Elevator traveling cable high-frequency time domain AI early warning method and device

Abstract

The invention provides an elevator accompanying cable high-frequency time domain AI early warning method and device, and the method comprises the steps: transmitting a 10MHz-100MHz square wave excitation signal (including a CAN2.0 / RS485 protocol threshold critical point), synchronously capturing the waveforms of the input and output ends of a cable at a 1Gsps sampling rate, and improving the sensitivity of tiny wire breakage and insulation aging faults; carrying out secondary FFT (primary fundamental wave / harmonic wave separation and secondary square wave interference elimination) or wavelet transform on the time domain signal based on the FPGA to generate a spectrogram / time-frequency diagram, and inputting the spectrogram / time-frequency diagram into a pre-trained YOLOv5 model to output a fault type and a fault position; reflection characteristic differences are compared through a reference cable to correct environmental interference, and the device integrates modules such as an MCU + FPGA + upper computer architecture, a voltage-controlled oscillator (supporting 100MHz output), a high-speed acquisition card and the like, so that portable intelligent detection is realized. The scheme breaks through the limitation that a traditional method is low in sensitivity and depends on manpower, and is suitable for real-time early warning of the elevator cable dynamic bending scene.

Description

Technical Field

[0001] The present invention belongs to the technical field of elevator safety detection and fault warning, and specifically relates to an active warning method and device for elevator accompanying cable faults based on high-frequency time domain signals and artificial intelligence. The solution integrates high-frequency signal excitation (10MHz~100MHz square wave), time-frequency domain feature extraction technology (quadratic FFT / wavelet transform) and deep learning model (YOLOv5), and achieves high-sensitivity detection of minor faults such as broken wires and insulation aging of elevator accompanying cables. At the same time, through hardware integration (MCU+FPGA+host computer) and dynamic reference calibration, it solves the problem of insufficient reliability of traditional detection methods in complex environments. Background Art

[0002] Elevator accompanying communication cables are mainly used for the transmission of lifting control signals between civil or industrial elevators. Usually, the main transmission signals are CAN2.0 and RS485, with twisted pair as the transmission medium. The accompanying cable usually uses a twisted pair with a shielded layer and a characteristic impedance of 120ohm. Due to the increase in the use time of the elevator, some cables have different degrees of aging under the conditions of repeated bending, temperature and humidity changes. The main problems are: cable wire breakage, insulation layer damage, electric shock oxidation of the terminal, etc. The specific manifestation is intermittent failure of elevator communication and control, cable failure is difficult to repeat, and the cable status cannot be monitored by conventional experimental means such as multimeters. Therefore, it is urgent to propose an elevator accompanying cable detection scheme based on a new principle.

[0003] At present, the conventional detection methods for elevator accompanying cables are mainly the following: (1) On-resistance measurement: Use a multimeter or a special cable tester to measure the resistance of the cable. The resistance value should be within a certain range. Too high or too low resistance values ​​may indicate that the cable has problems such as poor contact, open circuit or short circuit.

[0004] (2) Insulation resistance measurement: Use a professional insulation resistance tester to measure the insulation resistance of the cable. The insulation resistance value should be within a certain range. Too low an insulation resistance value may indicate that there is an insulation problem with the cable, which may cause leakage or failure.

[0005] (3) Signal transmission measurement: Detect the stability and accuracy of cable transmission signals by sending and receiving signals. Instruments such as oscilloscopes can be used to detect signal waveforms, frequency, amplitude and other parameters to determine whether signal transmission is normal.

[0006] (4) Shielding performance measurement: For cables with shielding layers, a professional shielding performance tester can be used to measure the shielding effect of the cable. During the test, connect the tester to the shielding layer and the grounding terminal of the cable to check whether the shielding performance meets the standard requirements.

[0007] However, these conventional detection methods may fail to judge cable gap faults. Even with bending tests or temperature tests, they may not be able to fully reproduce the fault conditions, resulting in missed detections. At the same time, existing detection methods rely on the elevator's own signal source and communication terminal, lacking active warning means from the outside.

[0008] In recent years, some new cable fault detection and location methods have been gradually put into use, mainly including: (1) Time Domain Reflectometry (TDR) The time domain reflection method locates fault points such as open circuit, short circuit or poor contact by measuring the time and amplitude of the reflected signal in the signal cable. Zhan Li et al. "Application of cable detection based on TDR technology in underground communication in coal mines" Industrial Control Computer, 2023, 36(1): 56-61. It is proposed to use the Ethernet physical layer chip 88E1512P (Marvell) with time domain reflection measurement function to locate the fault of Ethernet communication lines in coal mines. Shu Hongchun et al. "A new method for time domain fault location of double-circuit transmission lines in wind farms" Power System Technology, 2023, 47(4): 1451-1459. It is proposed to use the time domain finite difference method to discretize the transmission line equation expressed in the form of partial differential equations to construct a fault location function to achieve fault location, and the PSCAD / EMTDC simulation software and hardware-in-the-loop test experimental platform are used for analysis to verify the effectiveness of the proposed algorithm.

[0009] (2) Frequency Domain Reflectometry (FDR) and Reflection Coefficient Spectrum Method The frequency domain reflection method mainly analyzes the cable defect status through broadband impedance spectrum and reflection coefficient spectrum. The sensitivity and detection accuracy of this method to high-frequency signals are better than those of the time domain reflection method. In recent years, the research on fault detection methods based on reflection coefficient spectrum has been more concentrated. Zhao Hongshan et al. "Cable defect diagnosis method based on reflection coefficient spectrum integration" Power System Technology, 2022, 46(11): 4548-4556. A cable local defect location method based on reflection coefficient spectrum integration was proposed. This method can achieve high-precision positioning while reducing the signal frequency and the number of sampling points, with an error of less than 1%.

[0010] (3) Ultrasonic testing technology Yang Jie et al., "Research on partial discharge fault analysis of 10kV switchgear cable based on ultrasonic detection", Power Equipment Management, 2018, 11: 41-45. It is proposed to use ultrasonic sensors to detect ultrasonic pulses generated by partial discharge of 10kV cables. However, the voltage used in communication cables is relatively low, and the discharge phenomenon is not obvious, so the application prospects in communication cables are limited.

[0011] Among the above cable detection methods, the ultrasonic detection method is only applicable to high-voltage cables, while the voltage in communication cables is relatively small. Even if there is leakage, corona is usually not generated. The time domain reflection method is limited by the cable length. When the cable length is insufficient, it is impossible to effectively distinguish the reflected signal and the input signal, and its sensitivity to faults is insufficient. The frequency domain emission method and the dielectric impedance spectrum method are mainly used for insulation fault detection of AC power frequency high-voltage cables. The measurement results rely on the user's personal experience interpretation, and the distortion characteristics of the time domain signal waveform cannot be directly obtained from the frequency domain signal detection results. In addition, obtaining the reflection spectrum or dielectric impedance spectrum requires the use of a relatively expensive network analyzer or vector analyzer, which has a high cost and a limited use environment. Therefore, it is urgent to propose an intelligent detection solution that has accurate measurement, reasonable cost, takes into account both time domain and frequency domain signal measurement, and has the ability to intelligently determine the type and degree of faults. Summary of the invention

[0012] In view of the defects and shortcomings of the prior art, the present invention provides a high-frequency time-domain AI early warning method and device for elevator accompanying cables, which solves the problems of low sensitivity to minor faults, poor adaptability to environmental interference, and reliance on manual experience of traditional detection methods through the following technical solutions: High-frequency excitation and multi-frequency point measurement: 10MHz~100MHz square wave signal is transmitted to the elevator accompanying cable (characteristic impedance 120Ω), and the input and output time domain waveforms are synchronously captured through the 1Gsps high-speed acquisition module to enhance the sensitivity to minor faults such as broken wires and insulation aging; Secondary FFT / wavelet transform feature extraction: Based on FPGA, perform secondary Fourier transform (first separation of fundamental wave and harmonics, second removal of square wave spectrum interference) or continuous wavelet transform on the time domain signal to generate a spectrum diagram / time-frequency diagram that characterizes the fault reflection characteristics; YOLOv5 intelligent classification and dynamic calibration: Input feature maps into the pre-trained YOLOv5 model, output the fault type, location and distortion degree (location accuracy error ≤ 1%), and correct environmental interference such as temperature and bending by comparing with the reference cable to generate a high-reliability diagnosis report; Hardware system integration: The device adopts MCU+FPGA+host computer architecture, integrating components such as signal transmission module, voltage-controlled oscillator (100MHz square wave output), power half-bridge, etc. that support CAN2.0 / RS485 protocol, realizing the portability and intelligence of laboratory-level detection functions.

[0013] This solution combines the advantages of high precision, strong anti-interference ability and real-time warning, and is suitable for dynamic bending scenarios of elevator accompanying cables.

[0014] The technical solution specifically adopted by the present invention to solve the technical problem is: A high-frequency time-domain AI early warning method for elevator accompanying cables comprises the following steps: Transmit a square wave excitation signal with a frequency ≥10MHz to the traveling cable of the elevator to be tested, and synchronously collect the time domain waveform signals at the input and output ends; Performing a secondary Fourier transform or a continuous wavelet transform on the time domain waveform to generate a spectrum diagram or a time-frequency diagram; The frequency spectrum or time-frequency spectrum is input into a target detection model, and fault type and location information is output.

[0015] Furthermore, the target detection model is a pre-trained YOLOv5 model.

[0016] Furthermore, the frequency range of the square wave excitation signal is 10 MHz to 100 MHz, and complies with the threshold critical point of the elevator communication protocol CAN2.0 or RS485.

[0017] Furthermore, the quadratic Fourier transform separates the fundamental wave and the harmonics through the first FFT; and through the second FFT: Remove square wave spectrum interference; where C is the characteristic signal after the second FFT, is the frequency domain signal after the first FFT, represents the square wave spectrum interference term, n is the harmonic number, ω is the angular frequency, τ is the square wave pulse width time parameter, is the inverse Fourier transform factor.

[0018] Furthermore, the training data set of the target detection model includes: More than 200 elevator cable samples with different fault types, including broken wires and insulation aging; More than 30 normal cable samples, collected to simulate the bending state during elevator operation; Input waveform and output waveform data, covering 1MHz~100MHz square wave signals.

[0019] Furthermore, by comparing the test results of the reference cable with those of the cable to be tested, errors caused by environmental interference can be corrected.

[0020] And, an elevator accompanying cable high-frequency time domain AI warning device, comprising: The signal transmitting module is used to transmit a square wave excitation signal with a frequency ≥10MHz to the cable under test, supports CAN2.0 and RS485 communication protocols, and supports switching output at multiple frequency points; High-speed acquisition module, including acquisition card with sampling rate ≥1Gsps, synchronously captures the time domain waveform signals at the input and output ends of the cable; The signal processing module performs a second Fourier transform or a continuous wavelet transform on the time domain waveform based on FPGA to generate a spectrum diagram or a time-frequency diagram; The AI ​​judgment module is deployed on the host computer, and integrates the target detection model to classify the spectrum diagram or time-frequency diagram, and outputs the fault type and location.

[0021] Furthermore, the target detection model is a pre-trained YOLOv5 model, and the fault location accuracy output by the YOLOv5 model is verified through a detection experiment of a broken wire 1 / 3 copper core.

[0022] Furthermore, the signal transmission module includes a voltage-controlled oscillator and a power half-bridge, supporting a square wave output of up to 100 MHz.

[0023] Further, the apparatus includes: a first handheld device, configured in a transmission mode, integrating a signal transmission module, a communication interface and an energy supply module; A second handheld device, configured in a receiving mode, integrating a signal receiving module, a communication interface and an energy supply module; The first handheld device and the second handheld device work together through a communication interface to complete dynamic calibration or multi-frequency point measurement.

[0024] Compared with the prior art, the present invention and its preferred embodiments include at least the following beneficial effects: Significantly improve the sensitivity of minor fault detection: through high-frequency time domain signal excitation (10MHz and above square wave) combined with secondary FFT / wavelet transform, it can effectively capture the reflection characteristics caused by minor impedance changes such as broken wires and insulation aging inside the cable, and overcome the problem of missed detection caused by the traditional method's reliance on low-frequency signals; Realize automatic and precise fault judgment: Based on the YOLOv5 model (which can be replaced with other similar or higher versions of the model as needed), end-to-end classification of spectrum / time-frequency graphs is performed to directly output the fault type, location, and degree of distortion, reducing reliance on manual experience and avoiding the risk of subjective misjudgment in traditional spectrum analysis methods; Enhanced adaptability to complex environments: Through dynamic comparison with reference cables and multi-frequency signal measurement, the impact of environmental interference such as temperature fluctuations and cable bending on the detection results is suppressed, and the detection reliability in dynamic scenarios such as elevator shafts is improved; Taking into account both high precision and practicality: The device adopts a lightweight design of MCU+FPGA+host computer, integrating high-frequency signal transmission, high-speed acquisition and intelligent analysis modules, achieving portable deployment while ensuring laboratory-level accuracy, and adapting to the needs of elevator maintenance sites. The preferred split design adapts to the narrow space of the elevator shaft, solving the problems of bulky traditional equipment and complex wiring; a single device integrates transmission and reception functions, supports independent self-test and dual-machine collaborative detection, and improves on-site deployment flexibility; the dual-device independent power supply design optimizes the power management logic to ensure dynamic calibration stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments: Figure 1 It is a schematic diagram of the reflection and transmission principle of a damaged cable transmission line (the broken part is section 2); Figure 2 It is a structure diagram of YOLO - v5; Figure 3 It is a test result graph of a 1MHz signal square wave. The left graph corresponds to a normal cable, and the right graph corresponds to a damaged cable; Figure 4 It is a test result graph of a 10MHz signal square wave. The left graph corresponds to a normal cable, and the right graph corresponds to a damaged cable; Figure 5 It is for a good cable and a damaged cable / Signal comparison graph; Figure 6 It is an FFT transform spectrum graph of a good cable and a damaged cable: where the solid pink line represents the good cable, and the dashed red line represents the damaged cable; Figure 7 It is a continuous wavelet transform image of the time - domain signal: the left graph corresponds to a normal cable, and the right graph corresponds to a damaged cable; Figure 8 It is a schematic diagram of the device principle of the embodiment of the present invention; Fig. 9 It is a hardware system engineering structure diagram of the elevator trailing cable active warning detection device of the embodiment of the present invention; Fig.10 It is a hardware design block diagram of the signal detection end of the embodiment of the present invention; Fig.11 It is a hardware design block diagram of the signal transmitting end of the embodiment of the present invention. Specific embodiments

[0026] To make the features and advantages of the present invention more obvious and understandable, specific embodiments are given below for detailed description as follows: It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs.

[0027] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0028] The embodiments of the present invention address the deficiencies of the time domain reflection method and the frequency domain reflection method. Based on the transmission and signal integrity theory, a new detection method and experimental device are proposed: a reliability stress test and active fault warning device for elevator communication cables based on high-frequency time domain signals. The scheme sets the actual analog or digital communication signal at the threshold critical point of the actual standard, greatly increases the signal transmission frequency, and makes the minor faults inside the cable, such as broken wires and insulation damage, reflected in the output waveform, thereby realizing active fault warning of the elevator communication cable. The scheme also uses artificial intelligence methods instead of manual recognition, which is conducive to improving the accuracy of determining the fault type and fault location.

[0029] The basic principle of this embodiment is based on the transmission line theory: due to the oxidation of connector contacts, cable wire breakage, aging of insulation layer, etc., the capacitance and inductance of the cable transmission line will change, causing the characteristic impedance of the fault point to change, resulting in impedance mismatch and signal reflection. The potential hidden dangers of signal reflection to signal transmission include: (1) Multiple reflected signals and actual transmission signals are superimposed, resulting in ringing effects on the rising and falling edges of the signal, resulting in digital signal decoding errors. (2) The signal actually transmitted to the cable terminal is attenuated as a whole, causing misjudgment of high-level signals.

[0030] In order to analyze the impact of cable defects on the actual waveform of the transmission signal, the broken wires and cracks in the twisted pair transmission line of the cable are uniformly modeled as Figure 1 Structure shown: Assume that the signal source S is an AC sinusoidal signal and the output impedance is Z 0 , the frequency is f , the circular frequency is ω= 2 πf， The cable is divided into three sections, namely normal section 1 (directly connected to the signal source), damaged section 2 and normal section 3. The cable end is connected to a 120Ω matching resistor to prevent end signal reflection and keep consistent with the actual RS485 and CAN transmission twisted pair configuration. Since the end connection matching resistor will not generate a reflected signal, the actual signal reflection surface includes: (1) the connection between the signal source and the cable F0; (2) the direct interface F1 between section 1 and section 2; and (3) the interface F2 between section 2 and section 3. The F0 reflection signal direction is to the right, which is the positive direction, and the F1 and F2 reflection directions are to the left, which is the negative direction.

[0031] Since the reflection surfaces F0, F1 and F0, F2 respectively constitute two groups of signal reflection cavities to reflect the signal multiple times, according to the transmission line theory, the signal received by the cable terminal is divided into three parts, namely, the signal that passes directly through the cable without reflection , and the signal transmitted after multiple reflections in the F0F1 and F0F2 cavities , : (1) (2) (3) Among them, is the transfer function of the g1 segment, , , are the reflection coefficients of the three reflecting surfaces respectively, is the transmission coefficient of the F0 surface, and are the forward transfer function and the reverse transfer function of the g1 and b segments, and A is the signal reception coefficient of the cable. Under high-frequency conditions, the influence of the cable resistance can be ignored, and the frequency-domain response of the signal actually received at the cable terminal is: (4) This signal is the basic expression of the high-width-frequency transmission spectrum signal, and the fault information is reflected in and , that is, the influence of the reflections at the two end faces of the defect on the signal at the output cable terminal.

[0032] For the time-domain signal transmitted by the cable, it can usually be simplified to a square wave. In a discrete sampling system, the frequency-domain response of the square wave is: (5) Therefore, the actual output signal of the cable is affected by the defect, and its time-domain waveform is: (6) The above formula shows that the defect of the cable will cause the transmission intensity of the signal to change at different frequencies, thus causing distortion of the actual output signal of the cable. It can be seen from formula (5) that if the frequency of the square wave signal is low, the overall frequency of the frequencies covered in its spectrum is also low, and the frequency-domain coverage is insufficient, and the complete defect reflection information cannot be reflected in the time-domain signal. Therefore, in this embodiment, it is considered that when detecting the signal, the frequency of the square wave signal needs to be adjusted to 10 MHz or above, and the time-domain signals of multiple frequency points of the square wave need to be measured.

[0033] Fault feature extraction method based on integral transform In order to extract the characteristics of the fault defect, the time-domain waveform can be processed by integral transform. The Fourier transform is the most commonly used integral transform method, and the fast Fourier transform method has a fast processing speed and a mature implementation scheme. Performing a fast Fourier transform (FFT) on the output signal of the faulty cable can obtain: (5) Substituting (1)(2)(3)(5) into (6) and expanding the number of defects to n, we can obtain: (6) It can be seen that the FFT transform spectrum still contains The sine and cosine components of the square wave are obtained by FFT transformation. As is known, the signal is removed, so the signal is subjected to a second Fourier transform to extract features, and the formula is as follows: (7) Although the signal features contain all the spectrum features directly related to the fault distance zi, Located in the denominator, the reflected signal at a fault end face contains several high-order harmonics in its expansion, and may produce aliasing and broadening. It is ideal to use artificial intelligence methods to determine faults.

[0034] In addition, the effect can also be achieved by directly using continuous wavelet transform to analyze the time domain signal and make judgments through artificial intelligence methods. The continuous transformation formula is: (8) Among them, the general expression of the basis function of continuous wavelet transform is: (9).

[0035] Defect and fault determination method based on artificial intelligence method Considering the actual defect shape and the influence of the cable bending degree, the transformed peak will be broadened, and the harmonic peaks at different defects will overlap with the fundamental peak. Therefore, only inferring the cable fault state through FFT transformation may lead to misjudgment. Considering that the actual cable state is affected by many factors, such as temperature, bending degree, etc., extracting the effective signal based on wavelet and Fourier transform from the real spectrum signal and converting it into actual fault information requires heavy reliance on the personal experience of the user. In order to better classify, reduce noise, extract and distinguish the fault information features extracted by integral transform.

[0036] Whether extracting features by performing two consecutive transformations through FFT or by performing one-step wavelet transformation, the extracted feature image can be converted into a two-dimensional image, and the fault features can be extracted by yolo-v5 or a higher-order image recognition model. Figure 2 The network structure of YOLOv5 is given, which mainly includes four parts: Input, Backbone, Neck and Prediction.

[0037] Its working principle is briefly described as follows: At the input end, the image first undergoes a series of preprocessing steps, including adjusting to a size suitable for network training and normalization. The architecture of the input end consists of three key components: Mosaic data enhancement, adaptive anchor box calculation, and adaptive image scaling. Mosaic data enhancement technology integrates multiple images to not only enhance the diversity of training samples, but also significantly improves the model's detection performance for small targets, while accelerating the training process and improving detection accuracy. Adaptive anchor box calculation dynamically adjusts the prediction box size in each iteration to make the model's prediction more accurate and reasonable. Adaptive image scaling ensures the consistency of the input image, facilitating efficient processing by the network. Backbone, as a baseline network, is responsible for extracting common feature representations from images. Backbone consists of the Focus module, CBS module, CSP module, and SPPF module. The Focus module preprocesses the image through slicing operations, laying the foundation for subsequent feature extraction. The CBS module applies convolution, batch normalization, and SiLU activation to process input information and produce output results. The CSP module optimizes the problem of large computational workload during inference and improves the computational efficiency of the network. The SPPF module performs adaptive size output by pooling the input feature map into a fixed-size map. The Neck network is located between the Backbone and the Head, aiming to further improve the diversity and robustness of the features. The FPN-PAN structure is adopted. This design enhances the network's ability to fuse multi-scale features, enabling the model to better cope with target objects of different sizes. Finally, the Head part is responsible for generating the final detection results. It adopts a coupled head structure, uses the BCE loss function for classification tasks, and uses the CIoU loss function for bounding box regression tasks.

[0038] In order to train the model, a large-scale test data set needs to be constructed manually. First, it is necessary to collect multiple cables in different fault states and classify them according to the fault type and fault degree. The total number of cables with different fault types should be no less than 200. There should be no less than 30 new cables for comparison. The cables are placed according to the bending position during actual operation, CAN2.0, RS485, and square wave signals with different transmission frequencies are provided at the input end of the cable, and the received signal at the output end is measured. The data set information should include the following elevator cable information: (1) production date, (2) cumulative working time, (3) fault type and fault location information, (4) detection signal parameters, (5) input waveform (including defect reflection signal), (6) output waveform, (7) maintenance personnel's description of the communication fault and other information.

[0039] Take the 3.5-meter-long standard elevator cable as an example. At the 1-meter position of the red signal line in its shielded twisted pair, the copper wire is cut off by a blade by about 1 / 3. The copper core diameter of the twisted pair cable is 0.5mm, and the characteristic impedance is 120ohm.

[0040] The input and output signals of the elevator accompanying cable are measured by the acquisition card. The former contains the superposition of the original signal and the reflected signal. The latter is the cable output signal affected by defects. The experimental results show that Figure 3 As shown in Figure 1, for a 1MHz square wave signal, the waveforms of the good cable and the fault current cannot be distinguished. Figure 4 As shown in the figure, for a 10MHz square wave signal, due to the influence of the cable inductance parameters, the high-frequency signal is filtered, and the square wave signal degenerates into a nearly sinusoidal wave model. In addition, the waveform of the transmission signal of the faulty cable is clearly distinguished from the input waveform, as shown in the yellow box. Note that for Figure 4 Although the waveform in the middle of the right figure is not obviously distorted, the incident waveform is distorted. Therefore, by detecting the high-frequency time domain signal, it is possible to determine whether the cable is faulty.

[0041] In order to determine the fault type and location, for a 3.5-meter-long elevator cable, a 1MHz~100MHz square wave signal is used as the detected signal, and the signal is collected at a 1G sampling rate. The signal is subjected to the first Fourier transform to obtain / Signals, such as Figure 5 You can see the difference between the signal from the damaged cable and the good cable, characterized by reflections from the damage in the cable.

[0042] right Figure 5 The signal in the second Fourier transform is carried out, and the second FFT transform spectrum of the good cable and the damaged cable is as follows Figure 6 The pink solid line and the red dotted line in the figure are shown. The curve is input into the yolo-v5 network to determine the degree and characteristics of the fault.

[0043] In addition, it is possible to Figure 4 The cable output signal in is subjected to continuous wavelet transform, where the transform result is as follows Figure 7 As shown, it is also used as the input of the yolo-v5 judgment network.

[0044] The one-dimensional or two-dimensional signals extracted by the integral transform are uniformly converted into two-dimensional images, which are input into the trained yolo-v5 network for judgment, and the judgment results are output, including key fault information such as fault type, number of faults, degree of damage, characteristic impedance of the damaged part, etc.

[0045] like Figure 8As shown, based on the above method design provided in this embodiment, a system device for implementing this method is further designed. By inputting a high-speed square wave signal into the cable, the frequency range is in the range of 1MHz~100MHz, and the waveforms at the input and output ends are measured simultaneously at a sampling rate of 1Gsps. These waveforms contain reflected signals at the damaged part of the cable. Feature extraction is achieved by adopting quadratic Fourier transform technology or two-dimensional continuous wavelet transform, reducing the judgment difficulty of the deep learning network. Then, the feature extraction image is identified through the yolo-v5 network, and a diagnostic report of the cable is given, including: defect type, location, degree and type of distortion, expected service life, etc.

[0046] Further hardware principles of the device are as follows Fig. 9 As shown, this implementation is implemented using a handheld device based on MCU, FPGA and host computer, including a transmitter and a receiver. When the actual system device is in use, two devices are required to complete a test, one for transmission and the other for reception. One device contains a complete transmitter and receiver power supply.

[0047] When the device executes the detection program, the signal transmitter of the system generates a threshold communication signal (485, CAN or analog position signal) and connects it to the input end of the cable to be tested and the reference cable. The reference cable is used for offline or online self-test of the system. The output end of the cable is connected to the signal detection end to perform digital decoding, logic analysis, waveform detection and high-frequency detection on the input signal. The host computer is used to set the parameters of the transmission signal and receive the test results. The test results of the cable to be tested and the reference cable are compared with the cable diagnostic database of the host computer, and the cable diagnostic results and system self-test results are given. The establishment of the cable diagnostic database requires the advance entry of a large amount of normal cable and various fault cable data through the system signal transmitter and detection end.

[0048] The signal transmitter and detection hardware mainly include the minimum system module, power management module, 485 and CAN physical layer transceiver, power half bridge, voltage controlled oscillator, photoelectric tube, high-speed acquisition card, etc. The hardware structure is as follows Fig.10 , Fig.11 shown.

[0049] 1) Minimum system module: including clock circuit, reset circuit, JTAG and serial port debugging circuit. It is the core of the entire hardware system and is responsible for providing an interface to communicate with other modules.

[0050] 2) Power management module: provides the required power for each hardware module and sensor to ensure the normal operation of the system.

[0051] 3) Debug interface module: Use JTAG or UART serial port to connect the hardware circuit board to the computer to provide program downloading and online debugging for the device.

[0052] 4) Signal output module: It is composed of CAN transceiver, 485 converter, photoelectric tube, voltage-controlled oscillator, power half-bridge, etc. It is controlled by a microcontroller to generate the pressure threshold detection signal required by the system.

[0053] 5) High-speed acquisition card: used to analyze the waveform distortion and noise interference of digital signals or analog pulse signals after cable transmission.

[0054] 6) Host computer communication module: uses serial port or Ethernet communication to control the lower computer and transmit measurement data.

[0055] The device implements the high-frequency time domain signal cable warning method proposed in the embodiment of the present invention through the following methods and steps: (1) The above device measures the high-frequency time domain signal of the accompanying cable, and can support the output of standard CAN2.0 arbitrary rate signals and RS485 arbitrary signals, as well as the output and time domain measurement of various high-frequency square waves and arbitrary waveforms. This provides a high-quality signal source that meets specific requirements for this method. The maximum frequency of the square wave can be output to 100MHz. In addition, the signal receiving end also includes CAN and RS485 standard signal decoding devices, which can verify whether the standard communication signal under the threshold condition can be decoded normally, as a basis for judging the cable quality.

[0056] (2) The signal transmitting end and the signal receiving end of the measuring device both contain acquisition terminals with a sampling rate of 1Gsps, which can sample high-frequency method signals or arbitrary waveform signals. It is used to continuously acquire the time domain waveform signals at the input and output ends of the cable. These acquired waveforms are used to determine and locate the faults of the current cable. On the other hand, they are also stored in the local database of the device for training the local yolo-v5 model.

[0057] (3) The detection waveform is uploaded from the lower computer to the industrial computer host computer through the communication interface. The quadratic Fourier transform feature extraction and wavelet feature extraction algorithms are executed by the host computer through the Python 3 program. The Yolo-V5 feature judgment model is also deployed in the local host computer, and the host computer CPU can be used to run the detection or the built-in small GPU of the host computer, such as the mobile version of RTX 4060, mobile version of RTX 3060, etc., for judgment.

[0058] (4) The energy supply used by the system can be called through the user graphical interface, which can realize one-click detection of the system and automatically generate a cable detection report.

[0059] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0060] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0061] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other various forms of high-frequency time-domain AI warning methods and devices for elevator accompanying cables under the inspiration of the present invention. All equal changes and modifications made within the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A high-frequency time-domain AI early warning method for elevator accompanying cables, characterized in that: The following steps are involved: Transmit a square wave excitation signal with a frequency ≥10MHz to the traveling cable of the elevator to be tested, and synchronously collect the time domain waveform signals at the input and output ends; Performing a secondary Fourier transform or a continuous wavelet transform on the time domain waveform to generate a spectrum diagram or a time-frequency diagram; The frequency spectrum or time-frequency spectrum is input into a target detection model, and fault type and location information is output.

2. The elevator accompanying cable high-frequency time domain AI early warning method according to claim 1 is characterized by: The target detection model is a pre-trained YOLOv5 model.

3. The elevator accompanying cable high-frequency time-domain AI early warning method according to claim 1 is characterized by: The frequency range of the square wave excitation signal is 10 MHz to 100 MHz, and complies with the threshold critical point of the elevator communication protocol CAN2.0 or RS485.

4. The elevator accompanying cable high-frequency time-domain AI early warning method according to claim 1 is characterized by: The quadratic Fourier transform separates the fundamental wave and harmonics through the first FFT; through the second FFT: Remove square wave spectrum interference; where C is the characteristic signal after the second FFT, is the frequency domain signal after the first FFT, represents the square wave spectrum interference term, n is the harmonic number, ω is the angular frequency, is the square wave pulse width time parameter, is the inverse Fourier transform factor.

5. The elevator accompanying cable high-frequency time-domain AI early warning method according to claim 1 is characterized by: The training data set of the target detection model includes: More than 200 elevator cable samples with different fault types, including broken wires and insulation aging; More than 30 normal cable samples, collected to simulate the bending state during elevator operation; Input waveform and output waveform data, covering 1MHz~100MHz square wave signals.

6. The elevator accompanying cable high-frequency time-domain AI early warning method according to claim 1 is characterized by: By comparing the test results of the reference cable with those of the cable under test, errors caused by environmental interference can be corrected.

7. A high-frequency time-domain AI early warning device for elevator accompanying cables, characterized in that: include: The signal transmitting module is used to transmit a square wave excitation signal with a frequency ≥10MHz to the cable under test, supports CAN2.0 and RS485 communication protocols, and supports switching output at multiple frequency points; High-speed acquisition module, including acquisition card with sampling rate ≥1Gsps, synchronously captures the time domain waveform signals at the input and output ends of the cable; The signal processing module performs a second Fourier transform or a continuous wavelet transform on the time domain waveform based on FPGA to generate a spectrum diagram or a time-frequency diagram; The AI ​​judgment module is deployed on the host computer, and integrates the target detection model to classify the spectrum diagram or time-frequency diagram, and outputs the fault type and location.

8. The elevator accompanying cable high-frequency time-domain AI early warning device according to claim 7 is characterized in that: The target detection model is a pre-trained YOLOv5 model, and the fault location accuracy output by the YOLOv5 model is verified by a detection experiment of a broken wire 1 / 3 copper core.

9. The elevator accompanying cable high-frequency time-domain AI early warning device according to claim 7 is characterized in that: The signal transmission module includes a voltage-controlled oscillator and a power half-bridge, and supports square wave output at a maximum of 100 MHz.

10. The elevator accompanying cable high-frequency time-domain AI early warning device according to claim 7 is characterized in that: include: A first handheld device, configured in a transmission mode, integrating a signal transmission module, a communication interface and an energy supply module; A second handheld device, configured in a receiving mode, integrating a signal receiving module, a communication interface and an energy supply module; The first handheld device and the second handheld device work together through a communication interface to complete dynamic calibration or multi-frequency point measurement.

Images