Intelligent cell gate fault early warning system based on combinatorial algorithm

Abstract

The invention discloses an intelligent cell gate fault early warning system based on a combinatorial algorithm, and relates to the technical field of data processing, and the system comprises a standard periodic signal determination module which is used for interacting with a power transmission module of an intelligent cell gate and determining a standard periodic signal; the periodic reference signal determination module is used for identifying and identifying a signal inflection point window aiming at a standard periodic signal and determining a periodic reference signal; the fault decision training module is used for supervising and training the fault decision module; and the fault detection module is used for performing periodic matching judgment and fault trend prediction based on the signal inflection point window. The technical problems that a traditional intelligent unit gate fault detection method in the prior art is lagged in response, insufficient in precision and poor in multi-scene environment adaptability are solved, the real-time performance and precision of fault detection are improved through periodic signal matching, fault trend prediction and transfer learning, and the fault detection efficiency is improved. And meanwhile, the adaptability of the multi-scene environment is improved.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to an intelligent unit door fault early warning system based on a combination algorithm. Background Art

[0002] With the widespread application of intelligent equipment, intelligent unit doors have become an indispensable part of building automation. However, its core power transmission modules (such as motors, reducers, hanging wheels, belts) are easily affected by changes in environmental temperature and humidity, material aging and long-term use during operation, resulting in performance degradation and failure. Traditional fault detection methods mostly rely on regular manual inspections or simple threshold alarms, which have problems such as detection lag, insufficient accuracy and difficulty in predicting potential faults. In addition, different installation environments (such as regional climate differences) and equipment characteristics (such as assembly process differences) make the adaptability of fault detection systems a major technical challenge. Summary of the invention

[0003] The present application provides an intelligent unit door fault warning system based on a combined algorithm, which is used to solve the technical problems of delayed response, insufficient accuracy and poor adaptability to multiple scenario environments of traditional intelligent unit door fault detection methods in the prior art.

[0004] The present application provides a smart unit door fault warning system based on a combination algorithm, the system comprising: a standard periodic signal determination module, used for interacting with a power transmission module of the smart unit door, determining a standard periodic signal based on periodic transmission characteristics, wherein the power transmission module at least includes a motor, a reducer, a hanging wheel and a belt, and the standard periodic signal is determined based on a detection signal, and the detection signal at least includes a sound signal; a periodic reference signal determination module, used for identifying and labeling a signal inflection point window for the standard periodic signal, and determining a periodic reference signal, wherein the signal inflection point window is determined by a preset window size centered on the signal inflection point; a fault decision training module, used for supervised training of a fault decision module based on the periodic reference signal, wherein the fault decision module is constructed through transfer learning; a fault detection module, used for receiving a real-time detection signal of the smart unit door, and combining with the fault decision module, performing periodic matching judgment and fault trend prediction based on the signal inflection point window to determine fault detection data, wherein the fault detection data includes real-time faults and predicted faults.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages: The intelligent unit door fault warning system based on the combined algorithm provided by the present application relates to the field of data processing technology. By extracting the standard periodic signal of the power transmission module, generating a periodic reference signal, supervising the training of the fault decision module, and enhancing the adaptability through transfer learning. The real-time detection module combines the decision model to perform period matching and trend prediction, generates real-time fault and predicted fault data, and realizes accurate fault identification and warning of the intelligent unit door, which solves the technical problems of the traditional intelligent unit door fault detection method in the prior art, such as delayed response, insufficient accuracy, and poor adaptability to multiple scene environments, and realizes the technical effect of improving the real-time and accuracy of fault detection through periodic signal matching, fault trend prediction and transfer learning, while improving the adaptability to multiple scene environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0007] Figure 1 A schematic diagram of the structure of an intelligent unit door fault warning system based on a combined algorithm provided in an embodiment of the present application; Figure 2 A schematic diagram of a flow chart of supervised training of a fault decision training module in a smart unit door fault warning system based on a combination algorithm provided in an embodiment of the present application.

[0008] Explanation of the reference numerals: standard periodic signal determination module 11 , periodic reference signal determination module 12 , fault decision training module 13 , fault detection module 14 . DETAILED DESCRIPTION

[0009] The present application provides an intelligent unit door fault warning system based on a combined algorithm, which is used to solve the technical problems of delayed response, insufficient accuracy and poor adaptability to multiple scenario environments of traditional intelligent unit door fault detection methods in the prior art.

[0010] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0011] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices.

[0012] Embodiment 1, as Figure 1 As shown, the present application provides an intelligent unit door fault warning system based on a combined algorithm, the system comprising: The standard periodic signal determination module 11 is used for the power transmission module of the interactive intelligent unit door to determine the standard periodic signal based on the periodic transmission characteristics, wherein the power transmission module at least includes a motor, a reducer, a hanging wheel and a belt, and the standard periodic signal is determined based on a detection signal, and the detection signal at least includes a sound signal.

[0013] Optionally, the standard periodic signal determination module 11 of the present application is responsible for generating a stable standard periodic signal for the operating characteristics of the power transmission module of the intelligent unit door, providing a reliable benchmark for subsequent fault detection and early warning. Among them, the power transmission module, as the core component of the unit door, usually includes a motor, a reducer, a hanging wheel and a belt, which complete the opening and closing action of the door body through coordinated operation. When the equipment is operating normally, the periodic mechanical movement of these components will generate regular physical signals, among which the sound signal is the key detection signal, because the sound can directly reflect the working state of the mechanical components and their power transmission characteristics.

[0014] Specifically, we first use a highly sensitive acoustic sensor to collect the sound signals of the power transmission module in real time during operation. These sound signals are a comprehensive reflection of the mechanical vibration, friction and power transmission during the operation of the equipment. However, the original signal is often mixed with environmental noise (such as wind noise, human interference) and irrelevant background sound. Therefore, the signal is purified through noise filtering to remove interference signals, and smoothing is used to improve the stability of the signal.

[0015] The collected sound signal is then converted into frequency domain data through fast Fourier transform (FFT) technology to analyze the main frequency and harmonic characteristics of the signal. The periodic characteristics of the power transmission module are reflected in these frequency components. For example, the rotation speed of the motor will form a significant main frequency peak in the spectrum, while the movement of the hanging wheel and the belt will superimpose the characteristic signals of lower frequencies. By combining spectrum analysis and time domain signals, the periodic transmission characteristics of the power transmission module can be accurately captured.

[0016] Based on the extraction of periodic characteristics, a standard periodic signal model is established through multiple sampling and data statistics. This model not only includes the average value and frequency distribution of the signal, but also the dynamic range of the signal (such as the fluctuation of the amplitude). In order to adapt to the influence of different regional environments (such as temperature and humidity changes, aging effects), the module designs a dynamic calibration mechanism to adjust the standard signal model in real time to ensure its stability and adaptability in multiple scenarios.

[0017] To ensure the high reliability of the generated standard periodic signal, it is compared with the historical signal database and the periodic consistency of the signal is verified by the dynamic time warping (DTW) algorithm. If the signal deviation exceeds the set threshold, the module will automatically recalibrate, optimize the characteristic parameters of the standard periodic signal, and finally generate a reference signal suitable for the current device status.

[0018] Through the above process, the standard periodic signal generated by the standard periodic signal determination module 11 not only reflects the normal operating characteristics of the power transmission module, but also has strong environmental adaptability and robustness. As a comparison benchmark for the health status of the equipment, this signal will play a key role in subsequent fault detection and trend prediction.

[0019] The periodic reference signal determination module 12 is used to identify and mark a signal inflection point window for the standard periodic signal, and determine a periodic reference signal, wherein the signal inflection point window is determined by a preset window size centered on the signal inflection point.

[0020] Furthermore, the periodic reference signal determination module 12 is further configured to perform the following steps: P21: Determine the signal spectrum of the standard periodic signal and identify the signal inflection point, wherein the signal inflection point is marked with a periodic timing; P22: Based on the preset window size, perform a signal window interception on the signal inflection point and determine the signal inflection point window; P23: For the signal inflection point window, determine the inflection point feature vector, mark the standard periodic signal, and determine the periodic reference signal.

[0021] It should be understood that the main function of the cycle reference signal determination module 12 of the present application is to identify key characteristic inflection points, extract inflection point windows, and generate complete cycle reference signals based on standard cycle signals. The cycle reference signal is used to characterize the operating characteristics of the device completing a complete gating cycle, providing a benchmark for fault detection and trend prediction.

[0022] Specifically, the signal spectrum of the standard periodic signal is first constructed, usually in the form of a waveform graph or a spectrum graph to show the changing trend of the signal in the time domain or frequency domain. Through spectrum analysis, the points with significant changes in the signal, that is, the signal inflection points, can be intuitively identified. Signal inflection points refer to extreme points (such as peaks and valleys) or trend change points (such as slope mutation points) that appear in the signal waveform during operation. These inflection points reflect important characteristic information in the equipment operation cycle. A combination of extreme value detection and first-order derivative analysis can be used to accurately locate the inflection points in the signal, and a periodic timing label can be assigned to each inflection point to mark the specific position of the inflection point in the periodic signal. The generation of signal spectrum and inflection point identification lay the foundation for subsequent signal analysis.

[0023] After the inflection point is identified, the module takes each signal inflection point as the center and intercepts the signal according to the preset window size to form a signal inflection point window. The signal inflection point window is a small range of signal data with the inflection point as the core, which can retain the local features near the inflection point to ensure the accuracy of the analysis. The setting of the window size depends on the operating characteristics of the equipment and the sampling frequency of the signal, and usually needs to take into account both analysis accuracy and computational efficiency. The module extracts the inflection point windows one by one through the sliding window technology to ensure that the window coverage is complete and non-overlapping, while avoiding feature loss caused by signal truncation. Through window interception, the module realizes the focus on local features in the signal and provides high-quality input data for the extraction of inflection point features.

[0024] Next, each signal inflection point window is analyzed in detail to extract feature vectors including amplitude, frequency component, slope change rate, energy distribution, etc. These inflection point feature vectors are used to quantitatively describe the local characteristics of the inflection point and are key data for generating periodic reference signals. To ensure the simplicity and representativeness of the feature vectors, the principal component analysis (PCA) technology can be introduced to reduce the dimensionality of the extracted features, retaining the core information while reducing redundancy. Next, the standard periodic signal is identified according to the extracted feature vectors, the signal is segmented and the inflection point features are matched one by one, and finally a complete periodic reference signal is generated. This reference signal can fully reflect the overall characteristics of the powertrain module completing a gating cycle during operation, providing an accurate and reliable benchmark for system fault detection and trend prediction.

[0025] Further, when determining the inflection point feature vector for the signal inflection point window, step P23 of the embodiment of the present application further includes: P23-1: For the signal inflection point window, divide and determine the first front window, inflection point and first rear window, wherein the first front window is determined based on the front boundary and inflection point position of the signal inflection point window; P23-2: Based on the first front window and the inflection point, calculate the first gradient feature vector of the signal window, wherein the first gradient feature vector includes a gradient value and a gradient direction; P23-3: Based on the inflection point and the first rear window, calculate the second gradient feature vector of the signal window; P23-4: Integrate the first gradient feature vector and the second gradient feature vector as the inflection point feature vector.

[0026] Optionally, when determining the inflection point feature vector of the signal inflection point window, the window may be further divided, gradient features may be calculated and integrated into a final inflection point feature vector to fully characterize the local variation characteristics of the signal inflection point.

[0027] First, for each signal inflection point window, it is divided into three parts: the first front window, the inflection point, and the first rear window. The first front window is the signal segment from the front boundary of the inflection point window to the inflection point position, which mainly captures the change characteristics before the inflection point; the inflection point is the key position of the signal fluctuation, usually manifested as an extreme point or a trend turning point; the first rear window is the signal segment from the inflection point position to the rear boundary of the inflection point window, which is used to analyze the change trend after the inflection point. This division method divides the local area of ​​the signal into parts with different physical meanings, so that the subsequent gradient calculation can describe the signal changes more specifically.

[0028] Next, the module calculates the first gradient feature vector of the area based on the first front window and the inflection point position. The gradient feature vector describes the rate and direction of change of the signal, and its core indicators include the gradient value and gradient direction. The gradient value is calculated by the first-order derivative, indicating the intensity of the signal change; the gradient direction is determined by the positive or negative value of the gradient value, indicating the rising or falling trend of the signal. The module uses the sliding window method to calculate the gradient features of the first front window point by point, focusing on capturing the characteristics of the faster or slower rate of change of the signal near the inflection point, thereby revealing the change pattern before the inflection point.

[0029] Furthermore, for the inflection point and the first post-window, the module calculates the second gradient eigenvector in a similar way to the calculation of the first gradient eigenvector. The second gradient eigenvector focuses on describing the change characteristics of the signal after the inflection point, and reflects the fluctuation amplitude and trend of the signal after the inflection point by comparing the rate and direction of change of the signal. The gradient characteristics before and after the inflection point can intuitively show the symmetry or asymmetry of the signal, and help analyze the significance of the inflection point as a periodic feature.

[0030] Finally, the first gradient eigenvector and the second gradient eigenvector are integrated into the final inflection point eigenvector to fully characterize the local characteristics of the inflection point. During the integration process, the module merges the front and rear gradient eigenvectors by feature splicing, and normalizes the gradient value and gradient direction to ensure the consistency of the eigenvector. In addition, in order to improve the noise resistance of the eigenvector, low-pass filtering technology can be introduced to smooth the gradient data, thereby reducing the error caused by signal fluctuations or noise. The final generated inflection point eigenvector can accurately reflect the changing trend and regional characteristics of the signal inflection point, providing accurate data support for the construction of periodic reference signals.

[0031] The fault decision training module 13 is used to supervise the training of a fault decision module based on the periodic reference signal, wherein the fault decision module is constructed through transfer learning.

[0032] Further, such as Figure 2 As shown, the fault decision training module 13 is also used to perform the following steps: P31: Call the fault records of the intelligent unit door and mine the transmission fault samples, wherein each transmission fault sample includes the real-time fault stage and the fault evolution stage; P32: Based on the transmission fault samples, perform sample supervision training to determine the fault decision model; P33: Interact with the personalized assembly information of the intelligent unit door to determine the personalized features; P34: Add an inflection point analysis layer, combine the personalized features, perform transfer learning on the fault decision model, and determine the fault decision module.

[0033] It should be understood that the fault decision training module 13 of the present application is used to train the fault decision module by supervised learning and transfer learning methods based on the periodic reference signal, so as to realize accurate fault identification and prediction of the operation status of the intelligent unit door. The module constructs a highly adaptable fault decision module by calling historical fault data, combining personalized assembly information and environmental characteristics.

[0034] Specifically, we first call the historical fault records of the intelligent unit door to mine representative transmission fault samples. The transmission fault samples cover the signal data of the power transmission module at different fault stages, which are mainly divided into real-time fault stage and fault evolution stage. The signal in the real-time fault stage reflects the characteristic changes when the fault occurs, such as abnormal sound signals or vibration signals; while the data in the fault evolution stage reveals the dynamic characteristics of the fault from germination to full manifestation. By sorting and classifying these samples, the module generates a comprehensive transmission fault database, providing high-quality training data for subsequent supervised learning.

[0035] Next, based on the transmission fault samples, the module performs supervised training to build a fault decision model. Supervised training is a machine learning method based on labeled samples. The module uses the label information (such as fault type, stage identification) in the fault samples to train the model to ensure that the model can identify fault characteristics and evolution laws. During the training process, advanced algorithms such as deep neural networks (DNN) or gradient boosting trees (GBDT) can be used, and dynamic time warping can be introduced to match and analyze periodic signals to enhance the model's ability to process time series data and generate the fault decision model. The fault decision model has a strong ability to recognize common fault modes and can provide fault judgment support for different types of transmission modules.

[0036] Next, in order to adapt to the impact of different installation environments on the operating characteristics of the transmission module, the personalized assembly information of the interactive intelligent unit door is further combined to extract personalized features. The personalized assembly information includes regional environmental differences (such as high temperature, humidity, ultraviolet exposure), material aging, and specific configuration parameters of the equipment (such as belt tension, reducer model). By analyzing this information, the module adjusts the characteristics of the standard periodic signal and extracts a set of characteristic parameters with environmental adaptability as personalized features. These personalized features enhance the local adaptability of the model and provide a basis for the precision of fault decision-making.

[0037] After completing the personalized feature extraction, an inflection point analysis layer is introduced into the fault decision model, and transfer learning is performed on the fault decision model in combination with personalized features. Transfer learning is a method of quickly training local models in a new environment by reusing existing model knowledge, which can significantly improve training efficiency. The inflection point analysis layer focuses on the key inflection point features of the signal, deeply combines the changing trend of the periodic reference signal with the personalized features, and generates a feature representation that is more sensitive to the operating status of the local equipment. Finally, through transfer learning, the module adapts the fault decision model to a fault decision module that adapts to the local installation environment and personalized assembly information, and has a strong ability to recognize local fault modes.

[0038] Through the above steps, the fault decision training module 13 realizes the complete process from learning common fault modes to localized adaptation, ensuring that the fault decision module can not only accurately identify common fault modes, but also quickly adapt to the differences in personalized environments, greatly improving the fault diagnosis and prediction capabilities of the intelligent unit door.

[0039] Furthermore, when performing sample supervision training and determining the fault decision model, step P32 of the embodiment of the present application further includes: P32-1: Based on the transmission fault samples in the real-time fault stage, supervise the training of N fault decision makers, wherein the result with the largest proportion of the N decision results is output; P32-2: Obtain the transmission operation code of the power transmission module and train the code self-tester; P32-3: Integrate the N fault decision makers as the first branch, based on the code self-tester as the second branch, run the first branch and the second branch in parallel to determine the fault identification layer; P32-4: Based on the transmission fault samples in the fault evolution stage, perform prediction training under time series to determine the evolution prediction layer; P32-5: Post-connect the evolution prediction layer to the fault identification layer to generate the fault decision model.

[0040] In a possible embodiment of the present application, in the process of sample supervised training to determine the fault decision model, this embodiment comprehensively realizes the real-time identification and evolution prediction capabilities of faults through multi-level training and branch integration.

[0041] First, based on the transmission fault samples in the real-time fault stage, N fault decision makers are trained. Each decision maker uses a different algorithm or feature processing method (such as support vector machine, random forest, convolutional neural network, etc.) to supervise the fault features in the sample. These decision makers each output independent fault judgment results. The module uses the ensemble learning method to take the result with the largest proportion as the final output, that is, the fault category that appears the most times in the N decision results is selected as the current recognition result. This method of parallel training of multiple decision makers enhances the model's robustness in identifying different fault modes and avoids the deviation or misjudgment that may be caused by a single algorithm.

[0042] Next, the transmission operation code of the power transmission module is further called, that is, the control instructions and status feedback data during the operation of the equipment, to train a special code self-checker. The code self-checker identifies potential operation anomalies or code errors by analyzing the code execution characteristics of the transmission module under different operating states (such as logic jumps, parameter call frequency, instruction execution time, etc.). This code-based fault analysis method provides a supplementary perspective for fault identification that is different from physical signals, which helps to fully cover possible fault sources of the equipment.

[0043] Furthermore, the N-item fault decision maker trained above is used as the first branch, and the code self-checker is used as the second branch to build a fault identification layer in parallel. The first branch focuses on fault identification at the physical level by fusing multiple algorithms of real-time signals; the second branch provides additional fault judgment basis from the control level through code execution analysis. The parallel design of the two branches can complement each other, improve the accuracy and comprehensiveness of fault identification, and ensure that the model still has high reliability under complex operating conditions.

[0044] Next, we conduct time series analysis on transmission fault samples in the fault evolution stage and train the evolution prediction layer. Time series analysis focuses on the dynamic change characteristics of the signal during the evolution process. We can use deep learning methods such as long short-term memory network (LSTM) or time convolution network (TCN) to model and predict the process from early signs to full outbreak of faults. By learning the trend characteristics of evolution samples, the evolution prediction layer can output prediction signals in advance when the fault has not yet completely occurred, thus providing a basis for preventive maintenance.

[0045] Finally, the evolution prediction layer is connected to the fault identification layer to form a complete fault decision model. The fault identification layer provides real-time fault classification results, and the evolution prediction layer outputs trend prediction signals in the time series dimension. The combination of the two realizes accurate identification and dynamic prediction of faults. Through this hierarchical structure, the large model has both high-precision recognition capabilities for the current state and can effectively cope with the long-term evolution trend of complex faults.

[0046] Through the above steps, the fault decision model of this embodiment successfully integrates multi-algorithm supervised training, code-level self-check analysis and time series prediction technology, forming a comprehensive and efficient fault identification and prediction solution, which provides strong technical support for fault warning and precise maintenance of intelligent unit doors.

[0047] The fault detection module 14 is used to receive the real-time detection signal of the intelligent unit door, and in combination with the fault decision module, perform periodic matching judgment and fault trend prediction based on the signal inflection point window to determine the fault detection data, wherein the fault detection data includes real-time faults and predicted faults.

[0048] Furthermore, the fault detection module 14 is further configured to perform the following steps: P41: Receive the real-time detection signal, transmit it back to the fault identification layer of the fault decision module, execute program code self-check and judgment based on the real-time inflection point window, and determine the real-time fault; P42: If the real-time fault is empty, output a normal gating signal; P43: If the real-time fault is not empty, transfer the real-time fault flow to the evolution prediction layer, perform fault trend prediction, and determine the predicted fault; P44: Map the real-time fault and the predicted fault, determine the fault detection data and output it.

[0049] It should be understood that the fault detection module 14 of the present application takes the real-time detection signal of the intelligent unit door as input, and combines with the fault decision module to realize real-time judgment and trend prediction of faults, generate comprehensive fault detection data including real-time faults and predicted faults, and ensure the system's operating stability and fault response efficiency.

[0050] Specifically, the real-time detection signal of the intelligent unit door is first received, including sound signals, vibration signals, and program running codes, etc. These signals fully reflect the current operating status of the power transmission module. The received signal is transmitted back to the fault identification layer of the fault decision module to perform two key tasks: the first is to analyze the program running code through the code self-test device to determine whether the instruction execution is normal and identify potential abnormalities at the control level; the second is to combine the periodic matching technology based on the signal inflection point window to compare the real-time detection signal with the periodic reference signal to determine whether there are abnormal signal characteristics at the physical level. Through the dual judgment of the code layer and the physical layer, the module accurately determines whether there is a real-time fault, laying the foundation for subsequent processing.

[0051] Next, if the fault identification layer determines that the real-time fault is empty (i.e. the current operation status of the device is normal and has no abnormal characteristics), the module immediately outputs a normal gating signal. This signal feeds back the health status of the current device to the control system to confirm the normality of the gating operation. At the same time, the module continuously monitors the operating status of the device in real time to ensure that abnormal signals can be captured and processed in time to avoid missed judgments or delays.

[0052] Next, when the fault identification layer determines that there is a real-time fault (that is, abnormal features appear in the detection signal), the module transfers the real-time fault data to the evolution prediction layer of the fault decision module, triggering the fault trend prediction process. The evolution prediction layer is based on time series analysis and analyzes the dynamic change pattern and trend characteristics of the fault signal based on the training model of the fault evolution stage. By learning the characteristics of the current abnormal signal such as amplitude, frequency change and duration, the module predicts the possible evolution direction, severity and occurrence time of the fault, providing important decision-making basis for preventive maintenance and emergency response.

[0053] Furthermore, real-time faults and predicted faults are mapped and analyzed, and the two types of data are integrated to generate complete fault detection data. This process evaluates the correlation between the current fault and the potential risk by matching the real-time signal characteristics and the prediction trend model. For example, the module can determine the fault development speed and possible impact range based on the spectral characteristics, amplitude changes and trend curves of the abnormal signal. Finally, the generated fault detection data consists of two parts: real-time faults, which are used to describe the current abnormal state of the equipment; predicted faults, which are used to provide risk warnings that may occur in the future. The module outputs this comprehensive data to the monitoring terminal or maintenance system to provide precise support for fault management and operation optimization.

[0054] Through the above steps, a closed-loop analysis of the entire process from real-time signal acquisition to fault determination and trend prediction is achieved. The module combines program code self-checking with signal periodic matching technology to ensure comprehensive identification of faults. At the same time, through trend analysis of the evolution prediction layer, the system's fault warning capability and response efficiency are significantly improved, providing a strong guarantee for the stable operation of the intelligent unit door.

[0055] Furthermore, when executing the program code self-check and determining based on the real-time inflection point window to determine the real-time fault, step P41 of the embodiment of the present application further includes: P41-1: Interact with the code running data of the power transmission module, execute program code self-check, and determine the first detection result; P41-2: Based on the real-time detection signal, locate the real-time inflection point window, match the real-time inflection point window with the signal inflection point window, and determine the window signal pair; P41-3: Traverse the window signal pair, make a judgment at the periodic time node, make a second judgment based on the inflection point feature vector, and determine the second detection result; P41-4: Integrate the first detection result and the second detection result to determine the real-time fault.

[0056] Optionally, when executing program code self-check and judgment based on the real-time inflection point window to determine real-time faults, the embodiment of the present application gradually refines the judgment through a dual detection process of code running data and signal feature matching, and finally integrates the detection results to achieve accurate identification of real-time faults.

[0057] First, the code running data of the power transmission module is obtained through interaction. These data include real-time executed control instructions, status feedback, and execution logic information. The code running data reflects the control layer status of the equipment and is an important basis for identifying potential logical faults. The module uses the program code self-checker to analyze these data to check the logical integrity of the running code, the correctness of parameter calls, and the response time of instruction execution. Through abnormal pattern recognition methods (such as logic jump abnormality detection and execution path analysis), the module determines whether there are problems such as instruction conflicts, response delays, or illegal operations, generates the first detection result, and preliminarily determines whether there is a fault in the code layer.

[0058] Next, based on the real-time detection signal (such as sound or vibration signal), identify the key feature points in the signal and locate the real-time inflection point window. The real-time inflection point window is a local signal extracted with the inflection point of the current detection signal as the center, which is used to characterize the current operating status characteristics of the equipment. Subsequently, the module matches the real-time inflection point window with the previously generated signal inflection point window (the characteristic window of the reference signal) to determine the window signal pair. This matching process is achieved through the dynamic time warping (DTW) algorithm, which compares the consistency of the two signal windows in timing and characteristics, providing basic data for subsequent fault judgment.

[0059] Furthermore, the module traverses the window signal pairs one by one and completes two-layer judgments in sequence. First, a judgment is made based on the periodic time node. For each pair of window signals, the periodic sequence nodes of the real-time inflection point are compared to see whether they are consistent with the periodic nodes of the reference signal, and the time offset of the inflection point (such as the time offset corresponding to a certain pulse peak) is analyzed in detail. If the inflection point is significantly offset, it may indicate that the operating cycle of the equipment is abnormal. In addition, the state distribution of the inflection point window is proofread to check the consistency of the real-time signal with the reference signal in amplitude, frequency and morphology. Then, based on the first judgment, a second judgment is made using the characteristic vector of the inflection point (including gradient features, amplitude features, etc.) to check the degree of matching between the real-time signal and the reference signal in local features, and refine the ability to capture signal anomalies through deep feature comparison technology. Through two-layer judgment, a second detection result is generated to reflect the real-time signal status at the physical level.

[0060] Furthermore, the first detection result generated by the program code self-check and the second detection result generated by the signal matching analysis are integrated. During the integration process, a weighted fusion method can be used to assign different weights to the two results according to the possible source of the fault (code layer or physical layer), and the final fault judgment result is obtained comprehensively. If both results point to an abnormality, the module confirms that there is a real-time fault; if one of the results is normal and the other is abnormal, the module records the abnormal signal and marks it as a potential fault. The final real-time fault judgment result is output for subsequent fault processing and trend prediction.

[0061] Through the above steps, dual detection of the code layer and the physical layer is achieved, and accurate identification of real-time faults is ensured through layer-by-layer refinement. The module combines program code self-checking with signal inflection point window matching technology, effectively covering a variety of control anomalies and mechanical failures, and significantly improving the accuracy and robustness of fault detection.

[0062] Further, after determining the fault detection data, the fault detection module is further configured to execute step P40a, wherein step P40a further comprises: P41a: Integrate the fault detection data, generate a fault detection sheet, and display the fault on a visualization platform; P42a: Traverse the fault detection sheet, and if the risk level is greater than the preset risk threshold, generate an alarm pop-up window and perform pop-up alarm management.

[0063] In a possible embodiment of the present application, after the fault detection module completes the generation of fault detection data, it can further integrate fault data, generate detection orders, and perform visual alarm management based on risk levels, thereby enhancing the user's comprehensive perception and rapid response capabilities of fault status.

[0064] First, the generated fault detection data is integrated, including real-time fault information (currently detected abnormal state) and predicted fault information (fault trends generated by the evolutionary prediction layer). The integrated data is organized into a standardized fault detection sheet, which records in detail the type of fault, time of occurrence, location (such as specific components in the power transmission module: motor, belt, etc.), fault characteristics (such as abnormal signal amplitude, periodic offset, etc.) and predicted risk trends (such as potential fault deterioration time window). The generated fault detection sheet is uploaded to the visualization platform through the interface, and the fault information is displayed in an intuitive way in the platform interface, such as marking the fault location and status through a fault list, timeline or hot map. The core of the visualization display is to enhance the user's understanding of the fault, so that they can quickly locate the problem and evaluate its impact on the operation of the equipment.

[0065] After the fault detection list is generated, the module traverses the fault items one by one and dynamically manages them according to the risk level of each fault. The risk level is a comprehensive score calculated based on the fault type, equipment operating status and predicted trend, which is used to measure the impact of the fault on the use and safety of the equipment.

[0066] The risk level is compared with the preset risk threshold. If the risk level of a certain fault is lower than the preset threshold (such as slight vibration or sound abnormality, which does not significantly affect the function of the equipment), the module will upload it to the visualization platform for the user to view. The status information of such faults is only used as a reference for operation and maintenance to avoid excessive disturbance to user operations. If the risk level of a certain fault is higher than the preset threshold (such as serious power transmission module failure or high safety risk fault), the alarm pop-up window is immediately triggered to generate a prominent alarm prompt on the user operation interface. For example, the pop-up window displays the fault details (including fault name, risk level, recommended handling measures, etc.) and is accompanied by sound and light warnings to ensure that users can quickly pay attention and take action. The pop-up window management function places the highest risk faults in a prominent position through priority sorting to ensure that important information is processed in a timely manner.

[0067] Through the above steps, the whole process management from data integration to risk assessment and dynamic alarm is realized. The integrated fault detection list enhances the systematicness and traceability of fault information, while the pop-up alarm management optimizes the user's operation experience through the risk classification mechanism, which not only avoids the frequent interference of low-risk faults, but also ensures that high-risk faults can quickly attract attention, thus improving the overall safety and operation efficiency of the intelligent unit door system.

[0068] In summary, the embodiments of the present application have at least the following technical effects: This application analyzes the detection signal (such as sound signal) of the power transmission module of the intelligent unit door, extracts the standard periodic signal, and generates a periodic reference signal based on it to capture the key feature points in the signal. Subsequently, the fault decision module is supervised and trained using the periodic reference signal, and the environmental adaptability of the model is improved through transfer learning. In actual operation, the fault detection module receives the real-time detection signal, and through linkage with the fault decision module, it performs signal period matching and fault trend prediction, and finally generates detection data containing real-time faults and predicted faults, thereby realizing accurate fault identification and early warning of the intelligent unit door.

[0069] The technical effect of improving the real-time and accuracy of fault detection while enhancing the adaptability to multi-scenario environments is achieved through periodic signal matching, fault trend prediction and transfer learning.

[0070] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0072] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. Intelligent unit door fault warning system based on combined algorithm, characterized in that: The system comprises: A standard periodic signal determination module is used for a power transmission module of an interactive intelligent unit door to determine a standard periodic signal based on periodic transmission characteristics, wherein the power transmission module at least includes a motor, a reducer, a hanging wheel and a belt, and the standard periodic signal is determined based on a detection signal, and the detection signal at least includes a sound signal; A periodic reference signal determination module, used for identifying and labeling a signal inflection point window for the standard periodic signal, and determining a periodic reference signal, wherein the signal inflection point window is determined by a preset window size centered on the signal inflection point; A fault decision training module, used for supervising the training of a fault decision module based on the periodic reference signal, wherein the fault decision module is constructed by transfer learning; The fault detection module is used to receive the real-time detection signal of the intelligent unit door, and in combination with the fault decision module, perform periodic matching judgment and fault trend prediction based on the signal inflection point window to determine the fault detection data, wherein the fault detection data includes real-time faults and predicted faults.

2. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 1, characterized in that: The periodic reference signal determination module is also used for: Determine a signal spectrum of the standard periodic signal and identify a signal inflection point, wherein the signal inflection point is marked with a periodic timing; Based on the preset window size, performing signal window interception on the signal inflection point to determine a signal inflection point window; For the signal inflection point window, an inflection point feature vector is determined, and the standard periodic signal is identified to determine the periodic reference signal.

3. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 2, characterized in that: When the periodic reference signal determination module determines the inflection point feature vector for the signal inflection point window, the periodic reference signal determination module further includes: For the signal inflection point window, a first front window, an inflection point and a first rear window are divided and determined, wherein the first front window is determined based on the front boundary and the inflection point position of the signal inflection point window; Calculating a first gradient feature vector of a signal window based on the first front window and the inflection point, wherein the first gradient feature vector includes a gradient value and a gradient direction; Calculating a second gradient eigenvector of the signal window based on the inflection point and the first rear window; The first gradient feature vector and the second gradient feature vector are integrated to form the inflection point feature vector.

4. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 1, characterized in that: The fault decision training module is also used for: Call the fault records of the intelligent unit door to mine transmission fault samples, where each transmission fault sample includes the real-time fault stage and the fault evolution stage; Based on the transmission fault samples, sample supervision training is performed to determine a fault decision model; Personalized assembly information of interactive intelligent unit doors to determine personalized features; An inflection point analysis layer is added, and transfer learning is performed on the fault decision model in combination with the personalized features to determine the fault decision module.

5. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 4 is characterized in that: When the fault decision training module performs sample supervision training and determines the fault decision model, it also includes: Based on the transmission fault samples in the real-time fault stage, supervise the training of N fault decision makers, wherein the result with the largest proportion of the N decision results is output; Obtain the transmission operation code of the power transmission module and train the code self-tester; Integrate the N fault decision makers as a first branch, use the code self-checker as a second branch, and parallelize the first branch and the second branch to determine a fault identification layer; Based on the transmission fault samples in the fault evolution stage, prediction training is performed in time series to determine the evolution prediction layer; The evolution prediction layer is post-connected to the fault identification layer to generate the fault decision model.

6. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 5, characterized in that: The fault detection module is also used for: Receive the real-time detection signal, transmit it back to the fault identification layer of the fault decision module, perform program code self-check and determination based on the real-time inflection point window, and determine the real-time fault; If the real-time fault is empty, output a normal gating signal; If the real-time fault is not empty, the real-time fault is transferred to the evolution prediction layer to perform fault trend prediction and determine the predicted fault; The real-time fault and the predicted fault are mapped, and the fault detection data is determined and output.

7. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 6, characterized in that: When the fault detection module performs program code self-checking and determination based on the real-time inflection point window to determine the real-time fault, it also includes: Interacting the code running data of the power transmission module, executing the program code self-check, and determining the first detection result; Based on the real-time detection signal, locate the real-time inflection point window, match the real-time inflection point window with the signal inflection point window, and determine a window signal pair; Traversing the window signal pair, performing a primary determination based on a periodic time node, performing a secondary determination based on an inflection point feature vector, and determining a second detection result; The first detection result and the second detection result are integrated to determine a real-time fault.

8. The intelligent unit door fault early warning system based on the combined algorithm as claimed in claim 1, characterized in that: After determining the fault detection data, the fault detection module is further used to: Integrate the fault detection data, generate a fault detection sheet, and display the fault on a visualization platform; The fault detection list is traversed, and if the risk level is greater than the preset risk threshold, an alarm pop-up window is generated and pop-up alarm management is performed.

Images