Gate hoist fault prediction method and system based on data analysis

Abstract

The present invention relates to the technical field of gate hoist fault prediction, and in particular to a gate hoist fault prediction method and system based on data analysis. The method comprises the following steps: collecting the vibration signal of the gate hoist, performing segmented regularization and outlier correction, analyzing the opening and closing timing, judging the normal opening and closing cycle based on the timing, evolving a healthy opening and closing trajectory, performing normal opening and closing simulation to generate simulation data, analyzing the wear of the gate movement, performing state estimation on the wear data according to the length of use, determining the gap between the ideal gate state and the current state, performing fault backtracking and analysis based on the gap to identify abnormal parameters, and using a random forest model to perform fault prediction. The present invention realizes active monitoring and early warning of the health status of the equipment, improves the intelligent level of equipment management, reduces the impact of human factors on equipment maintenance, and promotes the safety of the overall operation of the water conservancy project.

Description

Technical Field

[0001] The present invention relates to the technical field of gate hoist fault prediction, and in particular to a gate hoist fault prediction method and system based on data analysis. Background Art

[0002] Traditional gate hoist fault monitoring methods often rely on manual maintenance and experience-based judgment, resulting in insufficient real-time control of equipment status. Traditional methods are difficult to cope with complex and dynamic working environments, thus greatly increasing the risk of potential equipment failures, leading to soaring management costs and unstable equipment operation, seriously affecting the safety and efficiency of the project. At the same time, existing vibration monitoring technologies often have signal noise interference, inability to effectively identify abnormal values, and lack of systematic analysis methods in actual applications. The monitoring signal is easily affected by environmental noise interference, resulting in unstable signal quality, and insufficient ability to identify abnormal values, making it difficult to timely and accurately judge sudden abnormal situations. Many monitoring systems lack real-time performance and are unable to detect abnormal conditions of equipment during operation in a timely manner. They cannot provide comprehensive health status assessments, which brings hidden dangers to the safe use of gates. Traditional methods are insufficient in adaptability and accuracy, resulting in many potential problems not being effectively dealt with in a timely manner. Summary of the Invention

[0003] Based on this, it is necessary to provide a gate hoist fault prediction method and system based on data analysis to solve at least one of the above technical problems.

[0004] To achieve the above objectives, a gate hoist fault prediction method based on data analysis includes the following steps:

[0005] Step S1: collecting the vibration signal of the gate hoist; performing segmented regularization processing on the vibration signal of the gate hoist and correcting the outliers to obtain a cleaning characteristic signal; analyzing the gate opening and closing timing according to the clear characteristic signal;

[0006] Step S2: Determine the normal opening and closing cycle based on the gate opening and closing sequence; evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle;

[0007] Step S3: Perform gate normal opening and closing simulation according to the gate healthy opening and closing trajectory to obtain simulated gate normal opening and closing data; analyze gate motion wear data based on the simulated gate normal opening and closing data;

[0008] Step S4: Obtain gate status data; estimate gate status based on gate movement wear data according to a preset gate usage time to generate an ideal gate status; determine the gate status difference based on the gate status data and the ideal gate status;

[0009] Step S5: Fault backtracking is performed based on the gate status difference, and the hoist transmission gear fault data is analyzed based on the backtracked fault data; abnormal parameters of the gate hoist are identified based on the hoist transmission gear fault data;

[0010] Step S6: Based on the random forest model, the gate hoist abnormal parameters are used to predict the gate hoist failure to generate predicted gate hoist failure data.

[0011] The present invention collects the vibration signal of the gate opening and closing machine and monitors the operating status of the equipment in real time, making the health assessment of the gate opening and closing machine more reliable. The segmented regularization processing and outlier correction effectively improve the signal quality and ensure the accuracy of subsequent analysis. The extraction of cleaning characteristic signals provides a solid data foundation for accurately identifying the gate opening and closing timing. The analysis based on clear characteristic signals enables timely discovery of potential fault signs, thus creating conditions for maintenance work. The judgment of regular opening and closing cycles based on the gate opening and closing timing helps to build a healthy opening and closing trajectory and provides a standard basis for the health management of the equipment. The evolved healthy opening and closing trajectory lays a solid foundation for subsequent fault prediction. The simulation data generated by the normal opening and closing simulation of the gate helps to understand the performance of the equipment under normal operating conditions. The analysis of the gate movement wear data enables better identification of potential performance degradation problems. The function of obtaining the current status data of the gate ensures a comprehensive understanding of the actual status of the equipment. The preset usage time is used for status estimation. The generated ideal state provides comparison parameters for the actual situation. Based on the comparison between the ideal state and the current data, the status gap of the equipment can be clarified, and faults can be predicted. The random forest model enables intelligent prediction of abnormal parameters of the gate hoist. The generated prediction data provides a reliable basis for technical personnel, effectively reducing the risk of equipment failure and significantly improving the stability and reliability of the equipment. By integrating signal monitoring, data analysis and fault prediction, active monitoring and early warning of equipment health status are achieved, the intelligent level of equipment management is improved, the impact of human factors on equipment maintenance is reduced, and the overall safety of water conservancy projects is promoted. A new data-driven management model has been established for the industry, which has improved the operating efficiency and lifespan of equipment, promoted the in-depth application and expansion of intelligent technology in the water conservancy industry, and provided important technical support for the scientific and effective management of equipment in the future. Ultimately, dynamic adaptation and optimization of equipment health management in complex environments have been achieved, forming a sustainable intelligent monitoring system.

[0012] The present invention also provides a gate hoist fault prediction system based on data analysis, which is used to execute the gate hoist fault prediction method based on data analysis as described above. The gate hoist fault prediction system based on data analysis includes:

[0013] The signal cleaning module is used to collect the vibration signal of the gate hoist; perform segmented regularization on the vibration signal of the gate hoist and correct outliers to obtain a clean feature signal; and analyze the gate opening and closing timing based on the clear feature signal;

[0014] The cycle determination module is used to determine the normal opening and closing cycle based on the gate opening and closing sequence; and to evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle;

[0015] The opening and closing simulation module is used to simulate the normal opening and closing of the gate according to the gate's healthy opening and closing trajectory to obtain simulated gate normal opening and closing data; and analyze the gate movement wear data based on the simulated gate normal opening and closing data;

[0016] The status assessment module is used to obtain the gate status data; estimate the gate status based on the gate movement wear data according to the preset gate usage time and generate the ideal gate status; determine the gate status gap based on the gate status data and the ideal gate status;

[0017] The fault backtracking module is used to trace back faults based on gate status differences and analyze the gate hoist transmission gear fault data based on the backtracked fault data; it also identifies abnormal parameters of the gate hoist based on the gate hoist transmission gear fault data;

[0018] The fault prediction module is used to predict the gate hoist fault based on the abnormal parameters of the gate hoist based on the random forest model to generate predicted gate hoist fault data.

[0019] Through the application of the signal cleaning module, the present invention can accurately collect and process the vibration signal of the gate hoist, ensuring that the obtained signal data has high reliability. The segmented regularization processing and outlier correction significantly improve the signal quality, providing an accurate basis for subsequent analysis. The extraction of the cleaning characteristic signal can provide the necessary data support for the effective analysis of the gate opening and closing timing. The introduction of the cycle determination module can scientifically judge the conventional opening and closing cycle, effectively evolve the healthy opening and closing trajectory, and provide a standard basis for the health management of the equipment. The simulation data generated by the opening and closing simulation module effectively reflects the characteristics of the equipment under normal operating conditions, which helps to deeply analyze the gate movement wear data and then identify and solve potential wear problems. The state assessment module provides a comprehensive perspective on the equipment status by obtaining the current status data of the gate. When performing state estimation, the preset usage time is combined with the wear data to generate the ideal gate state, so that the gap between the ideal state and the current state becomes quantifiable. The fault backtracking module can trace historical faults and extract key data by analyzing the state gap, promote in-depth analysis of the gate hoist transmission gear fault, and identify the gate hoist. The abnormal parameters of the fault prediction module provide strong support for subsequent fault prediction. The fault prediction module is based on the application of data analysis, which enables highly accurate fault prediction of abnormal parameters, thereby generating reliable fault prediction data, providing a scientific basis for preventive maintenance of equipment, significantly reducing the risk of sudden failures, and improving the overall reliability and safety of the equipment. It integrates various links of signal monitoring, data analysis and fault prediction, realizes active monitoring and early warning of equipment health status, promotes the intelligent level of equipment management, reduces the interference of human factors on equipment maintenance, enhances the predictability of equipment failures, and improves the efficiency of maintenance work. Through the data-driven management model, it lays the foundation for the safe and stable operation of water conservancy projects, not only improves the operating efficiency of equipment and the scientific nature of life cycle management, but also promotes the widespread application of intelligent technology in the water conservancy industry, forming an effective and sustainable intelligent fault early warning mechanism, providing valuable experience and foundation for future technological innovation and application promotion of equipment management, promoting the process of modernization of equipment management in the water conservancy field, and ultimately realizing dynamic adaptation and optimization of equipment health management in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention is a schematic diagram of the steps of a gate hoist fault prediction method based on data analysis;

[0021] Figure 2 Detailed implementation flow chart of step S2;

[0022] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0023] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0024] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0025] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0026] To achieve this, please refer to Figures 1 to 2 A gate hoist fault prediction method based on data analysis includes the following steps:

[0027] Step S1: collecting the vibration signal of the gate hoist; performing segmented regularization processing on the vibration signal of the gate hoist and correcting the outliers to obtain a cleaning characteristic signal; analyzing the gate opening and closing timing according to the clear characteristic signal;

[0028] Step S2: Determine the normal opening and closing cycle based on the gate opening and closing sequence; evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle;

[0029] Step S3: Perform gate normal opening and closing simulation according to the gate healthy opening and closing trajectory to obtain simulated gate normal opening and closing data; analyze gate motion wear data based on the simulated gate normal opening and closing data;

[0030] Step S4: Obtain gate status data; estimate gate status based on gate movement wear data according to a preset gate usage time to generate an ideal gate status; determine the gate status difference based on the gate status data and the ideal gate status;

[0031] Step S5: Fault backtracking is performed based on the gate status difference, and the hoist transmission gear fault data is analyzed based on the backtracked fault data; abnormal parameters of the gate hoist are identified based on the hoist transmission gear fault data;

[0032] Step S6: Based on the random forest model, the gate hoist abnormal parameters are used to predict the gate hoist failure to generate predicted gate hoist failure data.

[0033] The present invention collects the vibration signal of the gate opening and closing machine and monitors the operating status of the equipment in real time, making the health assessment of the gate opening and closing machine more reliable. The segmented regularization processing and outlier correction effectively improve the signal quality and ensure the accuracy of subsequent analysis. The extraction of cleaning characteristic signals provides a solid data foundation for accurately identifying the gate opening and closing timing. The analysis based on clear characteristic signals enables timely discovery of potential fault signs, thus creating conditions for maintenance work. The judgment of regular opening and closing cycles based on the gate opening and closing timing helps to build a healthy opening and closing trajectory and provides a standard basis for the health management of the equipment. The evolved healthy opening and closing trajectory lays a solid foundation for subsequent fault prediction. The simulation data generated by the normal opening and closing simulation of the gate helps to understand the performance of the equipment under normal operating conditions. The analysis of the gate movement wear data enables better identification of potential performance degradation problems. The function of obtaining the current status data of the gate ensures a comprehensive understanding of the actual status of the equipment. The preset usage time is used for status estimation. The generated ideal state provides comparison parameters for the actual situation. Based on the comparison between the ideal state and the current data, the status gap of the equipment can be clarified, and faults can be predicted. The random forest model enables intelligent prediction of abnormal parameters of the gate hoist. The generated prediction data provides a reliable basis for technical personnel, effectively reducing the risk of equipment failure and significantly improving the stability and reliability of the equipment. By integrating signal monitoring, data analysis and fault prediction, active monitoring and early warning of equipment health status are achieved, the intelligent level of equipment management is improved, the impact of human factors on equipment maintenance is reduced, and the overall safety of water conservancy projects is promoted. A new data-driven management model has been established for the industry, which has improved the operating efficiency and lifespan of equipment, promoted the in-depth application and expansion of intelligent technology in the water conservancy industry, and provided important technical support for the scientific and effective management of equipment in the future. Ultimately, dynamic adaptation and optimization of equipment health management in complex environments have been achieved, forming a sustainable intelligent monitoring system.

[0034] In an embodiment of the present invention, the gate hoist fault prediction method based on data analysis includes the following steps:

[0035] Step S1: collecting the vibration signal of the gate hoist; performing segmented regularization processing on the vibration signal of the gate hoist and correcting the outliers to obtain a cleaning characteristic signal; analyzing the gate opening and closing timing according to the clear characteristic signal;

[0036] In this embodiment, a dynamic signal test and analysis system with model DH5922 is used, the sampling frequency is set to 5120Hz, and the three-axis acceleration sensor model is PCB-356A16, which are respectively arranged on the main shaft box, motor base and reducer housing surface of the gate opening and closing machine. The three-way vibration signal is synchronously collected by the synchronous signal acquisition module. After the acquisition is completed, the original vibration signal is segmented using the built-in NumPy library of the Python language. The length of each segment is set to 1024 sampling points. The interval standardization method is used to regularize the signal data of each segment. The specific method is to subtract the mean of all points in each signal segment from the mean of the signal segment and divide it by the standard deviation to obtain a regularized signal sequence with a mean of 0 and a variance of 1. The outlier correction is performed by 3 Principle execution, calculate the mean of the regularized signal of each segment and standard deviation , will exceed the interval [ , ] is replaced by the median value in the signal segment, and finally the cleaning characteristic signal is formed. The find_peaks function in the scipy library is used to extract the main peak points in the working cycle of the gate hoist, and the time intervals between adjacent peak points are calculated. The integrity of the gate hoist's opening and closing timing is determined based on the interval stability.

[0037] Step S2: Determine the normal opening and closing cycle based on the gate opening and closing sequence; evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle;

[0038] In this embodiment, after the opening and closing timing is determined, based on the opening and closing cycle analysis, the average cycle value T_avg of all opening and closing cycles is calculated, and the judgment threshold is set to T_avg±5%T_avg. The cycle data within this threshold range is screened out as the normal opening and closing cycle. According to this normal opening and closing cycle, the opening and closing displacement change curve of the gate hoist is time-normalized, and the opening and closing data in different cycles are normalized to the same time axis using the cubic spline interpolation method (Cubic Spline Interpolation). The three key motion parameters of displacement, velocity, and acceleration are extracted. Finally, the healthy opening and closing trajectory curve of the gate is constructed by taking the mean of each normalized cycle curve. The specific numerical processing is based on the interp1d function in the SciPy library to complete the interpolation, and the numpy.mean function is used to calculate the mean. Each time node of the healthy trajectory contains three parameters: normalized time t, displacement s(t), velocity v(t), and acceleration a(t). The number of sampling points is set to 1000 points.

[0039] Step S3: Perform gate normal opening and closing simulation according to the gate healthy opening and closing trajectory to obtain simulated gate normal opening and closing data; analyze gate motion wear data based on the simulated gate normal opening and closing data;

[0040] In this embodiment, according to the generated healthy opening and closing trajectory, the MATLAB Simulink platform is called to construct a kinematic simulation model of the gate hoist. The model input uses the displacement-time data in the healthy trajectory, the simulation step is set to 0.001s, the motor speed range is set to 500-900rpm, and the reduction ratio is set to 50. Based on the simulation model, the driving torque, pulley rotational inertia, gate mass and friction resistance change data during the gate opening and closing process are output, and the normal opening and closing data of the simulated gate are sorted and output. The SimscapeMultibody module in MATLAB is used to simulate the pulley mechanism and the gate movement. The driving torque data output by the simulation is matched with the healthy trajectory speed curve, the energy consumed per unit displacement is calculated, and the energy consumption fluctuation data during the entire gate opening and closing process is recorded. The wear trend is further statistically analyzed, and the motion wear data is obtained by taking the average of the difference between the energy consumption data of multiple consecutive cycles. The unit of energy consumption is set to J / m.

[0041] Step S4: Obtain gate status data; estimate gate status based on gate movement wear data according to a preset gate usage time to generate an ideal gate status; determine the gate status difference based on the gate status data and the ideal gate status;

[0042] In this embodiment, during the stage of obtaining the current status data of the gate, a laser displacement sensor model KEYENCE LK-G5000 is used to measure the displacement curve of the gate opening and closing process. The displacement sampling frequency is set to 500Hz. At the same time, a vibration sensor is used to collect the vibration signal of the main shaft of the gate hoist, and the energy consumption and displacement data of the entire opening and closing process are recorded. The current status data is aligned point by point with the simulated healthy opening and closing data. The motion wear data is corrected in sections based on the cumulative use time of the gate. The service life data of the gate hoist is exported from the monitoring and management system database. The wear correction coefficient K_w=(T_u / T_d), T_u is the current cumulative use time of the gate, and T_d is the designed use time. The corrected wear data is used to re-estimate the ideal gate state curve, and the ideal opening and closing state displacement-time curve is subtracted from the current actual opening and closing curve to generate gate state gap data. A difference exceeding 3mm is recorded as an abnormal point, and the corresponding time period is marked.

[0043] Step S5: Fault backtracking is performed based on the gate status difference, and the hoist transmission gear fault data is analyzed based on the backtracked fault data; abnormal parameters of the gate hoist are identified based on the hoist transmission gear fault data;

[0044] In this embodiment, the gate status gap data is used to trace back the vibration signals of the previous 30 opening and closing cycles. Based on the frequency analysis of the vibration signal, the FFT (Fast Fourier Transform) method is used to convert the time domain signal to the frequency domain, and the amplitude peak frequency point in the frequency range of 0-500Hz is extracted to judge the frequency drift and amplitude abnormality. If the gear meshing frequency deviates from the normal 30Hz by more than ±3Hz, or the amplitude exceeds the conventional standard, it is determined that the transmission gear is abnormal. The abnormal amplitude, frequency and corresponding time are recorded, the abnormal data is analyzed, and the transmission gear fault parameters are identified. The fault parameters include gear meshing frequency, amplitude, meshing offset, and cumulative abnormal number of times. Based on the frequency-amplitude-offset characteristics, the gate opening and closing machine abnormal parameters are extracted and summarized into five columns of data: time, frequency, amplitude, offset, and abnormal category.

[0045] Step S6: Based on the random forest model, the gate hoist abnormal parameters are used to predict the gate hoist failure to generate predicted gate hoist failure data.

[0046] In this embodiment, based on the abnormal parameter data set, the RandomForestClassifier module in the Python sklearn library is called, the number of trees is set to 300, the maximum depth is 20, the random seed value is 42, the training data set is composed of the historical label data in the abnormal parameter data set, and the prediction data set is composed of the latest abnormal parameter set. The model training is performed, and after completion, the prediction is output as the gate hoist fault category label and probability. The prediction result fields include fault_type, probability, and cycle_id. If the predicted probability is higher than 0.7, it is determined to be a fault state, and the predicted gate hoist fault data is saved.

[0047] Preferably, step S1 includes the following steps:

[0048] Step S11: collecting vibration signals of the gate hoist in multiple frequency bands; performing wavelet denoising on the vibration signals of the gate hoist to obtain noise-reduced vibration signals;

[0049] Step S12: performing segmented regularization processing on the noise reduction vibration signal, and performing adaptive segmentation based on a preset segmentation threshold to obtain a segmented vibration signal;

[0050] Step S13: performing discrete point detection on the segmented vibration signal and identifying abnormal points; performing median filtering on the abnormal points to obtain a cleaning feature signal;

[0051] Step S14: performing spectrum conversion on the cleaning characteristic signal and extracting the signal characteristic envelope; extracting the time domain characteristics of the signal characteristic envelope, and identifying the opening and closing time nodes based on the time domain characteristics to generate the gate opening and closing timing sequence.

[0052] In this embodiment, when collecting vibration signals of the gate hoist in multiple frequency bands, the number of frequency bands is set within the range of three to five, the lower limit frequency of the frequency band is set to fifty Hz, and the upper limit frequency of the frequency band is set to one thousand Hz. A multi-channel acceleration sensor is used to simultaneously collect vibration signals in the X-axis, Y-axis and Z-axis directions during the movement of the gate hoist. The sampling frequency is set to two thousand and forty-eight points per second, and the sampling time is set to one hundred and twenty seconds. After analog-to-digital conversion, the vibration signal forms a set of discrete vibration data, and the index is from the first point to the total number of points. After the acquisition is completed, the vibration data is denoised using the wavelet decomposition method. The wavelet basis function is set to db6 type, and the number of decomposition layers is set to five layers. The vibration signal is decomposed into multiple discrete For sub-signals of the same frequency, all high-frequency signals are removed, and only the coefficients of the lowest frequency are retained. The retained coefficients are then reconstructed to finally obtain the noise-reduced vibration signal data. The noise-reduced vibration signal is segmented according to the set length, and the length of a single segment is 4,096 points. The maximum and minimum values ​​in each segment are calculated, and all the data in the segment are linearly normalized by the difference using the maximum and minimum differences in each segment. All values ​​fall between zero and one after linear transformation. Subsequently, the normalized signal is adaptively segmented based on a pre-set segmentation threshold, and the threshold value is set to 0.6. If the numerical difference between two adjacent points exceeds the threshold, the segment is segmented at that point. Finally, the signal is automatically divided into several sub-segments, and the result is obtained. To the segmented vibration signal, the average value and standard deviation of the segmented vibration signal are calculated segment by segment, and the abnormal point judgment threshold is set to three times the standard deviation. If the value of a point deviates from the average value in the current segment by more than this threshold, it will be judged as an abnormal point, and the specific location of the abnormal point is recorded. For all detected abnormal points, the median filter replacement method is used to replace the abnormal point with the median value of the three points before and after the point, after removing the abnormal point. After the replacement is completed, the cleaned characteristic signal is obtained, and the cleaned characteristic signal is subjected to spectrum transformation. The time domain signal is converted into a frequency domain signal using fast Fourier transform. The frequency domain range is zero to half of the sampling frequency. The frequency resolution is determined by dividing the sampling frequency by the number of sampling points in each segment. The maximum amplitude in the spectrum corresponds to the frequency and its corresponding signal envelope. The spectrum signal envelope is obtained using the Hilbert transform method, and then the envelope signal is inversely transformed back to the time domain to obtain the envelope time domain signal. The mean, root mean square value, maximum value, skewness and kurtosis of the signal are extracted, and the opening and closing action feature recognition criterion is set. When a certain instantaneous value of the envelope signal is higher than the mean value of the signal plus twice the root mean square value, the time position of the point is recorded. If several consecutive points meet the condition, the time of the first point that meets the condition is regarded as the starting point of opening and closing, and the first point that does not meet the condition is regarded as the end point of opening and closing. All opening and closing time points are converted to seconds according to the sampling frequency to obtain the time start and end sequence of the opening and closing action of the gate machine.

[0053] Preferably, step S2 includes the following steps:

[0054] Step S21: performing time series clustering on the gate opening and closing timing, and identifying the gate opening and closing mode based on the clustered time series;

[0055] Step S22: performing high-frequency pattern statistics on the gate opening and closing patterns, and determining a regular opening and closing cycle based on the high-frequency patterns;

[0056] Step S23: performing spatiotemporal mapping evolution according to the conventional opening and closing cycle to obtain an opening and closing trajectory change curve; confirming the opening and closing rhythm node according to the opening and closing trajectory change curve;

[0057] Step S24: performing multi-cycle trajectory alignment on the opening and closing trajectory change curve based on the opening and closing rhythm nodes to generate a gate healthy opening and closing trajectory.

[0058] In this embodiment, when performing time series clustering on the gate opening and closing time series data, first, the opening and closing start and end time point sequence obtained in the previous stage is organized into an opening and closing cycle sequence in chronological order, and the time length of a single cycle is set as the time difference between the opening and closing start and end points. A multidimensional time series matrix is ​​constructed based on all cycle time lengths, and the distance measurement method is set as the dynamic time warping (DTW) distance. After the DTW distances between all cycle time series are calculated, a distance matrix is ​​constructed. The clustering method uses K-Means clustering, and the number of clusters is set to four categories. According to the principle of minimizing the sum of squares of the DTW distance within the class, the clustering process is iteratively executed, and the number of iterations is set to three hundred times. Clustering is stopped when the change in the DTW distance between the current and subsequent cluster centers is less than one times ten to the power of negative five. After clustering is completed, the four gate opening and closing modes of normal opening and closing, slow opening and closing, fast opening and closing, and abnormal opening and closing are identified based on the morphological characteristics of the cluster center sequence. Each mode is represented by the mean of the opening and closing start and end time of the cluster center and The average rhythm characteristic value indicates that when the high-frequency pattern statistics are performed on the identified gate opening and closing patterns, the number of patterns belonging to all opening and closing cycles is counted, and the statistical period window length is set to thirty days. 0:00 is used as the starting point every day, and the patterns belonging to each opening and closing cycle within thirty days are recorded in chronological order. According to the frequency of pattern occurrence, in each statistical period, the pattern with the highest frequency is set as the high-frequency pattern in the period. If the high-frequency pattern appears for more than twenty days, the high-frequency pattern is determined as the regular opening and closing cycle in the period, and the average opening and closing start and end time, the average opening and closing rhythm, and the opening and closing start and end point interval corresponding to the regular opening and closing cycle are recorded. Mean, all statistical values ​​retain three decimal places, and the spatiotemporal mapping evolution is performed according to the mean of the opening and closing start and end time and the rhythm mean of the regular opening and closing cycle. The length of the mapping time axis is set to be equal to the regular opening and closing cycle, and the value range of the spatial axis is zero to one. The opening and closing start and end time are evenly mapped to the time axis, and the opening and closing rhythm mean is mapped to the amplitude value on the spatial axis. The opening and closing trajectory change curve is constructed. The curve is smoothed using the spline interpolation method, and the interpolation node is set to two points per second. After the curve is completed, the inflection point position of the curve is identified according to the changing trend of the opening and closing rhythm. If the difference in the amplitude change value within the five points before and after a certain inflection point is greater than 0.2, the inflection point position is marked as opening and closing. Rhythm nodes, all rhythm nodes are recorded in the form of time points, and the rhythm node interval is measured in seconds. Based on the confirmed opening and closing rhythm nodes, the opening and closing trajectory change curve is aligned with the multi-cycle trajectory. The alignment benchmark is set to the rhythm node position in the regular opening and closing cycle. All opening and closing trajectory change curves are translated along the time axis to make the corresponding rhythm nodes coincide. The translation step accuracy is set to 0.1 second. After the translation is completed, the aligned trajectory curve group calculates the mean of the spatial axis amplitude at each moment to obtain the healthy opening and closing trajectory of the multi-cycle trajectory. The healthy trajectory is recorded with a resolution of two points per second, and the amplitude value retains four decimal places. The healthy trajectory is saved to a dedicated trajectory sequence database.This will be used for random forest model training and fault prediction data benchmark comparison.

[0059] Preferably, performing the gate normal opening and closing simulation according to the gate healthy opening and closing trajectory in step S3 includes:

[0060] Perform Fourier transform decomposition on the gate opening and closing trajectory and extract the key frequency characteristic spectrum;

[0061] Construct the opening and closing motion equations based on the key frequency characteristic spectrum;

[0062] Simulate multiple groups of gate opening and closing trajectory samples through the opening and closing motion equations;

[0063] Perform morphological screening on gate opening and closing trajectory samples, eliminate abnormal trajectories, and obtain normal trajectory samples;

[0064] Trajectory fitting is performed based on the normal trajectory samples, and the normal opening and closing of the gate is simulated to obtain the simulated normal opening and closing data of the gate.

[0065] In this embodiment, the healthy opening and closing trajectory generated in the previous stage is sorted into a one-dimensional discrete signal sequence of the opening and closing amplitude changing with time in chronological order. The sampling frequency is set to two sampling points per second, and the total length of the opening and closing cycle is set to 600 seconds, resulting in 1200 time points. The trajectory signal is input into the Fast Fourier Transform (FFT) In the FFT function, the complex amplitude and phase information in the corresponding frequency domain is obtained, the spectrum data with a frequency range of 0 Hz to 1 Hz is intercepted, and the frequency points with amplitudes greater than 20% of the peak amplitude are screened out as key frequency points. The frequency value, amplitude and phase value corresponding to the frequency point are recorded, and three decimal places are retained to form a key frequency characteristic spectrum. The number of key frequency points is limited to 10. If there are less than 10, the closest frequency points are supplemented in descending order of amplitude. When constructing the opening and closing motion equation based on the key frequency characteristic spectrum, the frequency value f, amplitude A, and phase θ of the 10 key frequency points are used as parameters. The simple harmonic vibration superposition model is used to express each frequency component in the form of A multiplied by a sine function. The vibration period T is 1 divided by the frequency value f. The vibration term time variable is 0 to 600 seconds, and the time interval is 0.5 seconds. All key frequency components are accumulated to form the opening and closing motion equation. The equation outputs the amplitude value at each time point. The amplitude value range is limited to between 0 and 1. When the amplitude exceeds the boundary, the amplitude value is truncated. Boundary value, when simulating multiple groups of gate opening and closing trajectory samples through the opening and closing motion equation, keep the frequency value f unchanged, randomly perturb the amplitude A and phase θ, the amplitude perturbation range is limited to plus or minus 5% of the original amplitude, and the phase perturbation range is limited to plus or minus π / 10 of the original phase value. Based on the parameters after random perturbation, 1000 groups of opening and closing trajectory samples are generated. The length of each trajectory is 600 seconds, with two sampling points per second. All trajectories are saved in the form of a one-dimensional array of amplitude values ​​changing with time. The trajectory data accuracy retains four decimal places and is stored in the trajectory In the sample database, the naming method is "track_number". When performing morphological screening on gate opening and closing trajectory samples, the maximum amplitude, minimum amplitude, amplitude standard deviation and amplitude change rate of each trajectory group are calculated. The amplitude change rate is the sum of the absolute values ​​of the amplitude differences between adjacent sampling points divided by the total duration of the trajectory. The screening threshold is set, and the maximum amplitude is required to be between 0.95 and 1, the minimum amplitude is required to be between 0 and 0.05, the amplitude standard deviation is required to be between 0.1 and 0.5, and the amplitude change rate is required to be between 0.003 and 0.007, the trajectory samples that do not meet any of the conditions are directly eliminated, and the trajectory samples that meet all the conditions are retained. The number of samples after screening is recorded, and the final number of retained trajectories is not less than 800. If it is less than 800, the amplitude disturbance range is adjusted to plus or minus 3% and the trajectory samples are regenerated to make up for it. When fitting the trajectory based on the normal trajectory samples, the fifth-order B-spline curve fitting method is used, and the number of fitting nodes is set to 20. The node distribution is divided into equal intervals according to the trajectory time axis. The mean amplitude of all samples at each node is taken as the fitting target point. The least squares method is used in the fitting process. The control point positions of the fitting curve are determined using a method. After fitting is complete, the root mean square error (RMS) between the fitting curve and each normal trajectory sample is calculated. The error is required to be less than 0.02. If the error exceeds the limit, the number of nodes is increased to 25 and the fitting is repeated. Finally, a fitting curve is obtained with a fitting residual that meets the requirements. This curve is used as the reference trajectory for the normal gate opening and closing simulation. Based on this fitting trajectory, simulated normal gate opening and closing data is generated at two sampling points per second within a 600-second opening and closing cycle. The amplitude value is the amplitude value at the corresponding time point on the fitting curve. All data are rounded to four decimal places and written into the normal opening and closing trajectory database.

[0066] Preferably, analyzing the gate movement wear data based on the simulated gate normal opening and closing data in step S3 includes:

[0067] Determine the opening and closing speed change rate based on the simulated gate normal opening and closing data;

[0068] Analyze gate wear changes caused by speed changes based on preset gate physical data and opening and closing speed change rates;

[0069] The accumulated wear of the gate key points is calculated by superimposing the gate wear changes to obtain the gate key point wear data;

[0070] Construct a wear distribution network based on the wear data of the gate's key points;

[0071] Construct a wear gate framework based on the wear distribution network;

[0072] Identify the kinematic friction wear in the worn gate frame and eliminate the local fluctuation data to obtain the gate kinematic wear data.

[0073] In this embodiment, when determining the opening and closing speed change rate based on simulated gate normal opening and closing data, the simulated amplitude data of two sampling points per second within a 600-second period is first used as the displacement input. The finite difference method is used to divide the displacement difference between two consecutive sampling points by the time interval of 0.5 seconds to obtain the instantaneous opening and closing speed. The same method is then used to calculate the difference between two consecutive instantaneous speed values ​​and divide it by 0.5 seconds to obtain the opening and closing speed change rate. All speed change rate data are rounded to four decimal places. The speed unit is meters per second (m / s), and the speed change rate unit is meters per second squared (m / s²). To ensure data integrity, the speed change rates of the first and last sampling points are padded with zeros to generate a speed change rate sequence with the same length as the original sampling points. This sequence is then written into the speed change rate database. When analyzing gate wear changes caused by speed changes based on preset gate physical data and the opening and closing speed change rate, the preset gate physical parameters include a gate mass of 5000 kg, a friction coefficient of 0.15, a total length of the contact surface between the gate guide rail and the gate of 5 meters, and a wear coefficient per unit area of ​​0.0001 millimeters per Newton meter (mm / Nm), based on Newton's second law F=ma, the instantaneous inertia force of each sampling point is calculated using the gate mass and velocity change rate sequence, and then the instantaneous inertia force is multiplied by the friction coefficient to obtain the friction force, and the friction force is multiplied by the displacement increment and divided by the contact surface area to calculate the wear amount. The contact surface area is fixed to 5 square meters, and the instantaneous wear amount of each sampling point in a 600-second period is finally obtained in millimeters. All wear amounts are retained to six decimal places. The cumulative wear of the gate key points is calculated by superimposing the gate wear changes. When obtaining the gate key point wear data, the preset number of key points is 20, and the key point positions are evenly spaced along the 5-meter trajectory direction. The section where each key point is located contains There are 60 sampling points, and the cumulative wear is calculated by taking the arithmetic sum of the instantaneous wear in the section. After the calculation is completed, the cumulative wear values ​​of the 20 key points in the 600-second opening and closing cycle are obtained in millimeters. All cumulative wear values ​​retain six decimal places. The cumulative wear data are organized into a one-dimensional array. When constructing a wear distribution network based on the wear data of the gate key points, the cumulative wear values ​​of the 20 key points are used as node weights. The nodes are numbered 1 to 20, and the numbering sequence increases according to the trajectory direction. Directed edges are established between adjacent nodes. The edge weight takes the absolute value of the difference between the cumulative wear of the two nodes. The edge weight unit is millimeters. The network structure is stored as an adjacency matrix with a matrix dimension of 20 rows and 20 columns. The diagonal elements are zero and the non-diagonal elements are edge weights. When constructing the wear gate framework based on the wear distribution network, the wear gate framework is defined as a spatial grid model with key point numbers as nodes and wear distribution network edge weights as connection relationships. The node position coordinates are distributed along the x-axis direction. The x coordinate of each node is the equidistant position from the start point to the end point of the gate trajectory, in meters. The y coordinate is fixed to 0, and the z coordinate is the accumulated wear value of the node. All coordinate data retain four decimal places. The grid model is constructed using a three-dimensional coordinate modeling tool. The grid edge is determined by the spatial distance of adjacent nodes and the wear edge weight. The motion friction wear in the wear gate framework is identified, and the local fluctuation data is eliminated to obtain the gate motion wear data. First, the wear is aligned according to the direction of the gate motion trajectory. The gate frame grid model was projected in a directional manner to extract a wear curve showing the z-coordinate versus x-coordinate variation. Using a sliding window averaging method with a window length set to five key points, the wear curve was smoothed. The mean wear value within each window was calculated and used to replace the original wear value at the window center to generate a smoothed wear curve. The wear gradient change rate of the smoothed curve was then calculated through differentiation. A local fluctuation threshold was set at 0.005 mm. Any segment with an absolute value of the wear gradient change rate below the threshold was identified as a local fluctuation. The wear value for that segment was uniformly corrected to the mean of the adjacent segments before and after. The final wear curve was obtained after eliminating local fluctuations. All corrected wear values ​​were retained to four decimal places and compiled into a gate motion wear data series.

[0074] It is particularly important to construct a wear gate framework based on the wear distribution network including:

[0075] Perform structural node aggregation processing on the wear distribution network data to obtain the gate wear node array;

[0076] Perform regional density reconstruction based on the gate wear node array to generate wear density partition data;

[0077] Perform gate density projection based on the wear density partition data to obtain gate wear projection data;

[0078] Perform boundary contour fitting on gate wear projection data and extract wear contour curve;

[0079] Perform three-dimensional geometric mapping on the wear contour curve to generate a three-dimensional wear surface;

[0080] The three-dimensional wear surface is calibrated based on the gate wear node array to construct the wear gate frame.

[0081] In this embodiment, when performing structural node aggregation processing on the wear distribution network data, the wear distribution sensor array is first called to evenly distribute node points of 5mm×5mm on the surface of the gate structure, and the wear depth value at each point is collected in real time. The wear depth unit is mm. The matrix data structure is used to summarize all the measurement point data in the form of coordinate numbering to form a two-dimensional coordinate matrix. The row number and column number of the matrix correspond to the spatial coordinate index of the horizontal and vertical directions of the gate respectively. The element value in the data matrix represents the wear depth value of the corresponding coordinate point. The size of the data matrix is ​​determined by the actual point density. If the point range is 1000mm×2000mm, the node array dimension is set to 200×400. When performing regional density reconstruction based on the gate wear node array, the kernel density estimation method is adopted, and the kernel function type is set to Gaussian kernel. Kernel), the bandwidth parameter is 10mm, according to the wear depth value of each node in the two-dimensional node coordinate matrix, the wear density at each point is estimated according to the spatial density formula. The wear density unit is mm⁻². After the estimation is completed, the calculation area is divided into equal grids with 50mm×50mm as the grid unit. The average density value of all nodes inside each grid unit is calculated to generate the wear density partition data matrix. The output data format is TIFF (Tagged Image File Format) image raster file, the matrix dimension is determined by the distribution area range and grid scale. If the area range is 1000mm×2000mm and the number of grid units is 20×40, the matrix dimension is 20×40. When performing gate density projection based on wear density partition data, the two-dimensional wear density partition data is first projected to a single dimension along the gate height direction. The linear superposition method is used to accumulate the corresponding density values ​​along each column of grid units to obtain the horizontal projection value. The projection unit is mm⁻¹, and the projection step is set to 50mm. A total of 20 projection values ​​are generated for a gate width of 1000mm. The projection values ​​are arranged in the order of horizontal coordinates to form a gate wear projection data vector. When performing boundary contour fitting on the gate wear projection data, a spline curve is used. The wear contour curve is obtained by fitting the wear contour curve according to the projection value and horizontal coordinate points in the gate wear projection data vector. The wear contour curve represents the trend of the wear boundary change of the gate horizontal section. When the wear contour curve is subjected to three-dimensional geometric mapping, the surface reconstruction method based on NURBS (Non-Uniform Rational B-Splines) is used. The number of vertical segments of the gate is set to 40 and the segment spacing is 50mm. The horizontal contour curve is copied and stacked at a fixed vertical interval and arranged in sequence along the vertical coordinate. The continuous three-dimensional wear surface is generated using the NURBS surface fitting tool. The control point spacing is set to 20mm.The surface accuracy error is limited to within 0.5mm. When calibrating the framework section of the three-dimensional wear surface based on the gate wear node array, the segment numbers are first set at equal intervals in the horizontal and vertical directions according to the coordinate points in the wear node array. The number of horizontal segments is 20 and the number of vertical segments is 40. Corresponding to the same coordinate position on the wear surface, the spatial segment reference point is set. Based on the coordinates of the reference point and the corresponding wear depth value, the spatial position of each segment on the surface is calibrated to complete the spatial framework section division of the three-dimensional wear surface. After calibration, the segment calibration result data table is output in CSV format. The fields include segment number, reference coordinates, wear depth, and the coordinates of the corresponding surface control point.

[0082] Preferably, the gate state estimation based on the gate movement wear data according to the preset gate usage time in step S4 includes:

[0083] Extract the gradient features of gate motion wear data and generate surface morphology gradient features;

[0084] Conduct wear attenuation analysis based on surface topography gradient characteristics and construct a wear attenuation curve;

[0085] Perform gate wear simulation based on the preset gate usage time and the wear attenuation curve to obtain simulated gate wear data;

[0086] The gate wear state is estimated based on the simulated gate wear data and the preset initial gate state to generate the ideal gate state.

[0087] In this embodiment, to extract the gradient features of gate wear data and generate a surface topography gradient feature, the collected gate wear data is first preprocessed to remove noise and outliers. The data is then smoothed using a sliding window with a window size of 10 data points. The average value of the data within each sliding window is calculated to produce a smoothed wear curve. Next, the first-order derivative (i.e., gradient) of the wear curve in the horizontal direction is calculated. This gradient reflects the rate of wear change at different locations. Using a differential method, the wear gradient value at each location is obtained, expressed in millimeters per meter (mm / m). These gradient values ​​can be used to describe the degree and trend of wear on the gate surface. After obtaining the gradient data at each location, a wear gradient feature is formed. When performing wear attenuation analysis based on the surface topography gradient feature, a fitting method is used to perform regression analysis on the acquired wear gradient data. A cubic polynomial regression method is used to fit the wear gradient attenuation curve over time. This fitting method yields a wear attenuation coefficient, which represents the temporal variation of wear, specifically how the wear attenuation rate changes over time. To analyze the attenuation in detail, the changes in all gradient data are plotted as a attenuation curve, with the gate's usage time (in hours) on the x-axis and the wear gradient on the y-axis. The attenuation curve reveals the attenuation characteristics of gate wear under different usage times. To simulate gate wear based on a preset gate usage time and the wear attenuation curve, we first set the gate's usage time to 3000 hours. The usage time is discretized into hours, and the wear changes are simulated hour by hour. Based on the previously obtained wear attenuation curve, the relationship between usage time and the wear gradient is determined. Using linear interpolation, the wear value at each moment is calculated based on the usage time. Specifically, assuming the gate's usage time is t hours, the wear value at that moment is calculated based on the attenuation curve to obtain a new wear value. The wear value at the previous moment is then added to the current moment's wear value to obtain the current cumulative wear value. This process is repeated until the preset 3000 hours are reached. This hour-by-hour simulation method allows us to analyze the wear changes of the gate over its entire usage period. When estimating gate wear status based on simulated gate wear data and a preset initial gate state, the gate is initially assumed to be in a completely new state with an initial wear value of 0 mm. During the simulation, the calculated wear data sequence is used to calculate the current gate state at each moment based on the simulated wear value. The gate state is determined by the accumulated wear value and the wear gradient, with a threshold set to distinguish the gate's operating state. For example, when the accumulated wear value exceeds a certain threshold (e.g., 5 mm), the gate is considered to be in an "excessively worn" state.By calculating the wear state of the gate hour by hour, we can obtain the state change curve of the gate during use, and finally form a wear state curve, which reflects how the wear state of the gate changes with the increase of use time.

[0088] Preferably, determining the gate status gap based on the gate current status data and the ideal gate status in step S4 includes:

[0089] Phase alignment is performed on the gate status data and the ideal gate status to obtain gate status paired data;

[0090] Apply geometric deformation tensor transformation to gate state pairing data and compare gate topology differences;

[0091] Perform gradient segmentation on gate topology differences to generate gate difference gradients;

[0092] Determine gate state development differences based on gate difference gradients;

[0093] The gate status gap is determined by integrating the gate topology differences and gate status development differences.

[0094] In this embodiment, when phase-aligning the gate status data and the ideal gate state, the actually measured gate hoist motion data is first time-aligned with the ideal gate state data. The time alignment method used is Dynamic Time Warping (DTW). This method ensures the optimal time alignment of the gate status data and the ideal gate state data by calculating the minimum matching error between the two sets of time series. During the processing, the two data sets are first pre-processed to remove outliers and noise. Then, based on the displacement or velocity data of each time step, the time series are paired using the DTW algorithm. Finally, a set of gate status data that is completely aligned with the ideal state is obtained. This set of data will serve as the basis for subsequent analysis. The DTW parameters used are set to a window size of 50 and a distance metric of Euclidean distance. When applying geometric deformation tensor transformation to the gate state paired data, first, the paired data is transformed using a three-dimensional spatial coordinate system, and the gate current state data and ideal state data are represented as three-dimensional point cloud data respectively. Then, the rigid transformation model is applied to the current state data through transformations such as rotation, translation, and scaling to match the ideal state. When applying tensor transformation, the tensor used is a tensor based on rigid transformation, that is, a transformation that keeps the object shape unchanged. The goal is to reduce the topological difference between the two. The transformation method used is principal component analysis (PCA). By calculating the covariance matrix between the data points, the most important principal component direction is extracted, and geometric transformation is applied to the paired data accordingly. Finally, the transformed data is obtained. These data reflect the geometric difference between the current state and the ideal state of the gate. The topological difference of the gate state paired data is calculated, and the preliminary topological structure difference is obtained by comparing its coordinate difference. When the gate topology difference is gradient segmented, the topology difference data is divided into grids, and the eight-directional grid cutting method is used to divide the topological structure of the gate model into multiple areas. Each area corresponds to a specific part of the gate. Then the gradient value of each area is calculated. The gradient calculation is based on the spatial coordinate difference in the area. The height difference of each grid unit is calculated, and the gradient data is generated based on these height difference values. In the specific operation process, for each area, the difference between adjacent units is calculated by the difference method to obtain the gradient value, and the gradient data of each area is generated. The gradient value of each grid unit represents the deformation or wear degree of the part. When the gate state development difference is determined based on the gate difference gradient, the gradient data is weighted and integrated with the preset weights. The weight distribution is based on the importance of each part of the gate and the degree of influence in the opening and closing process. The gradient value of each grid unit is multiplied by the corresponding weight to obtain the comprehensive difference value of the area. The difference values ​​of all areas are accumulated to obtain the overall state difference of the gate.The weight factors used are set according to the design structure of the gate and the actual workload during use. A larger weight is given to the area with a greater mechanical load on the gate, so as to more accurately calculate the impact of the wear and deformation of this part on the overall state, and finally obtain the overall gate state development difference. When integrating the gate topology difference and the gate state development difference to determine the gate state gap, the weighted summation method is used to integrate the topology difference and the state development difference. The topology difference and state development difference of each area are first weighted. The weight is determined by the importance of each part in the gate design. All difference values ​​are weighted and accumulated to obtain the overall state gap value of the gate. This gap value represents the difference between the existing gate state and the ideal state. Finally, this gap value is used as one of the input features and combined with other working parameters. It is input into the Random Forest (RF) model for further fault prediction analysis. The weight factor setting in the whole process refers to the working environment and historical fault data of the gate.

[0095] Of particular importance is the determination of gate state development differences based on gate difference gradients, including:

[0096] Extract multi-segment gradient trajectories from gated difference gradients;

[0097] Perform directional offset distribution analysis on multi-segment gradient trajectories to generate gradient offset distribution data;

[0098] Perform differential evolution block cutting based on gradient offset distribution data to obtain gradient evolution blocks;

[0099] Merge the trend axes of the gradient evolution blocks to obtain the gate evolution axis;

[0100] Perform gradient vector fitting on the gate evolution principal axis to generate the gate state gradient vector;

[0101] The evolution trend is superimposed based on the gate state gradient vector to obtain the gate state development difference.

[0102] In this embodiment, when extracting the multi-segment gradient trajectory from the gate difference gradient, the gate is first divided into 40 equidistant segments in the vertical direction based on the gate wear node array data, and each segment is 50 mm high. It is divided into 20 equidistant segments in the horizontal direction, and each segment is 50 mm wide. Relying on the two-dimensional wear depth matrix, the wear depth difference between adjacent nodes in the horizontal and vertical directions is calculated, and the absolute value of the difference is taken as the local gradient value. The unit of the local gradient value is mm. The average gradient value is calculated for all nodes in each segment to obtain the average gradient trajectory value of the corresponding segment. The trajectory value is composed of a multi-segment gradient trajectory in the order of the spatial segment number. The output matrix format is CSV. The row number represents the vertical segment number, the column number represents the horizontal segment number, and the elements in the matrix represent the average gradient value of the corresponding segment. The value range is limited to between 0mm and 5mm. All gradient values ​​exceeding the upper limit are uniformly corrected to 5mm. When performing directional offset distribution analysis on multi-segment gradient trajectories, the aforementioned gradient trajectory matrix is ​​regarded as vector field data, and the direction angle of the gradient vector between adjacent segments is calculated respectively. The direction angle calculation is based on the inverse tangent function, with the horizontal gradient value and the vertical gradient value as independent variables. The angle value is in degrees, and the value range is limited to 0° to 360°. The gradient in all segments is calculated. Direction angle statistics summary, set the direction angle interval to 10°, divide the angle value into 36 intervals, count the number of segments in each interval, generate gradient offset distribution data, the data structure is a two-dimensional table, the fields include direction interval, number of segments, and number ratio, the output data table format is CSV, limit the sum of the number of segments to be equal to the total number of segments, and the sum of the number ratio is equal to 100%. When performing differential evolution block cutting based on gradient offset distribution data, the classification thresholds of frontier evolution zone, neutral evolution zone, and lagging evolution zone are set according to the distribution characteristics of the number ratio of segments in the direction interval. The frontier evolution zone is defined as the continuous direction with a direction interval ratio higher than 5%. The neutral evolution zone is defined as a zone with a directional interval ratio between 2% and 5%, and the lagging evolution zone is defined as a zone with a directional interval ratio less than 2%. All zones within the matrix are categorized and labeled according to their directional intervals to generate a gradient evolution block partitioning matrix. The matrix format is consistent with the original gradient trajectory matrix, and the matrix element value types include F (frontier), N (neutral), and L (lagging). The output evolution block matrix format is TXT. When merging the gradient evolution blocks along the main trend axis, continuous zones within the frontier evolution block are extracted as frontier block subsets. A straight line is fitted to each subset according to the block spatial coordinate order using the least squares method, with the upper limit of the fitting residual set to 0.2mm, the subset exceeding the threshold is split into smaller subsets and fitted separately, and all the parameters of the fitting straight line principal axis equation are recorded, including the slope, intercept, and fitting block coordinate range. The principal axes with the same or similar slope directions are merged, and the slope difference threshold is set to 5°. The principal axes with a slope difference of less than 5° are merged into a gate evolution principal axis, and the principal axis number, principal axis start and end coordinates, and fitting slope are recorded. When performing gradient vector fitting on the gate evolution principal axis, the local gradient change corresponding to each segment on the principal axis is calculated based on the principal axis start and end coordinates. The local gradient change takes the gradient difference of adjacent segments and converts the change into vector form along the principal axis direction. The vector modulus is the absolute value of the change, and the direction is from the gradient increasing direction to the gradient decreasing direction. The unit is mm, and the vector sampling spacing is set. The gate state gradient vector set is generated by evenly distributing sampling points on each principal axis, calculating the corresponding vectors, and summing them up to form the gate state gradient vector set. When performing evolution trend superposition based on the gate state gradient vectors, all principal axis gradient vectors are superimposed according to their spatial position. The superposition method is the vector sum operation of vectors at the same coordinate position, accumulating the gradient vectors of each principal axis at that position. The vector superposition result is in mm. After the superposition is completed, the modulus values ​​of all superimposed vectors are normalized. The normalization range is set to 0 to 1. The normalization formula is the original modulus value minus the minimum modulus value, divided by the difference between the maximum and minimum modulus values. After normalization, a gate state development difference data matrix is ​​generated. The matrix dimension is consistent with the original wear node array, and the matrix elements represent the normalized state development difference value at each point.

[0103] Preferably, step S5 includes the following steps:

[0104] Step S51: extracting the state difference feature of the gate state difference; performing backtracking of the use process based on the state difference feature, and inferring backtracking fault data based on the backtracking use process; mapping the backtracking fault data into the hoist fault data;

[0105] Step S52: acquiring the gear distribution data of the gate hoist; performing gear stress conversion processing on the gear distribution data of the gate hoist according to the gate hoist fault data to generate the gear force distribution;

[0106] Step S53: performing force simulation based on the force distribution of the gear, wherein the simulated load range is 50N~3000N, and estimating the gear surface deformation based on the simulated gear force data;

[0107] Step S54: Gear material fatigue analysis is performed based on gear surface deformation, and the hoist transmission gear failure is predicted based on gear fatigue data and retrospective fault data. The fatigue life data range is set to 1×10³ times~1×10 7 Secondary cycle;

[0108] Step S55: performing a slip-meshing angle compensation simulation on the hoist transmission gear fault according to the preset standard transmission gear data, and recording the compensation adjustment data, wherein the module range of the standard transmission gear data is 2-8 mm, the number of teeth range is 20-100 teeth, and the slip-meshing compensation angle range is set to -5° to +5°;

[0109] Step S56: Identify abnormal parameters of the gate hoist according to the compensation adjustment data.

[0110] In this embodiment, when extracting the state difference feature of the gate state difference, the historical gate opening and closing stroke data, real-time stroke sensor data and opening and closing speed data are first normalized. After unifying the dimensions, the stroke displacement difference of adjacent time points is differentially calculated with the standard displacement curve. The calculation formula is the difference between the stroke values ​​at two adjacent moments minus the corresponding position difference of the standard stroke curve. The obtained difference sequence is the initial value of the state difference. The difference sequence is divided into an upward segment, a downward segment and a stationary segment according to the opening and closing stroke. Five statistical features, including the mean, maximum, minimum, standard deviation and peak number of the difference values ​​of each segment, are extracted. These features are then used to form a state difference feature vector. The vector distance discrimination method is based on the Euclidean distance (Euclidean distance) to identify the state difference. Distance) method is used to calculate the difference between the current state and the historical healthy state. If the difference exceeds the set threshold of 5mm, it is determined to be an abnormal state. Then, the operation records of the gate hoist, the environmental water level change curve and the operation log are back-tracked according to the abnormal section time. All data time axes are aligned through the time synchronization method. Based on the 30-minute back-track window, the abnormal values ​​of the gate resistance, abnormal displacement values ​​and abnormal speed values ​​are extracted to construct a back-track fault data set. The fault characteristics are mapped to the internal fault type code of the gate hoist using the mapping rule table. The coding rules are such as resistance exceeding the limit is F1, displacement lag is F2, and speed fluctuation is F3. The fault code is then mapped to the fault feature code of the gate hoist gear transmission, bearing, and limiter to form a gate hoist fault data set. When obtaining the gate hoist gear distribution data, the 3D laser scanner model FARO Focus is used. The S350 performs high-precision modeling of the gear plate and transmission gear shaft, with a scanning resolution set to 0.2mm. After completing point cloud data acquisition, the point cloud fitting tool is used to fit the gear addendum circle, root circle, pitch circle, and tooth surface profile into a standard gear curve. Based on the fitting results, geometric parameters such as the number of teeth, module, tooth height, addendum height, root height, tooth thickness, and tooth spacing are extracted. The module value range is set to 2mm to 8mm, and the number of teeth range is 20 to 100 teeth. Gear stress conversion is performed based on the extracted geometric parameters and the hoist fault data. The finite element static analysis module is used to apply the load conditions corresponding to each fault type to the gear meshing surface and addendum circle. The gear model material is set to 45# steel with an elastic modulus of 210GPa and a Poisson's ratio of 0.3. Set different meshing loads according to the fault data. For F1 fault type, set the load to 1500N, for F2 fault type, set the load to 2000N, and for F3 fault type, set the load to 2500N. Calculate the contact stress of each tooth surface node, generate the gear force distribution data, output the maximum stress value and stress distribution cloud map at the tooth top, tooth root, and meshing point of each gear, and perform force simulation based on the gear force distribution. Set the simulated load range from 50N to 3000N, and set the load level in increments of 50N per gear. Use ANSYS A gear-gear contact model was established using a mechanical finite element simulation platform. The gear speed was set to 30 r / min, and the simulation time was set to 60 seconds. The force distribution of the gear was calculated under different load levels. The maximum stress, average stress, and stress concentration at the addendum and root of the tooth corresponding to each load were recorded. Based on the simulation results, the maximum deformation at the addendum and root of the gear were extracted. The maximum deformation value (unit: mm) was derived using the tooth surface displacement extraction tool. The deformation value of the gear addendum under a load of 50 N was 0.002 mm, and the deformation value at the root was 0.001 mm. The deformation value of the gear addendum under a load of 3000 N was 0.420 mm, and the deformation value at the root was 0.320 mm. The deformation curves under different loads were compiled and deformation trends were fitted. The linear increase range and nonlinear mutation range of the gear surface deformation with load were inferred. The gear surface deformation data were used to perform gear material fatigue analysis. The SN curve (stress-life curve) of 45 steel was selected, and the fatigue life data range was set from 1×10³ times to 1×10. 7 The maximum stress values ​​at the tooth top, tooth root, and meshing point are substituted into the SN curve to find the corresponding fatigue life value. If the stress value exceeds the yield limit, the fatigue life is set to 1×10³ times. If the stress value is in the elastic range, the fatigue life is calculated according to the SN curve interpolation method. The gear fatigue life data is extracted. Combined with the retrospective fault data, the fatigue life range corresponding to the same fault code is matched. The fatigue life range of the F1 fault is set to 5×10 4 times up to 1×10 6 times, and the F2 failure fatigue life interval is 1×10 4 times up to 5×10 5 times, and the F3 failure fatigue life range is 1×10³ times to 1×10 4The fault type and fatigue degree of the gate hoist transmission gear are predicted based on the corresponding relationship, and the fault prediction result is output. The gate hoist transmission gear fault is simulated by slip-meshing angle compensation according to the preset standard transmission gear data. The module range of the standard transmission gear data is set to 2mm to 8mm, the number of teeth range is 20 to 100 teeth, the meshing angle is 20°, and the slip-meshing compensation angle range is set to -5° to +5°. The finite element contact analysis module is used to establish a compensated gear meshing model, and the predicted fault gear stress, deformation, and fatigue life data are input into the model. The meshing angle is adjusted every 0.5°, and the gear force distribution, tooth top deformation and meshing point contact stress after each adjustment are recorded. The contact stress is lower than the material Yield limit, and extend the fatigue life of the meshing point to a compensation angle within the target life range, record the compensation adjustment data, output the final compensation meshing angle and the corresponding adjusted gear force, deformation, and fatigue life parameters, identify the gate hoist abnormal parameters according to the compensation adjustment data, extract the force distribution, deformation value, and fatigue life parameters in the compensated gear model, compare them with the original predicted data item by item, set the abnormal judgment threshold, and mark it as an abnormal parameter when the force increase exceeds 15%, the deformation increase exceeds 10%, and the fatigue life decreases by more than 30%. According to the abnormal parameters, the number of abnormal points, abnormal type, and abnormal degree are counted to form the gate hoist abnormal parameter set, and the transmission gear abnormality, bearing abnormality, and limiter abnormality are recorded in categories.

[0111] Preferably, step S6 includes the following steps:

[0112] Step S61: Perform interval label mapping processing on the gate hoist abnormal parameters to obtain gate hoist abnormality classification data, where the abnormality classification intervals are set as slight abnormality (0-5mm), moderate abnormality (5-10mm), and severe abnormality (10-20mm);

[0113] Step S62: Perform feature vector encoding processing on the gate hoist abnormality classification data to obtain the gate hoist feature code, wherein the number of abnormality types is set to 3, and the encoding vector dimensions are fixed to [1, 0, 0], [0, 1, 0], [0, 0, 1];

[0114] Step S63: performing input adaptation conversion on the gate hoist feature code based on the random forest model to obtain random forest input data, and limiting the input data standardization range to the interval of 0 to 1;

[0115] Step S64: Perform model prediction based on the random forest input data to obtain predicted gate hoist fault data, wherein the predicted probability output range is set between 0 and 1, and a single fault event predicted probability greater than 0.6 is determined as a fault warning.

[0116] In this embodiment, when performing interval label mapping processing on the gate hoist abnormal parameters, the compensation adjustment data generated in the previous stage is first called. The compensation adjustment data includes the offset value after gear slip-meshing angle compensation, and the unit is millimeter. All compensation adjustment data are sorted according to size, and a threshold is set according to the preset abnormal classification interval. Data with a compensation offset value less than 5 mm is mapped as a slight abnormality, data with an offset value in the range of 5 mm to 10 mm is mapped as a moderate abnormality, and data with an offset value in the range of 10 mm to 20 mm is mapped as a severe abnormality. The classification labels are expressed in the form of 0, 1, and 2, corresponding to slight abnormality, moderate abnormality, and severe abnormality, respectively. After the mapping is completed, a gate hoist abnormality classification data table is formed. The data table structure includes three columns: number, original compensation adjustment data, and classification label. Each row of data maintains a one-to-one correspondence with the corresponding gate hoist equipment number. When performing feature vector encoding processing on the gate hoist abnormality classification data, the abnormality type is one-hot encoded (One-Hot) according to the classification label. Encoding), when the classification label is 0, the corresponding encoding is [1,0,0], when the classification label is 1, the corresponding encoding is [0,1,0], when the classification label is 2, the corresponding encoding is [0,0,1], all classification labels are converted into corresponding encoding vectors one by one, and a gate hoist feature encoding data set is generated. The encoding vector dimension is fixed to 3, and the position corresponds to the number of abnormal types one by one, which contains 3 categories in total. Each feature code corresponds to the original gate hoist number, forming a three-column data table of number, classification label, and feature code. When the gate hoist feature code is input adapted based on the random forest model, the gate hoist feature code data generated in the previous stage is called, and the feature code is converted into a floating point format according to the model input standard. The dimension is 3, and the numerical value range in each encoding vector is limited to 0 to 1, maintaining the One-Hot encoding characteristics. The Min-Max normalization method is used to normalize other compensation adjustment data, and the original compensation adjustment data is normalized. The normalization formula y = (x - xmin) / (xmax - xmin) is used for conversion, with xmin set to 0 and xmax set to 20, ensuring that the converted values ​​of the compensation adjustment data are between 0 and 1. The normalized compensation adjustment data is concatenated with the corresponding feature encoding data to generate random forest input data of dimension 4, arranged in the order [encoding 1, encoding 2, encoding 3, normalized compensation value]. When making model predictions based on the random forest input data, a trained random forest model is used, with the number of model trees set to 100, the maximum tree depth set to 10, the minimum number of sample splits set to 2, and the minimum number of leaf node samples set to 1. Input data is passed to the model in batches, and the model predicts faults for each set of input data. The prediction output is a predicted probability value, set between 0 and 1. Each predicted value represents the probability of a gate hoist failure. The judgment threshold is set to 0.6. If the predicted probability of a single fault event is greater than 0, the prediction probability is 0.6. The device is judged to be in a fault warning state. All prediction results are kept in correspondence with the gate hoist number. Finally, a gate hoist fault prediction data table is generated. The data table contains three columns: number, prediction probability, and fault warning state, completing the fault prediction process.

[0117] The present invention also provides a gate hoist fault prediction system based on data analysis, which is used to execute the gate hoist fault prediction method based on data analysis as described above. The gate hoist fault prediction system based on data analysis includes:

[0118] The signal cleaning module is used to collect the vibration signal of the gate hoist; perform segmented regularization on the vibration signal of the gate hoist and correct outliers to obtain a clean feature signal; and analyze the gate opening and closing timing based on the clear feature signal;

[0119] The cycle determination module is used to determine the normal opening and closing cycle based on the gate opening and closing sequence; and to evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle;

[0120] The opening and closing simulation module is used to simulate the normal opening and closing of the gate according to the gate's healthy opening and closing trajectory to obtain simulated gate normal opening and closing data; and analyze the gate movement wear data based on the simulated gate normal opening and closing data;

[0121] The status assessment module is used to obtain the gate status data; estimate the gate status based on the gate movement wear data according to the preset gate usage time and generate the ideal gate status; determine the gate status gap based on the gate status data and the ideal gate status;

[0122] The fault backtracking module is used to trace back faults based on gate status differences and analyze the gate hoist transmission gear fault data based on the backtracked fault data; it also identifies abnormal parameters of the gate hoist based on the gate hoist transmission gear fault data;

[0123] The fault prediction module is used to predict the gate hoist fault based on the abnormal parameters of the gate hoist based on the random forest model to generate predicted gate hoist fault data.

[0124] Through the application of the signal cleaning module, the present invention can accurately collect and process the vibration signal of the gate hoist, ensuring that the obtained signal data has high reliability. The segmented regularization processing and outlier correction significantly improve the signal quality, providing an accurate basis for subsequent analysis. The extraction of the cleaning characteristic signal can provide the necessary data support for the effective analysis of the gate opening and closing timing. The introduction of the cycle determination module can scientifically judge the conventional opening and closing cycle, effectively evolve the healthy opening and closing trajectory, and provide a standard basis for the health management of the equipment. The simulation data generated by the opening and closing simulation module effectively reflects the characteristics of the equipment under normal operating conditions, which helps to deeply analyze the gate movement wear data and then identify and solve potential wear problems. The state assessment module provides a comprehensive perspective on the equipment status by obtaining the current status data of the gate. When performing state estimation, the preset usage time is combined with the wear data to generate the ideal gate state, so that the gap between the ideal state and the current state becomes quantifiable. The fault backtracking module can trace historical faults and extract key data by analyzing the state gap, promote in-depth analysis of the gate hoist transmission gear fault, and identify the gate hoist. The abnormal parameters of the fault prediction module provide strong support for subsequent fault prediction. The fault prediction module is based on the application of data analysis, which enables highly accurate fault prediction of abnormal parameters, thereby generating reliable fault prediction data, providing a scientific basis for preventive maintenance of equipment, significantly reducing the risk of sudden failures, and improving the overall reliability and safety of the equipment. It integrates various links of signal monitoring, data analysis and fault prediction, realizes active monitoring and early warning of equipment health status, promotes the intelligent level of equipment management, reduces the interference of human factors on equipment maintenance, enhances the predictability of equipment failures, and improves the efficiency of maintenance work. Through the data-driven management model, it lays the foundation for the safe and stable operation of water conservancy projects, not only improves the operating efficiency of equipment and the scientific nature of life cycle management, but also promotes the widespread application of intelligent technology in the water conservancy industry, forming an effective and sustainable intelligent fault early warning mechanism, providing valuable experience and foundation for future technological innovation and application promotion of equipment management, promoting the process of modernization of equipment management in the water conservancy field, and ultimately realizing dynamic adaptation and optimization of equipment health management in complex environments.

[0125] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0126] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A gate hoist fault prediction method based on data analysis, characterized in that: The following steps are involved: Step S1: collecting the vibration signal of the gate hoist; performing segmented regularization processing on the vibration signal of the gate hoist and correcting the abnormal value to obtain a cleaning characteristic signal; Analyze gate opening and closing timing based on clear characteristic signals; Step S2: Determine the normal opening and closing cycle based on the gate opening and closing sequence; evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle; Step S3: Based on the displacement-time data in the healthy opening and closing trajectory, the gate normal opening and closing simulation is performed through the kinematic simulation model to obtain the simulated gate normal opening and closing data; The gate movement wear data is obtained based on the simulation of the normal opening and closing data of the gate, including: Determine the opening and closing speed change rate based on the simulated gate normal opening and closing data; Analyze gate wear changes caused by speed changes based on preset gate physical data and opening and closing speed change rates; The accumulated wear of the gate key points is calculated by superimposing the gate wear changes to obtain the gate key point wear data; Construct a wear distribution network based on the wear data of the gate's key points; Construct a wear gate framework based on the wear distribution network; Identify the kinematic friction wear in the worn gate frame and eliminate the local fluctuation data to obtain the gate kinematic wear data; Step S4: Obtain gate status data; estimate gate status based on gate movement wear data according to a preset gate usage time to generate an ideal gate status; determine the gate status difference based on the gate status data and the ideal gate status; Step S5: Fault backtracking is performed based on the gate state gap, and hoist transmission gear fault data is obtained based on the backtracked fault data analysis; abnormal parameters of the gate hoist are identified based on the hoist transmission gear fault data; Step S6: Based on the random forest model, the gate hoist abnormal parameters are used to predict the gate hoist failure to generate predicted gate hoist failure data.

2. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting vibration signals of the gate hoist in multiple frequency bands; performing wavelet denoising on the vibration signals of the gate hoist to obtain noise-reduced vibration signals; Step S12: performing segmented regularization processing on the noise reduction vibration signal, and performing adaptive segmentation based on a preset segmentation threshold to obtain a segmented vibration signal; Step S13: performing discrete point detection on the segmented vibration signal and identifying abnormal points; performing median filtering on the abnormal points to obtain a cleaning feature signal; Step S14: performing spectrum conversion on the cleaning characteristic signal and extracting the signal characteristic envelope; extracting the time domain characteristics of the signal characteristic envelope, and identifying the opening and closing time nodes based on the time domain characteristics to generate the gate opening and closing timing sequence.

3. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: performing time series clustering on the gate opening and closing timing, and identifying the gate opening and closing mode based on the clustered time series; Step S22: performing high-frequency pattern statistics on the gate opening and closing patterns, and determining a regular opening and closing cycle based on the high-frequency patterns; Step S23: performing spatiotemporal mapping evolution according to the conventional opening and closing cycle to obtain an opening and closing trajectory change curve; confirming the opening and closing rhythm node according to the opening and closing trajectory change curve; Step S24: performing multi-cycle trajectory alignment on the opening and closing trajectory change curve based on the opening and closing rhythm nodes to generate a gate healthy opening and closing trajectory.

4. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: The normal gate opening and closing simulation according to the gate healthy opening and closing trajectory in step S3 includes: Perform Fourier transform decomposition on the gate opening and closing trajectory and extract the key frequency characteristic spectrum; Construct the opening and closing motion equations based on the key frequency characteristic spectrum; Simulate multiple groups of gate opening and closing trajectory samples through the opening and closing motion equations; Perform morphological screening on gate opening and closing trajectory samples, eliminate abnormal trajectories, and obtain normal trajectory samples; Trajectory fitting is performed based on the normal trajectory samples, and the normal opening and closing of the gate is simulated to obtain the simulated normal opening and closing data of the gate.

5. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: The gate state estimation based on the gate movement wear data according to the preset gate usage time in step S4 includes: Extract the gradient features of gate motion wear data and generate surface morphology gradient features; Conduct wear attenuation analysis based on surface topography gradient characteristics and construct a wear attenuation curve; Perform gate wear simulation based on the preset gate usage time and the wear attenuation curve to obtain simulated gate wear data; The gate wear state is estimated based on the simulated gate wear data and the preset initial gate state to generate the ideal gate state.

6. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: Determining the gate status gap based on the gate current status data and the ideal gate status in step S4 includes: Phase alignment is performed on the gate status data and the ideal gate status to obtain gate status paired data; Apply geometric deformation tensor transformation to gate state pairing data and compare gate topology differences; Perform gradient segmentation on gate topology differences to generate gate difference gradients; Determine gate state development differences based on gate difference gradients; The gate status gap is determined by integrating the gate topology differences and gate status development differences.

7. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: Step S5 includes the following steps: Step S51: extracting the state difference feature of the gate state difference; performing backtracking of the use process based on the state difference feature, and inferring backtracking fault data based on the backtracking use process; mapping the backtracking fault data into the hoist fault data; Step S52: acquiring the gear distribution data of the gate hoist; performing gear stress conversion processing on the gear distribution data of the gate hoist according to the gate hoist fault data to generate the gear force distribution; Step S53: performing force simulation based on the force distribution of the gear, wherein the simulated load range is 50N~3000N, and estimating the gear surface deformation based on the simulated gear force data; Step S54: Gear material fatigue analysis is performed based on gear surface deformation, and the hoist transmission gear failure is predicted based on gear fatigue data and retrospective fault data. The fatigue life data range is set to 1×10³ times~1×10 7 Secondary cycle; Step S55: performing a slip-meshing angle compensation simulation on the hoist transmission gear fault according to the preset standard transmission gear data, and recording the compensation adjustment data, wherein the module range of the standard transmission gear data is 2-8 mm, the number of teeth range is 20-100 teeth, and the slip-meshing compensation angle range is set to -5° to +5°; Step S56: Identify abnormal parameters of the gate hoist according to the compensation adjustment data.

8. The gate hoist fault prediction method based on data analysis according to claim 1 is characterized in that: Step S6 includes the following steps: Step S61: Perform interval label mapping processing on the gate hoist abnormal parameters to obtain gate hoist abnormality classification data, where the abnormality classification intervals are set as slight abnormality (0-5mm), moderate abnormality (5-10mm), and severe abnormality (10-20mm); Step S62: Perform feature vector encoding processing on the gate hoist abnormality classification data to obtain the gate hoist feature code, wherein the number of abnormality types is set to 3, and the encoding vector dimensions are fixed to [1, 0, 0], [0, 1, 0], [0, 0, 1]; Step S63: performing input adaptation conversion on the gate hoist feature code based on the random forest model to obtain random forest input data, and limiting the input data standardization range to the interval of 0 to 1; Step S64: Perform model prediction based on the random forest input data to obtain predicted gate hoist fault data, wherein the predicted probability output range is set between 0 and 1, and a single fault event predicted probability greater than 0.6 is determined as a fault warning.

9. A gate hoist fault prediction system based on data analysis, characterized in that: For executing the gate hoist fault prediction method based on data analysis according to claim 1, the gate hoist fault prediction system based on data analysis comprises: The signal cleaning module is used to collect the vibration signal of the gate hoist; perform segmented regularization on the vibration signal of the gate hoist and correct outliers to obtain a clean feature signal; and analyze the gate opening and closing timing based on the clear feature signal; The cycle determination module is used to determine the normal opening and closing cycle based on the gate opening and closing sequence; and to evolve the gate healthy opening and closing trajectory according to the normal opening and closing cycle; The opening and closing simulation module is used to simulate the normal opening and closing of the gate according to the gate's healthy opening and closing trajectory to obtain the simulated normal opening and closing data of the gate; the gate movement wear data is obtained based on the analysis of the simulated normal opening and closing data of the gate; The status assessment module is used to obtain the gate status data; estimate the gate status based on the gate movement wear data according to the preset gate usage time and generate the ideal gate status; determine the gate status gap based on the gate status data and the ideal gate status; The fault backtracking module is used to backtrack faults based on the gate status gap and obtain the gate hoist transmission gear fault data based on the backtracked fault data; and identify the gate hoist abnormal parameters based on the gate hoist transmission gear fault data; The fault prediction module is used to predict the gate hoist fault based on the abnormal parameters of the gate hoist based on the random forest model to generate predicted gate hoist fault data.

Images