances the requirements of sensitivity and stability by dynamically adjusting the signal strength compensation coefficient.

[0098] Generate target determination standard curve through segmented calculation and interpolation;

[0099] Based on the tolerance compensation parameter sequence, the embodiment of the present application performs segmented calculation of the upper threshold of the magnetic induction signal strength and uses a piecewise linear interpolation method to determine the threshold change slope of each interval. For example, the slope of the light wear interval is 0.01 per hour, which increases to 0.02 per hour for moderate wear and 0.05 per hour for severe wear. The system calculates the dynamic change of the upper threshold based on the slope value and sets the lower threshold to 80% of the upper limit to form a complete threshold interval. The threshold interval is smoothed using the cubic spline interpolation method to generate a continuous target judgment standard curve to ensure a smooth transition of the threshold boundary and avoid misjudgment due to sudden changes. The target judgment standard includes an upper threshold, a lower threshold, and a tolerance range to comprehensively characterize the operating status of the lock body.

[0100] Verify the stability of the judgment standard and trigger automatic calibration;

[0101] In the embodiment of the present application, the system sets a verification time window of 1 hour, continuously monitors the stability of the judgment standard curve, and calculates the standard deviation of the judgment results. If the standard deviation of 5 consecutive judgment results exceeds 10% of the mean, it indicates that the judgment result is unstable, and the system triggers an automatic calibration process. The calibration process includes re-collecting 100 sets of magnetic induction signal samples, updating the deep neural network parameters, and adjusting the tolerance compensation coefficient until the standard deviation is reduced to within 5% of the mean. The measured data shows that after calibration, the judgment accuracy rate is improved to more than 95%, and the misjudgment rate is reduced to less than 3%. The embodiment of the present application ensures the reliability of the judgment standard in long-term operation through a dynamic calibration mechanism, significantly reducing the risk of misjudgment caused by environmental changes or increased wear;

[0102] In an embodiment of the present application, through the combination of Kalman filtering, deep neural networks and piecewise interpolation methods, the system can accurately identify the tolerance range of the lock core and generate a stable target judgment standard. Compared with the traditional fixed threshold method, the embodiment of the present application significantly improves the adaptability of the smart lock under increased wear or environmental interference through dynamic compensation and automatic calibration. For example, under severe wear conditions, the system successfully avoids locking failure caused by increased gap by increasing the tolerance range and slope adjustment. The embodiment of the present application does not strictly limit the network structure or calibration cycle, and can be optimized by technical personnel according to the lock body characteristics and usage scenarios. The generated target judgment standard provides strong support for the reliable operation of the smart lock, effectively extending the equipment life and improving safety.

[0103] In this embodiment, the term "magnetic induction threshold" refers to a critical judgment range dynamically set based on the strength and waveform characteristics of the magnetic induction signal during the operation of the smart lock, which is used to determine whether the lock body is in a normal locked state. The core is to adjust the upper and lower threshold intervals in real time by analyzing parameters such as the strength, waveform distortion, and attenuation rate of the magnetic induction signal, combined with data such as the lock core gap, door body matching status, and wear degree. The threshold range reflects the magnetic field stability and mechanical matching status of the lock body under specific working conditions, and is the core basis for determining the effectiveness of lock control.

[0104] The trigger conditions during the application of the magnetic induction threshold are as follows:

[0105] Dynamically determine the lock status. Trigger condition: The real-time collected magnetic induction signal strength exceeds the reference threshold range (±20% of the nominal value). If the signal strength exceeds the upper limit (2.5 millitesla), the lock cylinder tolerance range identification is triggered (step S105). If the signal strength is lower than the lower limit (1.2 millitesla), the lock body is determined to be abnormally closed, triggering the lock control parameter adjustment (step S106).

[0106] Adapting to lock body wear and environmental interference, the trigger conditions are: lock body wear level classification (mild, moderate, severe) or door structure deformation (door frame displacement > 0.3 mm). Through random forest regression and deep neural network dynamic mapping of the relationship between gap and magnetic induction signal, the threshold range is adjusted (step S104). The threshold slope is calculated segmented according to the wear level (severe wear compensation coefficient is 1.5) to expand the tolerance range (step S105);

[0107] Optimize the control strategy. Trigger conditions: The magnetic induction signal fluctuation variance is greater than 10% (normal is 3%) or the misjudgment probability is greater than 5%. If the risk level is "high risk", the weight factor of the wear parameter is exponentially adjusted and the threshold is re-optimized (step S107).

[0108] like Figure 1-Figure 2 As shown, in step S106, the magnetic induction signal strength is monitored in real time to see whether it meets the target locking judgment standard, and the lock control parameters are dynamically adjusted to optimize the operating status of the smart lock.

[0109] Furthermore, in step S106, in the embodiment of the present application, the system dynamically optimizes the lock control parameters by real-time analysis of the matching between the magnetic induction signal strength and the target judgment criteria, thereby ensuring that the smart lock maintains an efficient and stable lock control state under different operating conditions;

[0110] Collect magnetic induction signals and divide them into threshold intervals to generate initial optimized values of locking control parameters;

[0111] In an embodiment of the present application, the system uses a high-speed differential comparator to collect magnetic induction signal strength data at a sampling frequency of 1000 Hz, and records the signal amplitude in real time. Based on the target lock judgment standard, the system divides the signal strength into three level intervals: high, medium, and low. The high intensity interval is greater than 2.0 millitesla, the medium intensity interval is 1.2 to 2.0 millitesla, and the low intensity interval is less than 1.2 millitesla. Each interval generates a corresponding initial optimization value of the lock control parameter through a mapping function. For example, the high intensity interval corresponds to a motor torque reference value of 0.8 Nm, the medium intensity interval is 0.6 Nm, and the low intensity interval is 0.4 Nm. This hierarchical mapping ensures that the lock control parameters are dynamically matched with the signal strength, thereby improving control accuracy. The embodiment of the present application does not strictly limit the interval division or the mapping function form, and can be optimized according to the lock body characteristics and application scenarios.

[0112] Combined with the lock body operating status data, a recursive neural network model is constructed to establish a dynamic mapping between motor torque and speed;

[0113] For the initial optimization values of the lock control parameters, the system combines the lock body operating status data to construct a three-layer recursive neural network model, which includes an input layer, a hidden layer with 64 neurons, and an output layer. The network adopts a cyclic structure with a time step of 10 milliseconds to capture the timing characteristics of the motor torque and speed. The input layer receives the torque reference value and current speed of the smart lock motor driving the lock cylinder or lock tongue in the smart lock operation range. The output layer outputs the optimized smart lock motor control signal, which controls the torque of the lock cylinder or lock tongue. The training data set contains 1000 sets of historical records, covering operating data under different load conditions. The random gradient descent method is used for training, and the learning rate is set to 0.001. After training, the network can accurately predict the torque adjustment requirements. For example, in the high-intensity range, the torque output deviation is controlled within 5%. This method significantly improves the response speed and stability of motor control through timing modeling.

[0114] Set fuzzy control rules to dynamically adjust the motor speed;

[0115] Based on the theoretical torque curve output by the recursive neural network, the embodiment of the present application designs seven fuzzy control rules for fine-tuning the motor speed. The rules use the speed deviation and the deviation change rate as input variables, which are divided into five fuzzy subsets: the speed deviation includes large negative, small negative, zero, small positive, and large positive; the deviation change rate includes fast decrease, slow decrease, stable, slow increase, and fast increase. The output variable is the speed correction amount, which is also divided into five levels. The fuzzy rules achieve dynamic control through 25 combinational logics. For example, when the speed deviation is small positive and the change rate is slow increase, a moderate positive correction amount is output. Actual measurements show that fuzzy control controls the speed fluctuation within 5% of the rated value, significantly improving the smoothness of the lock body operation. The embodiment of the present application does not strictly limit the number of rules or the division of fuzzy subsets, and can be adjusted according to actual needs;

[0116] Use the Kalman state observer to monitor the lock control state and trigger calibration;

[0117] In an embodiment of the present application, the system uses a high-precision position sensor to collect lock tongue position data with a resolution of 0.01 mm, recording the displacement curve during the unlocking and locking process. For example, the displacement of the lock tongue from locked to retracted is 20 mm. The system calculates the smoothness index of the displacement curve and evaluates the motion stability by the slope change rate. A smoothness of less than 0.1 indicates that the lock tongue moves smoothly. Based on this, the system constructs a Kalman state observer to simultaneously monitor three state variables: motor torque, speed, and lock tongue position. The observer sets the torque measurement noise standard deviation to 0.05 Nm, the speed to 10 rpm, and the position to 0.02 mm. The state noise covariance is one-tenth of the measurement noise. The system uses a 5-minute calibration window to record the standard deviation of the state parameters. If the torque standard deviation exceeds 10% of the mean, the speed exceeds 8%, or the position exceeds 5%, an automatic calibration process is triggered. The calibration includes collecting 100 sets of new samples and updating the neural network parameters and fuzzy rules until the parameters are stable. This method ensures the long-term reliability of the lock control state through multivariable observation and dynamic calibration.

[0118] In an embodiment of the present application, through the integration of high-speed signal acquisition, recursive neural networks, fuzzy control and Kalman state observation, the system realizes real-time optimization of lock control parameters. Compared with traditional static control methods, the present application can quickly adapt to changes in magnetic induction signals and fluctuations in the operating state of the lock body. For example, in the low-intensity range, the system avoids lock tongue jamming caused by overload by reducing torque output and fine speed adjustment. The embodiment of the present application does not strictly limit the sampling frequency or observer parameters, and can be optimized by technical personnel according to actual scenarios. The optimized lock control state data significantly improves the operating efficiency and stability of the smart lock, providing reliable protection for the safe operation of power facilities.

[0119] like Figure 1-Figure 2 As shown, in step S107, combined with the optimized lock control status data and change trend sequence, the probability of misjudgment is evaluated and it is determined whether the safety risk is lower than the preset threshold. If the risk exceeds the standard, the weight factor of the wear parameter of the magnetic induction threshold and the judgment tolerance is adjusted, and the lock control parameters are re-optimized to ensure the safe operation of the smart lock.

[0120] Furthermore, in step S107, in the embodiment of the present application, the system analyzes the optimized lock control state data and the change trend sequence, quantifies the probability of misjudgment and evaluates the security risk, dynamically adjusts the weight factor of the wear parameter of the magnetic induction threshold and the judgment tolerance, and optimizes the lock control parameters to improve the security and stability of the smart lock;

[0121] Extract lock control status data and calculate the misjudgment probability through deep neural network;

[0122] In an embodiment of the present application, the system extracts the smart lock motor torque curve, the smart lock motor speed change curve and the smart lock tongue position curve from the optimized lock control state data, and constructs a three-layer deep neural network model. The input layer receives 9 state parameters, including the smart lock motor torque, the smart lock motor speed, the smart lock motor current, the smart lock tongue position, the smart lock motor speed and the smart lock motor acceleration; the two hidden layers contain 128 and 64 neurons respectively, and use the ReLU activation function; the output layer generates the misjudgment score of the smart lock operation state. The network training data set contains 10,000 groups of historical records, of which 2,000 groups are misjudgment samples. The training adopts the stochastic gradient descent method, and the learning rate is set to 0.001. After training, the network can accurately identify state abnormalities. For example, when the lock tongue position deviation exceeds 0.1 mm, the misjudgment probability may increase to 0.1. This method significantly improves the accuracy of misjudgment prediction through multi-dimensional parameter modeling, providing a reliable basis for risk assessment;

[0123] Use Bayesian probability method to assess safety risks and classify risk levels;

[0124] Regarding the misjudgment scoring of the smart lock operation status, the embodiment of the present application uses the Bayesian probability calculation method to evaluate the security risk of the smart lock operation status, and divides the security risk level intervals of the smart lock operation status into low, medium and high: the misjudgment probability below 0.05 is low risk, 0.05 to 0.15 is medium risk, and above 0.15 is high risk. The system determines the probability distribution of each interval through historical data statistics. For example, during normal operation, 95% of the data is in the low-risk interval, and the proportion of medium and high-risk intervals increases significantly under abnormal conditions. The Bayesian method combines prior probability and real-time data to dynamically update the risk assessment results. For example, when the misjudgment probability exceeds 0.1 for 5 consecutive times, the system determines it as medium risk and triggers subsequent adjustments. The embodiment of the present application does not strictly limit the risk interval division and can be optimized according to actual scenarios;

[0125] Set adaptive weight update rules and dynamically compensate for magnetic induction judgment criteria;

[0126] Based on the risk assessment results, the system designs adaptive weight update rules. These rules determine the wear parameter weight factor based on the aforementioned risk level. This weight factor is then adjusted to dynamically compensate for the magnetic induction threshold and determination tolerance. The wear parameter weight factor can be understood as a parameter that compensates for these two factors. The wear parameter weight factor is dynamically adjusted based on the smart lock's security risk assessment results. It reflects the system's sensitivity adjustment requirements for different risk levels.

[0127] In the low-risk range, the weight factor for adjusting the wear parameter maintains its initial value of 1.0; in the medium-risk range, the weight factor for adjusting the wear parameter increases linearly by 0.1 for every 0.01 increase in the probability of misjudgment; in the high-risk range, the weight factor for adjusting the wear parameter is adjusted exponentially, with a base of 1.5 and an exponent equal to the multiple of the probability of misjudgment exceeding 0.15. Using the updated weight factor for adjusting the wear parameter, the system dynamically compensates for the magnetic induction threshold and the judgment tolerance. Assuming that the original magnetic induction threshold is set to 2.0 millitesla (mT), the threshold can be adjusted according to the new weight factor. For example, in the medium-risk case, if the weight factor becomes 1.05, the new magnetic induction threshold can be adjusted to 2.0 * 1.05 = 2.1 mT; and in the high-risk case, if the weight factor becomes 1.074, the new magnetic induction threshold is adjusted to 2.0 * 1.074 ≈ 2.148 mT.

[0128] Similarly, the same weighting factor can be applied to the decision tolerance. For example, if the original decision tolerance is ±0.2 mT, then in a medium-risk situation, the new decision tolerance can be adjusted to ±0.2 * 1.05 = ±0.21 mT; in a high-risk situation, the new decision tolerance is adjusted to ±0.2 * 1.074 ≈ ±0.2148 mT.

[0129] The magnetic induction threshold compensation coefficient is used to directly adjust the upper and lower thresholds of the magnetic induction signal strength, thereby ensuring that the smart lock can accurately determine its locked state. The magnetic induction threshold compensation coefficient is positively correlated with the probability of misjudgment. For example, when the probability of misjudgment is 0.1, the magnetic induction threshold compensation coefficient is 1.2, and the magnetic induction threshold compensation coefficient of the judgment tolerance is 1.3; when the probability increases to 0.2, the magnetic induction threshold coefficients increase to 1.4 and 1.6 respectively. The compensated judgment standard improves the sensitivity of abnormal state recognition. For example, in the high-risk area, the threshold range is narrowed by 10%, effectively reducing the risk of misjudgment.

[0130] Optimize locking control parameters through fuzzy control and state feedback;

[0131] In the embodiment of the present application, the system designs seven fuzzy control rules to update the locking control parameters based on the compensated magnetic induction judgment standard. The rules use the misjudgment probability and risk level as input variables and output parameter adjustment amounts, which are divided into five levels: when both are low, the adjustment amount is zero; when any indicator increases, the adjustment amount increases proportionally; when both are high, the maximum adjustment amount is used. The system further constructs a state feedback compensator and adopts a proportional integral control method to adjust the motor torque and speed in real time. The proportional coefficient is set to 0.8 and the integral time constant is 0.5 seconds. During the motor startup phase, the torque compensation amount can reach 20% of the nominal value; during stable operation, the speed compensation amount is controlled within 5%. This method ensures the accuracy and stability of parameter adjustment through fuzzy control and real-time feedback. The embodiment of the present application does not strictly limit the number of rules or control parameters, and can be adjusted according to actual needs;

[0132] Verify the stability of the lock control state and trigger parameter optimization;

[0133] For the optimized locking control parameters, the system sets a one-hour parameter verification cycle, continuously records the stability indicators of the locking control status data, and uses standard deviation to assess the degree of fluctuation. If the torque standard deviation exceeds 8% of the mean, the speed exceeds 6%, or the lock tongue position exceeds 4%, the optimization effect is judged to be unsatisfactory, and a new round of parameter optimization is triggered. The optimization process includes re-collecting 100 sets of status data, updating the deep neural network parameters, and adjusting the fuzzy control rules and compensation coefficients of the status parameters. Actual measurements show that after 3 to 4 rounds of optimization, the fluctuation of the status parameters can be controlled within the ideal range. For example, the torque standard deviation is reduced to within 5% of the mean. This method significantly improves the long-term reliability of the locking control system through periodic verification and dynamic optimization.

[0134] In the embodiment of the present application, accurate assessment of the probability of misjudgment and dynamic optimization of lock control parameters are achieved. Compared with the traditional static threshold method, the present application can quickly respond to abnormal status. For example, in the high-risk area, the misjudgment rate is reduced to below 3% through exponential weight adjustment and strict threshold compensation. The embodiment of the present application does not strictly limit the network structure or verification cycle, and can be optimized by technical personnel according to the lock body characteristics and operating environment. Ultimately, the optimized lock control parameters and control strategies provide strong support for the safe operation of the smart lock, significantly reduce safety hazards and extend the service life of the equipment.

[0135] Those skilled in the art can make various other corresponding changes and deformations based on the technical solutions and concepts described above, and all of these changes and deformations should fall within the scope of protection of the claims of this application.

Claims

1. A method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing, characterized in that: include: Obtain the magnetic induction signal sequence when the smart lock is running and then denoise it to generate a denoised signal sequence; Calculate and extract the attenuation parameters of the hyperbolic attenuation model parameters of the signal amplitude in the original magnetic induction signal sequence for the denoised signal sequence, and fit the change trend sequence; Taking the trend sequence as input, the difference between the current waveform distortion and the historical baseline value is calculated as the distortion increment and decomposed. A support vector regression model is constructed to output a quantitative wear score of the lock body wear degree, and the wear index is divided according to the score. After the wear index is divided, a gap change curve is generated based on the judgment result of whether the predicted value of the lock core gap exceeds the preset threshold; The predicted value of the lock core gap is used as a parameter to establish a door body matching status database; a gap-magnetic induction intensity mapping model is constructed, the signal intensity compensation coefficient is output and segmented optimization is performed, and the interval boundaries of the segmented optimization are dynamically adjusted; When the high-zone threshold value is continuously exceeded during segmented optimization, the tolerance increment of the gap between the lock cylinder and the door body for the next 7 days is output; the magnetic induction fluctuation compensation amount is calculated based on the tolerance increment; and a target judgment standard curve is generated; Based on the target judgment standard of the target judgment standard curve, the magnetic induction signal strength level is divided into levels, and the torque output of the smart lock motor driving the lock cylinder or lock tongue is dynamically adjusted; Based on the lock control status data of the torque output, a deep neural network misjudgment probability model is constructed to output a misjudgment score. The weight factor of the wear parameter is dynamically adjusted according to the risk level until the torque fluctuation of the smart lock motor driving the lock cylinder or lock tongue meets the preset parameters.

2. The method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing according to claim 1 is characterized in that: Obtain the magnetic induction signal sequence when the smart lock is running and then denoise it to generate a denoised signal sequence, including: The sequence is subjected to a third-order high-pass filter with a cutoff frequency of 5 Hz and a second-order low-pass filter with a cutoff frequency of 500 Hz using a Butterworth digital filter to generate a denoised signal sequence.

3. The method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing according to claim 1, characterized in that: After the wear index is divided, a gap change curve is generated according to a judgment result of whether the lock core gap prediction value exceeds a preset threshold, including the following operations: A prediction model is constructed using a long short-term memory neural network. The input layer receives the wear quantitative score, the rate of change of the lock body wear degree, and the average daily usage frequency of the lock body, and outputs the predicted value of the lock cylinder gap in the next 7 days. When the lock cylinder clearance prediction value exceeds the lock cylinder clearance prediction threshold, a maintenance warning is triggered and a lock cylinder clearance change curve including a timestamp is generated.

4. The method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing according to claim 1, characterized in that: When the high threshold exceeds the limit continuously, the following operations are performed: A deep neural network is used to predict the lock cylinder tolerance trend. The wear level and gap prediction value are input and the output is the tolerance increment of the gap between the lock cylinder and the door body in the next 7 days. The magnetic induction fluctuation compensation amount is calculated based on the tolerance increment and the signal strength compensation coefficient, where the signal strength compensation coefficient is set to mild, moderate, and severe according to the wear level; The target determination standard curve was generated by cubic spline interpolation.

5. The method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing according to claim 1, characterized in that: The target determination standard curve is used as a benchmark to divide the magnetic induction signal strength level and dynamically adjust the torque output of the smart lock motor driving the lock cylinder or lock tongue, including the following operations: Real-time acquisition of magnetic signal strength, which is divided into three levels: high (>2.0mT), medium (1.2-2.0mT), and low (<1.2mT); Combined with the lock body operating status data, a recursive neural network model is constructed, which inputs the torque reference value and speed of the smart lock motor driving the lock cylinder or lock tongue in the smart lock operating range, and outputs the optimized smart lock motor control signal; A fuzzy controller is used to dynamically adjust the torque output using the speed deviation and the deviation change rate as input variables.

6. The method for controlling smart locks of electric power facilities based on multi-dimensional authority data processing according to claim 1, characterized in that: The method includes constructing a deep neural network misjudgment probability model based on the lock control state data of the torque output, outputting a misjudgment score, and dynamically adjusting the weight factor of the wear parameter according to the risk level until the torque fluctuation of the smart lock motor driving the lock cylinder or lock tongue meets the preset parameters, including the following operations: Build a deep neural network misjudgment probability model, input the smart lock motor torque, smart lock motor speed, and smart lock tongue position parameters, and output a misjudgment score of the smart lock operation status in the range of 0-1; The Bayesian method is used to classify the security risk level of the smart lock operation status; Dynamically adjust the weighting factor of wear parameters based on risk level: low risk is maintained at 1.0, medium risk increases linearly, and high risk exponential compensation base is 1.5; The parameter re-optimization process is triggered until the torque fluctuation of the smart lock motor driving the lock cylinder or lock tongue meets the preset parameters.

Images