Elevator door lock short circuit detection method, electronic equipment and medium

Abstract

The invention provides an elevator door lock short circuit detection method, electronic equipment and a medium. The method comprises the steps that visual time sequence data, vibration time sequence data, spectrum time sequence data and environment time sequence data of an elevator door lock indicator light are obtained; extracting a first visual feature parameter and a second visual feature parameter; extracting a first vibration characteristic parameter and a second vibration characteristic parameter; extracting a first spectral characteristic parameter and a second spectral characteristic parameter; according to the first judgment condition, the initial state of the elevator door lock is determined; the optimization state of the elevator door lock is determined through a second judgment condition; and an improved Bayesian network is used for fusion decision making, and an elevator door lock short circuit detection result is obtained. According to the method, the spectral information of the indicator lamp is introduced, vision, vibration and environmental data are combined, deep physical characteristics such as wavelength drift and bandwidth change are extracted, an improved Bayesian network is constructed to realize dynamic fusion decision, an existing detection blind area is effectively broken through, and the detection accuracy and robustness in a complex environment are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of elevator safety testing technology, and more specifically, to an elevator door lock short-circuit detection method, electronic equipment, and medium. Background Technology

[0002] Elevator door lock systems are core components ensuring passenger safety. Whether the door lock circuit is short-circuited directly affects the prevention of fatal accidents such as doors opening and elevator movement. Traditional testing methods require two certified personnel to work together on the car top: one person performs jogging maintenance while the other disconnects the main and auxiliary door locks in sequence and observes whether the elevator stops to determine the short-circuit status. This method is time-consuming, labor-intensive, and inefficient. Frequent starts and stops exacerbate mechanical wear, and manual assistance can easily lead to misjudgments, making it difficult to meet the needs of large-scale, routine testing.

[0003] In recent years, machine vision-based elevator door lock status recognition technology has emerged. This technology uses cameras to capture images of door lock indicator lights on the control cabinet's mainboard, employs target detection networks to identify the on / off state, and combines this with door lock opening / closing signals to infer short circuits. However, elevator machine rooms are complex environments with severe background interference such as lighting changes and metal reflections. Relying solely on visual information can easily lead to false detections or missed detections. To address this, some studies have introduced vibration sensors to collect relay action signals, improving detection robustness through visual, vibration, and environmental influences. However, these methods are still limited to perceiving the appearance of the indicator lights, such as brightness and color, and cannot capture the microscopic changes in the internal luminous characteristics of the indicator lights. When the indicator lights age, the drive circuit malfunctions, or there are early faults, visual, vibration, and conventional light intensity detection methods are all insufficient to reflect the true state, creating detection blind spots. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method for detecting short circuits in elevator door locks, which can improve the problems of difficulty in determining the cause of the fault, low detection efficiency, large equipment wear and tear, and high labor costs in traditional detection methods.

[0005] To achieve the above technical objectives, the technical solution adopted in this application is as follows:

[0006] In a first aspect, this application provides a method for detecting short circuits in elevator door locks. The method includes: acquiring visual time-series data, vibration time-series data, spectral time-series data, and environmental time-series data of an elevator door lock indicator light; extracting a first visual feature parameter and a second visual feature parameter using the visual time-series data; extracting a first vibration feature parameter and a second vibration feature parameter using the vibration time-series data; extracting a first spectral feature parameter and a second spectral feature parameter using the spectral time-series data; determining a preliminary state of the elevator door lock based on the first visual feature parameter and the first vibration feature parameter, through a first judgment condition; determining an optimized state of the elevator door lock based on the preliminary state, through a second judgment condition; and performing a fusion decision using an improved Bayesian network based on the optimized state, the environmental time-series data, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter to obtain a short circuit detection result for the elevator door lock.

[0007] Optionally, extracting the first visual feature parameter and the second visual feature parameter using the visual time-series data includes: inputting the visual time-series data into a target detection network to extract the detection area of ​​the indicator light; based on the detection area, extracting the first visual feature parameter using a first algorithm, wherein the first visual feature parameter includes the brightness abrupt change rate; and based on the detection area, extracting the second visual feature parameter using a second algorithm, wherein the second visual feature parameter includes the chromaticity offset.

[0008] Optionally, extracting the first and second vibration characteristic parameters using the vibration time-series data includes: using the vibration time-series data to perform filtering and wavelet transform to obtain wavelet coefficients; mapping the squared modulus of the wavelet coefficients from the original domain to the high-resolution domain, and combining it with the redistribution scale spectrum to obtain the instantaneous energy spectrum of the vibration signal; based on the instantaneous energy spectrum, extracting the first vibration characteristic parameter using a third algorithm, the first vibration characteristic parameter including the peak value of the pull-in impact and the pull-in response time; and based on the instantaneous energy spectrum, extracting the second vibration characteristic parameter using a fourth algorithm, the second vibration characteristic parameter including the released echo energy and the oscillation attenuation rate.

[0009] Optionally, extracting the first and second spectral feature parameters using the spectral time-series data includes: preprocessing the spectral time-series data to obtain preprocessed spectral time-series data; extracting the peak wavelength and full width at half maximum (FWHM) of the spectrum at each moment based on the preprocessed spectral time-series data to obtain a center wavelength time-series sequence and a bandwidth time-series sequence; extracting the first spectral feature parameter based on the center wavelength time-series sequence using a fifth algorithm, wherein the first spectral feature parameter characterizes the spectral center wavelength drift characteristics of the indicator light; and extracting the second spectral feature parameter based on the bandwidth time-series sequence using a sixth algorithm, wherein the second spectral feature parameter characterizes the transient rate of change of the spectral bandwidth of the indicator light.

[0010] Optionally, the fifth algorithm includes: using the door lock status synchronization signal to determine the time interval during which the indicator light is stably illuminated; obtaining a residual sequence based on the center wavelength sequence of the time interval; and using the residual sequence to determine the center wavelength drift index, wherein the calculation formula for the center wavelength drift index is as follows:

[0011]

[0012] in, As a center wavelength drift index, The number of time scales to be divided, No. The set of sampling points corresponding to each time scale The number of samples in this set. For the first The mean of the residuals at each time scale For the first The window length is determined by a time scale. For reference time scale, The scale decay constant is As the range penalty factor, For the reference center wavelength, The center wavelength value in the residual sequence. The value is positive. The sixth algorithm includes: using the door lock status synchronization signal to locate the instant the door lock relay operates; determining the bandwidth sequence of the action transition process based on the time point; and determining the bandwidth mutation crispness index based on the bandwidth sequence. The calculation formula for the bandwidth mutation crispness index is as follows:

[0013]

[0014] in, This is an indicator of bandwidth abrupt change in crispness. and These represent the typical bandwidth values ​​for the stable phases before and after the action. The duration of the transition process. This represents the number of sampling points within the transition process interval. This is the measured bandwidth value. The fitted value is the ideal step response. The bandwidth step amplitude, For regularization terms, This represents the number of bandwidth pre-change events detected during the transition process. For the number of pre-change event points, This represents the total number of sampling points. This is a penalty factor for pre-change events.

[0015] Optionally, determining the preliminary state of the elevator door lock based on the first visual feature parameter and the first vibration feature parameter, and through a first judgment condition, includes: obtaining a first deviation degree based on the first visual feature parameter, and obtaining a second deviation degree based on the first vibration feature parameter; determining the comprehensive interference value of the current environment based on the environmental time series data; determining a fusion deviation degree based on the comprehensive interference value; comparing the fusion deviation degree with a first preset threshold; if the fusion deviation degree is less than the first preset threshold, determining the door lock to be in a normal state; if the fusion deviation degree is not less than the first preset threshold, determining the relative difference coefficient between the first deviation degree and the second deviation degree; comparing the relative difference coefficient with a second preset threshold; if the relative difference coefficient is less than the second preset threshold, determining the current state to be in a state requiring review; if the relative difference coefficient is not less than the second preset threshold, selecting the state corresponding to the feature parameter with the smaller value between the first deviation degree and the second deviation degree as the preliminary state.

[0016] Optionally, based on the preliminary state, determining the optimized state of the elevator door lock through the second judgment condition includes: based on the preliminary state, determining a confidence quantification value of the preliminary state using visual deviation, vibration deviation, fusion deviation, and relative difference coefficient; comparing the confidence quantification value with a third preset threshold; if the confidence quantification value is higher than the third preset threshold, then the preliminary state is taken as the optimized state; if the confidence quantification value is not higher than the third preset threshold, then outputting a consistency score between chromaticity offset and oscillation attenuation rate in time sequence; comparing the consistency score with a fourth preset threshold; if the consistency score is lower than the fourth preset threshold, then correcting the preliminary state according to environmental time sequence data, and taking the corrected state as the optimized state; if the consistency score is not lower than the fourth preset threshold, then maintaining the preliminary state as the optimized state.

[0017] Optionally, based on the optimized state, the environmental time-series data, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter, a fusion decision is made using an improved Bayesian network to obtain the short-circuit detection result of the elevator door lock. This includes: determining an improved Bayesian network based on the optimized state; generating a dynamic prior probability for the current moment using the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, the second vibration feature parameter, and the environmental time-series data; inputting the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter into the posterior probability calculation formula of the improved Bayesian network, and outputting the posterior probability, wherein the posterior probability includes a first posterior probability and a second posterior probability, and the calculation formula of the posterior probability incorporates an environmental perception attention modulation index and a cross-modal coupling coordination factor; comparing the magnitudes of the first posterior probability and the second posterior probability, and taking the state with the higher posterior probability value as the short-circuit detection result.

[0018] Secondly, this application also provides an electronic device, which includes a processor and a memory coupled to each other, wherein the memory stores a computer program, and when the computer program is executed by the processor, the electronic device performs the method described thereon.

[0019] Thirdly, this application also provides a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the method described therein.

[0020] Compared with existing technologies, the advantages of the technical solution in this application include:

[0021] (1) The spectral characteristics of indicator lights are introduced into the field of door lock status detection. By extracting deep physical characteristics such as center wavelength drift and bandwidth change, the limitations of existing detection are broken through. It can capture the evolution of microluminescence characteristics caused by aging, pollution or electrical abnormality, and make up for the defects of existing technology in sensing changes in the internal state of indicator lights.

[0022] (2) A four-dimensional synchronous acquisition and multimodal fusion framework containing visual, vibration, spectral and environmental information was constructed. Data quality was ensured through timestamp alignment and preprocessing, providing a rich and reliable foundation for feature extraction and decision-making.

[0023] (3) In visual feature extraction, a calculation model for brightness mutation rate and chromaticity offset was designed, which effectively suppressed background interference and achieved a robust characterization of the indicator light brightness dynamics and chromaticity stability.

[0024] (4) In vibration feature extraction, a high-resolution instantaneous energy spectrum is constructed by redistribution wavelet transform, and the characteristic parameters of absorption impact and release rebound are introduced. Combined with signal-to-noise ratio correction and wave penalty mechanism, the mechanical state identification capability in complex environment is significantly improved.

[0025] (5) By combining dynamic environmental interference assessment and adaptive weight fusion with modal consistency constraints and the principle of minimum anomaly, the intelligent fusion of visual and vibration features is realized, which effectively reduces the false alarm rate caused by environmental fluctuations.

[0026] (6) A multi-dimensional confidence quantification model and a secondary feature time series consistency analysis mechanism were constructed to intelligently correct the initial state confidence when it is insufficient, thereby enhancing the robustness of the decision-making process.

[0027] (7) An improved Bayesian network was proposed, which introduced environmental awareness attention weights and cross-modal coupling factors to realize dynamic soft reconstruction of the network structure, enabling spectral features to play a key discriminative role in high-noise scenarios, and significantly improving the accuracy and robustness of fusion decision-making. Attached Figure Description

[0028] This application can be further illustrated by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate some embodiments of this application and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without any inventive effort.

[0029] Figure 1 This is a flowchart of an elevator door lock short-circuit detection method provided in an embodiment of this application.

[0030] Figure 2 This is a flowchart of step S4 of an elevator door lock short circuit detection method provided in an embodiment of this application. Detailed Implementation

[0031] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts are referred to by the same reference numerals in the drawings or description. Implementations not shown or described in the drawings are forms known to those skilled in the art. In the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0032] Please refer to Figure 1 This application provides a method for detecting short circuits in elevator door locks, comprising the following steps:

[0033] S1. Acquire visual timing data, vibration timing data, spectral timing data, and environmental timing data of the elevator door lock indicator light.

[0034] In one embodiment, four types of time-series data are first collected synchronously using multiple sensors. Specifically, this includes: using an industrial camera to continuously capture image sequences at a fixed frame rate in the door lock indicator area on the elevator control cabinet motherboard to obtain visual time-series data; simultaneously, a high-sensitivity accelerometer is installed near the door lock relay to collect vibration signals generated during the relay's engagement and disengagement in real time, serving as vibration time-series data. The door lock indicator includes, but is not limited to, a power indicator and a hard disk activity indicator.

[0035] Furthermore, a miniature fiber optic spectrometer is installed unobstructed to the side of the elevator door lock indicator light to continuously collect spectral data from the indicator light at a high sampling rate, thus obtaining spectral time-series data. Simultaneously, environmental sensors collect light intensity, ambient temperature, and humidity parameters to form environmental time-series data. It should be noted that the technologies used in acquiring spectral data include, but are not limited to, spectroscopic techniques.

[0036] Furthermore, the above four types of data are subjected to timestamp alignment and data preprocessing, including noise reduction, filtering, and normalization.

[0037] The advantage of this method is that by integrating visual, vibration, spectral, and environmental information, it can comprehensively capture key state changes during the operation of elevator door locks, providing a rich and reliable data foundation for high-precision and high-robustness door lock short-circuit detection.

[0038] In a specific practical application, an industrial camera, at a frame rate of 30 frames per second, continuously captured 300 frames of images for 10 seconds, pointing directly at the door lock indicator light on the elevator control cabinet's mainboard. Simultaneously, an accelerometer recorded the vibration waveform of the relay's operation at a sampling rate of 12.8 kHz; a miniature fiber optic spectrometer acquired the indicator light's spectrum at a sampling rate of 1 kHz; and an environmental sensor recorded an illumination intensity of 320 Lux and a temperature of 24.5°C. The humidity was 55%RH; all data was precisely timestamped via a time synchronization server.

[0039] S2. Using the visual time-series data, extract the first visual feature parameter and the second visual feature parameter.

[0040] In one embodiment, the collected visual temporal data is first input into an improved YOLOv10n object detection network. It should be noted that YOLOv10n is a publicly available network. The improvements include: incorporating an ECA attention mechanism module into the backbone network using a C2f module to suppress background interference from the elevator control cabinet motherboard and enhance signal feature extraction of elevator indicator lights; replacing the PAN network with a Lite-BiFPN network in the neck network; and replacing the CloU function with a Focal-SloU loss function at the detection head end, incorporating angle and shape information, adding sample weights, enhancing the regression accuracy and high-precision detection of indicator light bounding boxes, and improving the robustness of elevator door lock indicator classification. This network has been pre-trained on an elevator door lock indicator light dataset and can accurately identify and locate the indicator light detection region in each frame of the image. The network output is a detection result containing the coordinates of the indicator light bounding box.

[0041] Further, based on the detection area, a first visual feature parameter is extracted using a first algorithm. This first visual feature parameter characterizes the dynamic brightness characteristics of the indicator light. The specific implementation steps of the first algorithm are as follows: First, calculate the average brightness of all pixels within the detection area of ​​each frame; then calculate the brightness difference between adjacent frames; next, calculate the variance of pixel brightness within the detection area as a measure of brightness fluctuation; then, construct a brightness mutation rate calculation model, which satisfies the following relationship:

[0042]

[0043] in, For the first The total number of pixels in the frame detection area; This represents the brightness difference between adjacent frames. For the first Frame brightness; It must be a very small positive number to prevent division by zero errors; This is a regularization term to ensure numerical stability; This is a preset reference variance threshold used to normalize brightness fluctuations. The variance of pixel brightness within the detection area is used as a measure of brightness variability.

[0044] Furthermore, based on the detection area, a second visual feature parameter is extracted using a second algorithm. This parameter characterizes the chromaticity stability of the indicator light. The specific implementation steps of the second algorithm are as follows: First, the image of the detection area is converted from the RGB color space to the HSV color space, and the hue and saturation components are extracted. Then, the mean hue and mean saturation of all pixels in the detection area are calculated. Next, the deviation of the hue and saturation from the preset reference values ​​is calculated. Finally, the formula for calculating the chromaticity offset is constructed as follows:

[0045]

[0046] in, This is the chromaticity offset. and These are the maximum permissible deviation ranges for hue and saturation, respectively, used for normalization processing; The standard deviation of saturation within the detection area, This is the penalty factor weighting coefficient, used to adjust the impact of saturation fluctuations on chromaticity stability; It is a very small positive number; To detect the mean hue of all pixels within the region, The average saturation value. and This is a preset reference value.

[0047] It should be noted that by introducing the standard deviation of saturation as a penalty factor, this formula can effectively identify chromatic aberration caused by changes in lighting or aging of indicator lights, thus improving the sensitivity to abnormal conditions.

[0048] The advantage of this method is that it can accurately locate small indicator light areas through an improved target detection network, and by combining a brightness mutation rate calculation model and a chromaticity shift calculation model, it can extract highly robust feature parameters from visual data, effectively suppressing background noise and illumination change interference.

[0049] In a specific real-world application, based on 300 frames of collected images, the improved YOLOv10n network accurately identified areas of [missing information] in each frame. The indicator light area of ​​the pixel; then, the first algorithm calculated that at frame 50, the average pixel brightness in the detection area was 208, the brightness difference between adjacent frames was 5, and the brightness fluctuation variance was 12.3; the brightness change rate of this frame was calculated to be 0.18; the second algorithm calculated the brightness offset to be 0.25, indicating that the indicator light color was stable and there was no obvious drift.

[0050] S3. Using the vibration time series data, extract the first vibration characteristic parameter and the second vibration characteristic parameter.

[0051] In one embodiment, the collected vibration time-series data is first... Bandpass filtering is performed to remove low-frequency drift and high-frequency noise interference, retaining the frequency components related to the door lock relay operation. Then, continuous wavelet transform is applied to the filtered signal to obtain the time-frequency expression of the vibration signal.

[0052]

[0053] in, For the time frequency of the vibration signal, As a scale factor, The translation factor is... The wavelet basis functions are used to construct an instantaneous energy spectrum based on the squared modulus of the wavelet coefficients, which is used to characterize the distribution characteristics of vibration energy in time and frequency.

[0054] It is important to note that the construction of the instantaneous energy spectrum is not a simple direct mapping of the squared modulus of the wavelet coefficients, but rather a transformation combining the redistribution scale spectrum and the wavelet framework to achieve high-precision localization of time-frequency energy. Its expression is as follows:

[0055]

[0056] in, For instantaneous energy spectrum, These are continuous wavelet transform coefficients. It is a scale factor (inversely proportional to frequency). This is the translation factor (corresponding to time). This is a scale normalization factor used to compensate for energy attenuation in wavelet transform at different scales, ensuring the physical consistency of the energy spectrum. The Dirac function is used to redistribute energy, repositioning energy that was originally scattered across the time-frequency plane to a more precise location. The local time reallocation operator is calculated using the following formula:

[0057]

[0058] in, Wavelet coefficients with respect to time The partial derivatives, Wavelet coefficients with respect to time The partial derivative yields the conjugate complex number. As a regularization factor, Indicates taking the real part, The instantaneous frequency redistribution operator is calculated using the following formula:

[0059]

[0060] in, The center frequency of the wavelet basis function. Wavelet coefficients for scaling The partial derivatives, The complex conjugate of the continuous wavelet transform coefficients. This indicates taking the imaginary part. This is a regularization factor; the algorithm uses a redistribution technique to adjust the squared modulus of the wavelet coefficients from the original... Domain mapping to high resolution The domain eliminates the time-frequency ambiguity caused by the Heisenberg uncertainty principle inherent in wavelet transform, significantly improving the time-frequency focusing of the instantaneous energy spectrum; the final obtained It can accurately depict the impact moment, frequency components, and energy distribution of a relay during operation.

[0061] Furthermore, based on the instantaneous energy spectrum, a first vibration characteristic parameter is extracted using a third algorithm. This parameter characterizes the closing impact characteristics of the door lock relay. The specific implementation steps of the third algorithm are as follows: First, locate the moment when the door lock action command is issued on the time axis. This moment can be synchronously obtained from the control system; then, in Subsequent time windows The system searches for local maxima in the instantaneous energy spectrum, identifies the impact waveform generated at the moment the relay engages, and records the moment the impact waveform appears. Next, the peak amplitude of the impact waveform is calculated. :

[0062]

[0063] in, The characteristic frequency where the impact energy is most concentrated is defined. Finally, a set of characteristic parameters for the attraction impact is constructed, including the peak value of the attraction impact and the attraction response time, and their calculation formulas are as follows:

[0064]

[0065] in, To absorb the peak impact, For the pull-in response time, This is a preset reference impact amplitude used for normalization. For regularization terms; The mean squared error of the background noise before and after the action; The average energy of the impact signal; and This is an adjustment factor used to control the impact of the signal-to-noise ratio on the peak confidence level. Nominal response time; This is the frequency offset penalty coefficient; The formula, which introduces a signal-to-noise ratio correction term and a frequency offset penalty term to the nominal characteristic frequency, can effectively distinguish between normal and abnormal engagement, such as the vibration characteristics of contact sticking and mechanical jamming.

[0066] Furthermore, based on the instantaneous energy spectrum, a second vibration characteristic parameter is extracted using a fourth algorithm. This parameter characterizes the release rebound characteristics of the door lock relay. The specific implementation steps of the fourth algorithm are as follows: First, at the moment the door lock release command is issued... Subsequent time window The system identifies the oscillating echo sequence generated after the relay releases. Then, it performs multi-scale energy decomposition on the oscillating echo, dividing the total energy into several sub-bands according to frequency bands. Next, it extracts the total energy value of the oscillating echo and the decay rate of the oscillation amplitude. Finally, it constructs a set of release rebound characteristic parameters, including the release echo energy and the oscillation decay rate, calculated using the following formula:

[0067]

[0068]

[0069] in, To release echo energy; This refers to the oscillation decay rate; This refers to the number of sub-bands. For the first The instantaneous energy of the individual belt; For the first The center frequency of each sub-band; This is the system's resonant frequency; For bandwidth parameters; The number of identified oscillation periods, For the first The peak amplitude of each oscillation cycle, The time interval between adjacent peak values. It should be a very small positive number to prevent logarithmic operations from diverging. Let be the standard deviation of the oscillation envelope. The average amplitude, The formula is a fluctuation penalty factor. It is calculated by weighting the frequency band energy and the logarithmic decay rate, and introduces an envelope fluctuation penalty term. It can accurately characterize the damping characteristics of the mechanical system after release and effectively identify abnormal rebound behavior caused by spring fatigue and contact wear.

[0070] The advantage of this method is that it extracts the attraction impact characteristics and release rebound characteristics with clear physical meaning from the vibration signal through time-frequency analysis technology. Combined with complex signal-to-noise ratio correction, frequency offset penalty and fluctuation penalty mechanism, it can effectively suppress environmental vibration interference and accurately capture the minute mechanical state changes of the door lock relay.

[0071] In a specific practical case, the system issued an attraction command at 14:32:15.820. The third algorithm searched the instantaneous energy spectrum within a 5-millisecond time window after the command was issued, and located the impact waveform at 14:32:15.822, with a peak value of 2.45 m / s² and a characteristic frequency of 3200 Hz. Combined with the background noise standard deviation of 0.12 m / s², the normalized peak value of the attraction impact was calculated to be 1.12. During the release phase, the fourth algorithm extracted the total energy of the release echo as 0.085, with an oscillation attenuation rate of 385. This indicates that the damping characteristics are normal after the relay is released.

[0072] S4. Using the spectral time series data, extract the first spectral feature parameter and the second spectral feature parameter.

[0073] Please see Figure 2 In one embodiment, a miniature fiber optic spectrometer is first used to continuously acquire the spectral data of the indicator light at a sampling rate of not less than 1 kHz. At each sampling moment, a set of light intensity distributions covering the visible light band is obtained. The raw spectral data is preprocessed by dark current subtraction, wavelength calibration and normalization to eliminate the influence of sensor noise and ambient light fluctuations. Then, the peak wavelength and full width at half maximum (FWHM) of the spectrum at each moment are accurately extracted by a Gaussian multi-peak fitting algorithm to obtain the center wavelength time series and bandwidth time series.

[0074] Furthermore, based on the aforementioned center wavelength time series, a first spectral feature parameter is extracted using a fifth algorithm. This parameter characterizes the spectral center wavelength drift characteristics of the indicator light. The specific implementation steps of the fifth algorithm are as follows: First, the time interval for stable illumination of the indicator light is determined using the door lock status synchronization signal, and the time interval corresponds to the stable closing or opening period of the door lock relay. Then, the trend term of the center wavelength sequence within this interval is removed, and the slow changes caused by temperature drift are eliminated using a local weighted regression scatter smoothing method to obtain the residual sequence. Next, multi-scale fluctuation analysis is performed on the residual sequence to calculate the wavelength jitter characteristics at different time scales. Finally, a center wavelength drift index is constructed, and its calculation formula is as follows:

[0075]

[0076] in, As a center wavelength drift index, The number of time scales to be divided, No. The set of sampling points corresponding to each time scale The number of samples in this set. For the first The mean of the residuals at each time scale For the first The length of a window on a time scale. For reference time scale, The scale attenuation constant; It is a range penalty factor used to enhance sensitivity to sudden large fluctuations; For the reference center wavelength, For the first The center wavelength value in each residual sequence; It is a constant.

[0077] Compared to existing technologies, this formula, through multi-scale fluctuation analysis combined with scale weighting and range penalty, can effectively identify and quantify minute wavelength drifts caused by indicator light aging, drive current fluctuations, or optical contamination.

[0078] Furthermore, based on the bandwidth time series, a sixth algorithm is used to extract spectral feature parameters, which characterize the transient change rate of the indicator light's spectral bandwidth. The specific implementation steps of the sixth algorithm are as follows: First, the time point of the door lock relay's action (engagement or release) is located using the door lock status synchronization signal; then, time windows of a certain length are taken before and after the action time point to construct the bandwidth sequence of the action transition process; next, adaptive segmented fitting is performed on this sequence to identify the bandwidth change process from a steady-state value to a new steady-state value; finally, a bandwidth abrupt change crispness index is constructed, the calculation formula of which is:

[0079]

[0080] in, This is an indicator of bandwidth abrupt change in crispness. and These represent the typical bandwidth values ​​for the stable phases before and after the action; The duration of the transition process is the time it takes for the bandwidth to change from 90% of the original steady-state value to 10% of the new steady-state value (or vice versa); This represents the number of sampling points within the transition process interval. This is the measured bandwidth value. The fitted value is the ideal step response; The bandwidth step amplitude, This is a regularization term used to control sensitivity to waveform distortion. This represents the number of bandwidth pre-change events detected during the transition process, which manifests as minute bandwidth jitter. For the number of pre-change event points, This represents the total number of sampling points; The penalty factor for pre-change events;

[0081] Compared to existing technologies, this formula, by introducing the reciprocal of the transition time as the basic rate of change and combining it with waveform distortion penalty terms and pre-change event penalty terms, can accurately characterize the crispness of the indicator light bandwidth change at the moment of action: if the bandwidth change is rapid and without warning jitter, then... The values ​​are relatively large; if there is unstable driving current or abnormal transient response of the luminescent material, the waveform distortion penalty and pre-change penalty will be significantly reduced. This value effectively identifies abnormal behavior of the indicator lights associated with the door lock status.

[0082] In a specific implementation case, during the time period from 14:32:15.800 to 14:32:15.850, the data collected by the spectrometer was fitted with Gaussian to obtain the center wavelength time sequence. The fifth algorithm analysis showed that during the stable illumination period of the indicator light, the calculated value of the center wavelength drift index was 0.032nm, indicating that the wavelength was very stable. The sixth algorithm focused on the moment of the suction action at 14:32:15.820, analyzed the bandwidth change, measured the duration of the transition process to be 4 milliseconds, and calculated the bandwidth change crispness index to be 2.85, indicating that the indicator light responded quickly and crisply.

[0083] S5. Based on the first visual feature parameters and the first vibration feature parameters, determine the preliminary state of the elevator door lock through the first judgment condition.

[0084] In one embodiment, the first visual feature parameters extracted in step S2 are obtained. The peak value of the pull-in impact in the first vibration characteristic parameter extracted in step S3 These two parameters are respectively compared with the preset visual standard parameters. and vibration standard parameters The two sides are compared, and their respective first deviations are calculated. The calculation of the first deviation not only considers the absolute deviation, but also incorporates historical statistical characteristics for adaptive normalization. The specific formula is as follows:

[0085]

[0086] in, The first deviation of the visual feature parameters. The first deviation of the vibration characteristic parameters; and These are the mean and standard deviation of the historical visual feature parameters, respectively; and These are the mean and standard deviation of the historical vibration characteristic parameters, respectively. and It is a very small positive number. It is a nonlinear enhancement factor. It is an exponential adjustment factor.

[0087] Compared to existing technologies, this formula normalizes by incorporating historical statistical characteristics and employs an exponential enhancement term, which can effectively amplify abnormal deviations and improve sensitivity to minor anomalies.

[0088] Furthermore, based on the environmental time-series data collected in step S1, the comprehensive interference value of the current environment is determined. The calculation of the comprehensive interference value integrates three dimensions: light intensity fluctuation, environmental vibration and noise level, and electrical power frequency interference level. The calculation formula is as follows:

[0089]

[0090] in, and These are the standard deviation and mean of light intensity, respectively, reflecting the degree of light intensity fluctuation; and These represent the standard deviation and mean of environmental vibration noise, respectively. and These are the standard deviation and mean of the power frequency interference amplitude, respectively. , , The weighting coefficients for each interference source can be calibrated according to the actual application scenario. This is a bias term used to adjust the baseline value of the interference value; It is a very small positive number; the formula uses the Sigmoid function to map multidimensional environmental interference to the (0,1) interval. The closer it is to 1, the more severe the current environmental interference is, and the closer it is to 0, the more ideal the environmental conditions are.

[0091] Furthermore, based on the comprehensive interference value, the fusion weights of visual deviation and vibration deviation are dynamically adjusted. The weight adjustment follows the principle of reducing the weight of sensitive modes under high interference environments. The specific calculation formula is as follows:

[0092]

[0093]

[0094] in, and These are the fusion weights for the visual and vibrational modes, respectively. and The modal sensitivity coefficients reflect the sensitivity of the visual and vibration modes to environmental disturbances, respectively, and can be determined through prior knowledge or experimental calibration. The Softmax form of the weight calculation formula ensures that the sum of the weights is 1, and when environmental disturbances increase, the weights of the more sensitive modes will adaptively decrease, thereby reducing their negative impact on the fusion results.

[0095] Furthermore, based on the dynamically adjusted weights, the fusion deviation between visual features and vibration features is calculated. The calculation formula is as follows:

[0096]

[0097] in, These are the cross-term enhancement coefficients, used to strengthen the confidence level when the two modes are consistent. It is a very small positive number; the cosine term is used to measure the consistency of the deviation between the two modes: when and When the values ​​are similar, the cosine value is close to 1, and the contribution of the cross term is positively enhanced; when the difference between the two is large, the cosine value is close to 0, and the contribution of the cross term is weakened. This formula not only realizes weighted fusion, but also introduces modal consistency constraints through the cross term, which effectively improves the robustness of the fusion result.

[0098] Next, the fusion deviation was compared. With the first preset threshold The size; the first preset threshold is determined based on the statistical distribution of fusion deviation under historical normal operating conditions. This is achieved by collecting a large number of samples of elevators operating under known fault-free and favorable environmental conditions, and calculating the mean of their fusion deviation. and standard deviation ,Will Set as ,in The confidence coefficient is used to ensure sufficient sensitivity to potential abnormal states while covering the vast majority of normal fluctuations.

[0099] like If so, the door lock is considered to be in a normal state; if Then, a more detailed discrimination process is initiated; at this point, the relative difference coefficient between visual deviation and vibration deviation is calculated, and its calculation formula is as follows:

[0100]

[0101] in, As an environmental modulation factor, The threshold is set as the environmental interference baseline. The Sigmoid term in the denominator is used to dynamically adjust the sensitivity of the difference coefficient according to the degree of environmental interference: when the environmental interference exceeds the threshold, the Sigmoid term increases, the denominator of the difference coefficient increases, thereby reducing the sensitivity to modal differences and avoiding misjudgments caused by environmental interference.

[0102] Furthermore, compare the relative coefficients of difference. Second preset threshold The size; the second preset threshold is calibrated by offline analysis of historical fault data and interference data. Specifically, data on visual misjudgment caused by single-mode interference, such as strong light flashing, and real physical faults, such as door lock jamming causing abnormal vibration, are collected in simulated or real environments. The corresponding relative difference coefficient is calculated, and with the goal of maximizing the distinction between modal divergent interference and modal consistent faults, the optimal classification boundary is found through ROC curve analysis or clustering algorithm, and this boundary value is set as the second preset threshold.

[0103] like This indicates that the deviation between the two modes is relatively consistent, but the fusion deviation exceeds the standard, which is due to systematic environmental interference or sensor drift. At this time, the current state is determined to be a state that needs to be reviewed, and the system will trigger the review process, such as increasing the sampling frequency or enabling the backup sensor.

[0104] like This indicates a significant discrepancy between the two modal deviations. In this case, the state corresponding to the feature parameter with the smaller value between the visual deviation and the vibration deviation is selected as the initial state. That is, following the principle of minimum anomaly, the mode that is closer to the normal mode is used for state determination in order to avoid unnecessary shutdown caused by false alarms of a single mode.

[0105] Compared to existing technologies, this method has the advantage of achieving intelligent fusion of visual and vibration features through dynamic environmental interference assessment and adaptive weight adjustment mechanisms. Combined with modal consistency constraints and the principle of minimum anomalies, it can accurately identify abnormal door lock states under complex and ever-changing environmental conditions, effectively reducing the false alarm rate caused by interference factors such as light fluctuations and environmental vibrations, and providing reliable preliminary state input for optimizing state judgment.

[0106] In a specific practical case, the visual brightness mutation rate of 0.18 and the vibration impact peak of 1.12 were used as inputs; the visual history mean was 0.15 with a standard deviation of 0.02; the vibration history mean was 1.05 with a standard deviation of 0.05; the calculated visual deviation was 1.8 and the vibration deviation was 1.6. Environmental sensor data showed that the standard deviation of light fluctuation was 8 Lux and the standard deviation of environmental vibration was 0.05 m / s², resulting in a calculated comprehensive interference value of 0.2 (good environment); after dynamic weight calculation, the visual weight was 0.48 and the vibration weight was 0.52, with a fusion deviation of 1.62. This value is less than the first preset threshold of 2.0, therefore the door lock is determined to be in a normal state.

[0107] S6. Based on the preliminary state, determine the optimized state of the elevator door lock through the second judgment condition.

[0108] In one embodiment, the preliminary state determined in step S5 is first obtained. The system incorporates intermediate variables in the decision-making process, including visual bias, vibration bias, fusion bias, and relative difference coefficient. Based on these variables, a confidence quantification of the initial state is constructed. This confidence quantification not only reflects the degree of certainty in the decision but also incorporates the influence of modal consistency and environmental disturbance factors. Its calculation formula is as follows:

[0109]

[0110] in, This is a confidence quantification of the initial state. For fusion deviation, For visual deviation degree, For vibration deviation, The penalty intensity coefficient, It is a very small positive number. The total interference value, and These are the standard deviation and mean of the fluctuation in historical decision-making, respectively.

[0111] Compared to existing technologies, this formula achieves a more refined quantification of confidence by multiplying and fusing three factors: when environmental interference is severe, modal divergence is significant, or fusion deviation is large, the confidence level decreases significantly; conversely, when the environment is ideal, the modalities are consistent, and the deviation is small, the confidence level approaches 1.

[0112] Further, determine the third preset threshold. Compare confidence quantification values With the third preset threshold The size; the third preset threshold is set based on the decision reliability requirements, and usually adopts a combination of field calibration and theoretical analysis: first, a large amount of data containing various working conditions, such as ideal environment, strong interference, minor fault, and serious fault, is collected, and the corresponding confidence level is calculated. Then, through expert experience or decision risk model, a minimum confidence threshold is determined that can ensure that high confidence level can be passed directly, low confidence level must be reviewed, and the risk of misjudgment is acceptable. This threshold is used as the third preset threshold.

[0113] like This indicates that the initial state has high reliability, and the system directly transmits the initial state. As an optimized state, no further processing is needed, thus improving decision-making efficiency; if This indicates that the confidence level of the initial state is insufficient, requiring the triggering of an optimization process based on environmental compensation. At this point, the second visual feature parameter (chromaticity shift) extracted in step S2 and the second vibration feature parameter (oscillation decay rate) extracted in step S3 are introduced. The temporal consistency score of these two parameters is calculated. The consistency score is calculated based on cross-correlation analysis and phase synchronization assessment within a sliding time window, specifically using a method that fuses the weighted cross-correlation coefficient with the instantaneous phase difference. The calculation formula is as follows:

[0114]

[0115] in, A consistency score is obtained for chromaticity shift and oscillation decay rate. The number of sliding time windows. For the first The set of sampling points within a time window and The first The average of chromaticity shift and oscillation decay rate within a time window. and The first The instantaneous principal phase values ​​of two parameter sequences within a time window (extracted by Hilbert transform). The value is a very small positive number to prevent division by zero errors. The first term of the formula is the weighted cross-correlation coefficient, which measures the linear correlation of the two parameter sequences in terms of amplitude. The second term is the phase synchronization penalty factor. When the instantaneous phase difference between the two sequences is close to 0 or π, it indicates that they have good synchronization and the penalty is small. When the phase difference is close to π / 2, the penalty is large. Through multiplication fusion, It can comprehensively assess the consistency of the two parameters in terms of amplitude variation patterns and phase evolution trends: when both change synchronously and their phases match, Approaching 1; when the two change patterns are contradictory or phase mismatched, Significantly reduced.

[0116] Further, determine the fourth preset threshold. And compare consistency scores With the fourth preset threshold The size of the fourth preset threshold is determined by quantifying the benchmark consistency between the second visual feature and the second vibration feature under normal operating conditions. Specifically, when the elevator installation and commissioning are completed or it is confirmed to be in a healthy state, a sufficiently long period of normal operating data is collected, the consistency score of the chromaticity offset and the oscillation attenuation rate is calculated, and after obtaining its statistical distribution, the fourth preset threshold is set as the lower percentile of the distribution, which is used as the dividing line for judging whether the secondary feature deviates significantly from the normal synchronization relationship.

[0117] like This indicates that the second visual feature and the second vibration feature have good temporal consistency, and even if the confidence of the initial state is insufficient, the initial state can still be maintained as the optimal state. This is because the secondary features of the two independent modes exhibit consistency, which corroborates the reliability of the initial state.

[0118] like This indicates a significant inconsistency between the second visual feature and the second vibration feature. Therefore, the initial state needs to be corrected based on environmental time-series data. The correction strategy employs an environmentally modulated state offset function, taking into account the current environmental interference values. Based on historical state transition probabilities, generate corrected optimal states. The correction rule is as follows: If the initial state is normal but the visual and vibration secondary characteristics are inconsistent, and the environmental interference value is high. If the initial state is abnormal but the secondary characteristics are inconsistent, and the environmental interference value is low, then the state will be corrected to a state requiring review; If the status is positive, the status will be corrected to a status requiring review; otherwise, the initial status will be maintained but a low confidence level will be added.

[0119] The advantage of this method is that it can accurately evaluate the credibility of the initial state by constructing a multi-dimensional confidence quantification model, and introduce the temporal consistency analysis of secondary features when the confidence is insufficient, and combine environmental interference information for intelligent correction, which effectively solves the problem of feature misjudgment that may occur in complex environments.

[0120] In a specific implementation case, the confidence level is calculated based on the initial normal state of S5. Specifically, the fusion deviation is 1.62, the difference between visual and vibration deviations is small, and environmental interference is low, resulting in a calculated confidence level of 0.85. This value is greater than the third threshold of 0.7, indicating that the initial state has high credibility. Therefore, the system directly uses the initial normal state as the optimized state without triggering the environmental compensation process. The entire decision-making process is efficient and reliable, confirming that the door lock is currently in a normal operating state without short circuits.

[0121] S7. Based on the optimized state, the environmental time series data, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter, a fusion decision is made using an improved Bayesian network to obtain the short circuit detection result of the elevator door lock.

[0122] In one embodiment, the optimization state determined in step S6 is first... This paper constructs the basic framework of an improved Bayesian network. Unlike existing Bayesian networks that use fixed naive Bayesian or tree structures, the improved Bayesian network proposed in this step introduces a dynamic attention modulation layer and a cross-modal coupling factor. Existing Bayesian networks usually assume that features are conditionally independent or are only connected by simple directed edges, making it difficult to dynamically adapt to the impact of environmental changes on feature reliability. Based on the traditional conditional probability table, this improved network adds an environment-aware attention modulation module. This module adjusts the likelihood probability weights of each feature node in real time according to the current environmental time series data, realizing dynamic soft reconstruction of the network structure.

[0123] Furthermore, based on the network structure and initial conditional probability parameters, an improved Bayesian network model is constructed; the network structure includes four feature nodes: a second visual feature parameter, a first spectral feature parameter, a second spectral feature parameter, and a second vibrational feature parameter; the state nodes are binary variables, taking values ​​of "short-circuited state" or "non-short-circuited state";

[0124] It is important to note that the core invention of introducing spectral feature parameters into the improved Bayesian network is: dynamically modulating the likelihood probability contribution of bispectral features (center wavelength drift and bandwidth abrupt change) through environmental awareness attention weights, so that it can adapt to environmental interference; at the same time, by using cross-modal coupling factors, the independence assumption is actively relaxed under specific environmental conditions, enhancing the joint discriminative power between spectral and vibration features, thereby accurately capturing the weak spectral distortion induced by door lock short circuit in high-noise scenarios.

[0125] Furthermore, based on the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, the second vibration feature parameter, and environmental time series data, the dynamic prior probability at the current moment is calculated by an adaptive prior probability generator. This generator integrates the confidence of historical statistical priors and the optimized state. The formula for calculating the confidence has been given in the description of S6 and will not be repeated here.

[0126] Finally, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter are used as observational evidence and input into the updated improved Bayesian network model for posterior probability inference. Compared with the existing Bayesian network, the posterior probability calculation formula of the improved Bayesian network introduces the environmental perception attention modulation index and the cross-modal coupling coordination factor. Its core formula is as follows:

[0127]

[0128] in, Given the observational evidence and environmental conditions, this represents the posterior probability. This represents the actual state of the elevator door lock; a value of 1 indicates a short-circuited state, while a value of 0 indicates a non-short-circuited state. For the specific values ​​that the state variables can take, For the second visual feature parameters, This is the first spectral characteristic parameter (spectral center wavelength drift index). This is the second spectral characteristic parameter (spectral bandwidth abrupt change in crispness index). The second vibration characteristic parameter, which characterizes the release rebound properties of the door lock relay, is specifically the oscillation attenuation rate extracted in step S3. This represents the dynamic prior probability at the current moment. Given a state Observed features Class conditional probability, For the first One observed characteristic variable, For environmental time series data, For the first The environmental awareness attention weights of each feature node are calculated using the following formula:

[0129]

[0130] in, The modal fundamental sensitivity coefficient reflects the first modal sensitivity coefficient. The degree to which each characteristic is sensitive to environmental changes; The modally optimal operating point indicates the environmental conditions under which this feature is most reliable; To adapt the bandwidth to the environment, control the smoothness of weight changes as the environment changes; It serves as a feature stability compensation factor, adjusting the degree to which the feature's own stability compensates for the weights. For the first The standard deviation of the historical values ​​of a feature reflects the volatility of the feature itself; For the first The mean of the historical values ​​of each feature; It is a constant; This is the cross-modal coupling modulation factor, which is an important difference between the improved Bayesian network and the traditional Bayesian network. It is used to relax the feature independence assumption under specific environmental conditions. When the environmental conditions are suitable and there is a strong correlation between modes, this factor is greater than 1 and acts as an exponent on the likelihood probability product term to amplify the joint contribution of the coupled modes.

[0131]

[0132] in, This is the coupling enhancement amplitude coefficient, which controls the maximum strength of the coupling effect; This is the environmental center point where the coupling effect is strongest, meaning that the coupling effect is most significant under these environmental conditions. The coupling bandwidth controls the rate at which the coupling effect decays as it deviates from the environment. Indicates the state given The conditional correlation coefficient between the first spectral feature and the second vibrational feature reflects the intrinsic coupling strength between the two modes. An environmental modulation steepness coefficient is used to control the steepness of the sigmoid function. The comprehensive interference value, taken from step S5, reflects the overall interference level of the current environment. This is the coupling activation threshold; when environmental disturbances exceed this threshold, the coupling effect is suppressed. Using Gaussian radial basis functions ensures that coupling effects are similar to those in the environment. At its strongest, This is a sigmoid function that implements nonlinear gating of the coupling effect by environmental disturbances. When the environmental disturbance is low, the output is close to 1, and when the environmental disturbance is high, the output is close to 0.

[0133] Furthermore, the calculated posterior probabilities are compared, if... If the condition is met, the final detection result will be a short-circuited state; otherwise, the non-short-circuited state will be output. This posterior probability inference process is executed in real time during each sampling period to ensure the timeliness and accuracy of the detection results.

[0134] The advantages of this method lie in the fact that the improved Bayesian network adds a dynamic attention modulation layer and a cross-modal coupling factor to the traditional structure. Unlike the fixed naive or tree structure and feature condition independence assumption of traditional Bayesian networks, this network includes four feature nodes: visual, bispectral, and vibration, and introduces an environment perception module. This module dynamically modulates the likelihood probability contribution of each feature node based on real-time environmental time-series data through attention weights. At the same time, it relaxes the independence assumption through the coupling factor in specific environments, enhances the joint discrimination power of spectral and vibration, realizes dynamic soft reconstruction of the network structure, significantly improves the robustness of fusion decision-making in complex environments, and significantly improves the accuracy and reliability of elevator door lock short circuit detection.

[0135] In a specific practical case, the optimized state (normal) and the center wavelength drift index of 0.032 nm and the bandwidth abrupt change crispness index of 2.85 extracted in step S4, as well as the oscillation attenuation rate of 385 s extracted in step S3, are considered. -1As input to the improved Bayesian network; the current environment is good (interference value 0.2), and the dynamic attention weight calculation shows that the spectral feature weight is slightly higher; the cross-modal coupling factor is calculated to be 1.01; the network inference yields the posterior probabilities: P(short-circuit state) = 0.12, P(non-short-circuit state) = 0.88; since the non-short-circuit state has a higher probability, the final output is the non-short-circuit state, confirming that the elevator door lock is operating normally.

[0136] In summary, this invention achieves accurate perception of changes in the internal state of indicator lights by introducing spectral features, and constructs a multi-level, multi-modal intelligent fusion and decision-making system. It significantly improves detection dimensions, anti-interference capabilities, and decision reliability. It can improve the problems of difficult fault cause determination, low detection efficiency, large equipment wear and high labor costs in traditional detection methods. It provides a comprehensive, accurate, and robust solution for elevator door lock short-circuit detection, and has important engineering application value and technological innovation significance.

[0137] The elevator door lock short-circuit detection method proposed in this embodiment, which integrates spectral features, introduces indicator light spectral data on the basis of existing visual, vibration mode, and ambient light data. It extracts deep physical features such as center wavelength drift and bandwidth changes, captures the evolution of the indicator light's luminous characteristics caused by aging, contamination, or electrical abnormalities, and performs multimodal fusion of spectral features with visual, vibration, and environmental information to achieve accurate identification of changes in the indicator light's internal state. This provides a more comprehensive and reliable solution for elevator door lock short-circuit detection.

[0138] This application provides an electronic device that may include a processing module and a storage module. The storage module stores a computer program, which, when executed by the processing module, enables the electronic device to perform the corresponding steps in the elevator door lock short-circuit detection method described above. The electronic device in this embodiment is not limited to desktop computers, laptops, tablets, etc.

[0139] In this embodiment, the processing module can be an integrated circuit chip with signal processing capabilities. The processing module can be a general-purpose processor. For example, the processor can be a central processing unit, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0140] The storage module can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc.

[0141] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device described above can be referred to the corresponding steps in the aforementioned method, and will not be elaborated further here.

[0142] This application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the elevator door lock short-circuit detection method as described in the above embodiments.

[0143] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium and includes several instructions to cause a computer device to execute the methods described in various implementation scenarios of this application.

[0144] In the embodiments provided in this application, it should be understood that the disclosed apparatus, systems, and methods can also be implemented in other ways. The apparatus, systems, and methods embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which includes one or more executable instructions for implementing a specified logical function. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0145] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for detecting short circuits in elevator door locks, characterized in that, The method includes the following steps: Acquire visual time-series data, vibration time-series data, spectral time-series data, and environmental time-series data of the elevator door lock indicator lights; Using the aforementioned visual time-series data, extract the first visual feature parameters and the second visual feature parameters; Using the vibration time series data, extract the first vibration characteristic parameter and the second vibration characteristic parameter; Using the spectral time-series data, extract the first spectral feature parameter and the second spectral feature parameter; Based on the first visual feature parameters and the first vibration feature parameters, the initial state of the elevator door lock is determined by the first judgment condition. Based on the initial state, the optimized state of the elevator door lock is determined by the second judgment condition; Based on the optimized state, the environmental time series data, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter, an improved Bayesian network is used to perform fusion decision-making to obtain the short circuit detection result of the elevator door lock.

2. The method according to claim 1, characterized in that, Extracting the first visual feature parameter and the second visual feature parameter using the aforementioned visual temporal data includes: The visual temporal data is input into the target detection network to extract the detection area of ​​the indicator light; Based on the detection area, a first visual feature parameter is extracted using a first algorithm. The first visual feature parameter includes the brightness abrupt change rate. Based on the detection area, a second visual feature parameter is extracted using a second algorithm. The second visual feature parameter includes a chromaticity offset.

3. The method according to claim 1, characterized in that, Using the vibration time series data, the extraction of the first vibration characteristic parameter and the second vibration characteristic parameter includes: The vibration time series data is used to perform filtering and wavelet transform to obtain wavelet coefficients; The squared modulus of the wavelet coefficients is mapped from the original domain to the high-resolution domain, and combined with the redistribution scale spectrum to obtain the instantaneous energy spectrum of the vibration signal. Based on the instantaneous energy spectrum, a first vibration characteristic parameter is extracted using a third algorithm. The first vibration characteristic parameter includes the peak value of the pull-in impact and the pull-in response time. Based on the instantaneous energy spectrum, a second vibration characteristic parameter is extracted using a fourth algorithm. The second vibration characteristic parameter includes the released echo energy and the oscillation attenuation rate.

4. The method according to claim 1, characterized in that, Using the aforementioned spectral time-series data, the extraction of the first spectral feature parameter and the second spectral feature parameter includes: The spectral time series data is preprocessed to obtain preprocessed spectral time series data. Based on the preprocessed spectral time series data, the peak wavelength and full width at half maximum (FWHM) of the spectrum at each moment are extracted to obtain the center wavelength time series and bandwidth time series. Based on the center wavelength time sequence, the first spectral feature parameter is extracted by the fifth algorithm. The first spectral feature parameter characterizes the spectral center wavelength drift characteristics of the indicator light. Based on the bandwidth time series, a second spectral feature parameter is extracted using a sixth algorithm. The second spectral feature parameter characterizes the transient rate of change of the indicator's spectral bandwidth.

5. The method according to claim 4, characterized in that, The fifth algorithm includes: The time interval for the indicator light to illuminate stably is determined by using the door lock status synchronization signal; Based on the time interval, the center wavelength sequence is obtained; Based on the central wavelength sequence, the residual sequence is obtained; Using the residual sequence, the center wavelength drift index is determined, and the calculation formula for the center wavelength drift index is as follows: ； in, As a center wavelength drift index, The number of time scales to be divided, No. The set of sampling points corresponding to each time scale The number of samples in this set. For the first The mean of the residuals at each time scale For the first The length of a window on a time scale. For reference time scale, The scale decay constant, As the range penalty factor, For the reference center wavelength, The center wavelength value in the residual sequence. The value is positive; the sixth algorithm includes: The timing of the door lock relay's action is determined by using the door lock status synchronization signal. Based on the stated time points, determine the bandwidth sequence of the action transition process; Based on the bandwidth sequence, a bandwidth mutation crispness index is determined, and the calculation formula for the bandwidth mutation crispness index is as follows: ； in, This is an indicator of bandwidth abrupt change in crispness. and These represent the typical bandwidth values ​​for the stable phases before and after the action. The duration of the transition process. This represents the number of sampling points within the transition process interval. This is the measured bandwidth value. The fitted value is the ideal step response. The bandwidth step amplitude, For regularization terms, This represents the number of bandwidth pre-change events detected during the transition process. For the number of pre-change event points, This represents the total number of sampling points. This is a penalty factor for pre-change events.

6. The method according to claim 1, characterized in that, Based on the first visual feature parameters and the first vibration feature parameters, the preliminary state of the elevator door lock is determined by the first judgment condition, including: Based on the first visual feature parameters, a first deviation degree is obtained. Based on the first vibration characteristic parameter, the second deviation degree is obtained; Based on the aforementioned environmental time-series data, determine the overall interference value of the current environment; Based on the comprehensive interference value, the fusion deviation is determined; The fusion deviation is compared with a first preset threshold. If the fusion deviation is less than the first preset threshold, the door lock is determined to be in a normal state. If the fusion deviation is not less than the first preset threshold, the relative difference coefficient between the first deviation and the second deviation is determined. The relative difference coefficient is compared with a second preset threshold; if the relative difference coefficient is less than the second preset threshold, the current state is determined to be a state requiring review; if the relative difference coefficient is not less than the second preset threshold, the state corresponding to the feature parameter with the smaller value between the first deviation degree and the second deviation degree is selected as the initial state.

7. The method according to claim 1, characterized in that, Based on the initial state, the optimized state of the elevator door lock is determined by the second judgment condition, including: Based on the preliminary state, the confidence quantification of the preliminary state is determined using visual deviation, vibration deviation, fusion deviation, and relative difference coefficient. The confidence quantification value is compared with the third preset threshold. If the confidence quantification value is higher than the third preset threshold, the initial state is taken as the optimization state. If the confidence quantification value is not higher than the third preset threshold, the consistency score of chroma offset and oscillation decay rate in time is output. The consistency score is compared with a fourth preset threshold. If the consistency score is lower than the fourth preset threshold, the initial state is corrected based on the environmental time series data, and the corrected state is taken as the optimized state. If the consistency score is not lower than the fourth preset threshold, the initial state is maintained as the optimized state.

8. The method according to claim 1, characterized in that, Based on the optimized state, the environmental time-series data, the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter, an improved Bayesian network is used for fusion decision-making to obtain the elevator door lock short-circuit detection result, including: Based on the optimization state, an improved Bayesian network is determined; Using the second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, the second vibration feature parameter, and the environmental time series data, a dynamic prior probability for the current moment is generated; The second visual feature parameter, the first spectral feature parameter, the second spectral feature parameter, and the second vibration feature parameter are input into the posterior probability calculation formula of the improved Bayesian network, and the posterior probability is output. The posterior probability includes the first posterior probability and the second posterior probability. The calculation formula of the posterior probability introduces the environmental perception attention modulation index and the cross-modal coupling coordination factor. Compare the first posterior probability with the second posterior probability, and take the state with the higher posterior probability value as the short-circuit detection result.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory coupled together, the memory storing a computer program that, when executed by the processor, causes the electronic device to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the method as described in any one of claims 1 to 8.

Images