A multi-directional monitoring method, device and terminal based on power material sampling inspection

Abstract

This invention discloses a multi-directional monitoring method, device, and terminal based on the random inspection of power materials. The method includes: acquiring random inspection data in real time, wherein the random inspection data is data collected by a multi-directional monitoring terminal during the random inspection of power materials, and the power materials are fixed on the multi-directional monitoring terminal during the random inspection; analyzing the random inspection data based on the random inspection task status of the power materials, and outputting the status and / or alarm information of the power materials to complete the monitoring of the power materials during the random inspection process. By monitoring the random inspection data generated during the random inspection of power materials, and combining it with the random inspection task status of the power materials, the random inspection data is analyzed to determine the status of the power materials and output the corresponding status and alarm information, thereby realizing the monitoring of each stage of the random inspection of power materials and improving the efficiency of random inspection of power materials.

Description

Technical Field

[0001] This invention belongs to the technical field of material monitoring, specifically relating to a multi-dimensional monitoring method, device, and terminal based on random sampling of power materials. Background Technology

[0002] In the process of power grid construction, the foundation and prerequisite for ensuring the safety and stable operation of the power system is the quality of materials. Doing a good job in material quality supervision and special spot checks is of great importance. In particular, the distribution link involves a wide variety of materials in large quantities, resulting in many suppliers with different quality control levels. Therefore, the spot check of distribution materials is especially important.

[0003] The material sampling inspection process involves different operational stages such as sampling, sample packaging, sample delivery and transportation, and sample receipt. Due to the lack of sound technical control measures, the process mainly relies on ordinary packaging methods such as QR codes and RFID tags to store and protect sensitive sample information. The personnel involved in the delivery process cannot be controlled, which poses a series of risks such as the destruction or substitution of sealed samples. This affects the accuracy and impartiality of the material sampling inspection results and leaves hidden dangers for the safe operation of power grid equipment.

[0004] Currently, monitoring of power materials focuses on security and warehousing, without analyzing and detecting problems encountered during random inspections. Patent CN110211318A discloses a fully automated integrated power material monitoring system with functions for power material detection, analysis, ventilation control, and automatic entry and exit. This system includes a power supply module, a main control module, an information storage module, a temperature and humidity detection module, a power material conveying module, a temperature and humidity regulation module, an anti-theft monitoring module, an entry and exit module, an alarm module, and a display module. The power supply module provides power to the warehouse monitoring system. The temperature and humidity detection module collects temperature and humidity data and transmits it back to the main control module. The anti-theft monitoring module collects signals and transmits them back to the main control module. The main control module receives, processes, and outputs the data to the information storage module, display module, and alarm module. By employing the temperature and humidity detection module, anti-theft monitoring module, main control module, and temperature and humidity regulation module, the system ensures that power materials are stored in a suitable condition, guarantees the safety of power material storage, saves manpower, and improves work efficiency.

[0005] Therefore, how to monitor each stage of the sampling inspection process for power materials and improve the efficiency of the sampling inspection is a problem that needs to be solved. Summary of the Invention

[0006] To address the shortcomings of the existing technology, this invention provides a multi-directional monitoring method, device, and terminal based on power material sampling inspection. The method includes: real-time acquisition of material sampling inspection data, wherein the material sampling inspection data is data collected by a multi-directional monitoring terminal during the sampling inspection of power materials, and the power materials are fixed on the multi-directional monitoring terminal during the sampling inspection; analyzing the material sampling inspection data based on the sampling task status of the power materials, and outputting the power material status and / or alarm information to complete the monitoring of the power materials during the sampling inspection process. By monitoring the material sampling inspection data generated during the sampling inspection of power materials, and combining it with the sampling task status of the power materials, the method analyzes the material sampling inspection data, determines the status of the power materials, and outputs the corresponding power material status and alarm information, thereby realizing the monitoring of each stage of the power material sampling inspection process and improving the efficiency of power material sampling inspection.

[0007] Firstly, the present invention provides a multi-faceted monitoring method based on random inspections of power materials, specifically including the following steps:

[0008] Real-time acquisition of material sampling inspection data, which is data collected by multi-directional monitoring terminals during the sampling inspection of power materials, and the power materials are fixed on the multi-directional monitoring terminals during the sampling inspection.

[0009] Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed, and the status and / or alarm information of the power materials are output to complete the monitoring of the power materials during the sampling inspection process.

[0010] Furthermore, the material sampling inspection data includes pressure switch status and light intensity data;

[0011] Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including:

[0012] If the sampling inspection task status of power materials is "seal sealing completed", then the pressure switch status and light intensity data are judged.

[0013] If the pressure switch is in the raised state or the light intensity data reaches the light sensing threshold, output the removal alarm information and the location information of the power materials.

[0014] If the pressure switch is in a down state and the light intensity data is less than the light threshold, the status of the power materials will be output based on the preset sealing time interval.

[0015] Furthermore, based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the power material status and / or alarm information, specifically including:

[0016] If the sampling inspection task for power materials is in the sample delivery stage, then the status of the pressure switch and the light intensity data are used to make a judgment.

[0017] If the pressure switch is in the raised state or the light intensity data reaches the light sensing threshold, output the removal alarm information and the location information of the power materials.

[0018] If the pressure switch is in a down state and the light intensity data is less than the light threshold, the power material status is output based on the preset sampling time interval.

[0019] Furthermore, the material sampling inspection data includes acceleration data and positioning data;

[0020] Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including:

[0021] If the sampling inspection task status of power materials is "seal sealing completed", then the acceleration data and positioning data are judged.

[0022] If the acceleration data exceeds the velocity threshold or the positioning data exceeds the preset sealing area, the output will include a position deviation alarm and the location information of the power materials.

[0023] If the acceleration data is less than the velocity threshold and the positioning data is within the preset sealing area, the status of the power materials is output based on the preset sealing time interval.

[0024] Furthermore, the material sampling inspection data includes vibration time-series data;

[0025] Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including:

[0026] Based on the status of the random inspection of power materials, the vibration time series data is denoised to obtain standard vibration time series data.

[0027] Based on a preset silence threshold, the standard vibration time series data is evaluated, and the evaluation results are given.

[0028] Based on the judgment results, the standard vibration time series data is windowed to obtain multiple standard vibration time series sub-data.

[0029] By integrating the frequency domain and time domain features from multiple standard vibration time series data, the vibration characteristics of the first target are determined.

[0030] Based on a pre-built vibration analysis model, and combining the vibration characteristics of the first target and the vibration characteristics of the second target, the status of power materials is given. The vibration characteristics of the second target are obtained through a pre-built time-series analysis model.

[0031] Based on the analysis of the status of power equipment, alarm information is output.

[0032] Furthermore, the first target vibration characteristic includes at least one of waveform factor, zero-crossing rate, and maximum spectral amplitude;

[0033] By fusing frequency and time domain features from multiple standard vibration time series data, the vibration characteristics of the first target are determined, specifically including:

[0034] The fluctuations of multiple standard time series sub-data in the time domain are analyzed, and the time domain characteristics are given;

[0035] The proportion of multiple standard time series sub-data in different frequency bands in the frequency domain is analyzed, and frequency domain characteristics are given;

[0036] Based on the time-domain and frequency-domain characteristics, provide at least one of the waveform factor, zero-crossing rate, and maximum spectral amplitude.

[0037] Furthermore, based on the pre-constructed vibration analysis model, and combining the vibration characteristics of the first and second targets, the status of the power materials is given, specifically including:

[0038] The vibration characteristics of the first target and the vibration characteristics of the second target are spliced ​​and normalized to obtain the comprehensive vibration characteristics.

[0039] Based on a pre-built vibration analysis model, a time-series analysis of the comprehensive vibration characteristics is performed to obtain the time-series analysis results;

[0040] Based on the time series analysis results, the classification layer of the vibration analysis model is used to classify the comprehensive vibration characteristics and give the status of power materials.

[0041] Furthermore, the pre-constructed vibration analysis model is determined in the following way:

[0042] Obtain the vibration training dataset, in which each data point is labeled with a vibration event label;

[0043] Construct the first memory network layer and the second memory network layer, and pre-train them using the vibration training dataset until convergence to obtain the vibration pre-analysis model;

[0044] The model parameters of the first and second memory network layers in the vibration pre-analysis model are frozen. Combined with the loss function, the fully connected layer and the classification layer are trained until convergence, thus obtaining the vibration analysis model.

[0045] Secondly, the present invention also provides a multi-directional monitoring device based on random inspection of power materials, employing any of the above-mentioned multi-directional monitoring methods based on random inspection of power materials, including:

[0046] The data acquisition unit is used to acquire material sampling inspection data in real time. The material sampling inspection data is the data collected by the multi-directional monitoring terminal during the sampling inspection of power materials. The power materials are fixed on the multi-directional monitoring terminal during the sampling inspection.

[0047] The data analysis unit is used to analyze the sampling inspection data of power materials based on the sampling inspection task status, and output the status and / or alarm information of power materials to complete the monitoring of power materials during the sampling inspection process.

[0048] Thirdly, the present invention also provides a multi-directional monitoring terminal based on spot checks of power materials, including a controller module, a pressure switch, a mobile communication module, a positioning module, a light sensor and a vibration sensor, wherein the pressure switch, the mobile communication module, the positioning module, the light sensor and the vibration sensor are all connected to the controller module;

[0049] Pressure switch, used to obtain the pressure switch status;

[0050] The mobile communication module is used to establish a communication channel with the controller module.

[0051] The positioning module is used to obtain the location information of power materials and output positioning data;

[0052] A light sensor is used to acquire light intensity data;

[0053] Vibration sensors are used to acquire acceleration data and vibration time-series data.

[0054] The controller module is used to acquire pressure switch status, positioning data, light intensity data, acceleration data, and vibration timing data, and execute a multi-directional monitoring method based on spot checks of power materials, as described above.

[0055] The present invention provides a multi-directional monitoring method, device, and terminal based on random inspection of power materials, which has at least the following beneficial effects:

[0056] (1) By monitoring the sampling data of power materials generated during the sampling process, and combining the sampling task status of power materials, the sampling data of power materials is analyzed, the status of power materials is judged, and the corresponding power material status and alarm information are output, so as to realize the monitoring of each link of power materials in the sampling process and improve the sampling efficiency of power materials.

[0057] (2) By directly splicing the first vibration target features and the second vibration target features, the vibration analysis model becomes more dependent on the comprehensive vibration features, thereby improving the ability to distinguish similar events (such as emergency braking and human collisions).

[0058] (3) By adopting a random forest model based on long short-term memory network, LSTM and random forest are combined. At the same time, the first target vibration feature and the second target vibration feature are fused. The first target vibration feature is a statistical feature that is sensitive to certain transient features. The second target vibration feature is obtained through LSTM. LSTM can capture dynamic changes in time, that is, the order or duration of events. The combination of the first target vibration feature and the second target vibration feature can describe the data more comprehensively, provide a more comprehensive basis for the random forest to make judgments, and improve the accuracy of classification. Attached Figure Description

[0059] Figure 1 This is a structural block diagram of a multi-directional monitoring terminal based on spot checks of power materials provided in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the structure of a multi-directional monitoring terminal based on spot checks of power materials provided in an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of the casing of a multi-directional monitoring terminal based on spot checks of power materials provided in an embodiment of the present invention;

[0062] Figure 4 This is a structural diagram of power equipment with a flat mounting surface and a multi-directional monitoring terminal provided in an embodiment of the present invention;

[0063] Figure 5 A schematic diagram of the structure of the power materials including the cylindrical surface and the multi-directional monitoring terminal provided for an embodiment of the present invention;

[0064] Figure 6 This is a structural schematic diagram of power materials including a transfer box cover and a multi-directional monitoring terminal provided in an embodiment of the present invention;

[0065] Figure 7 A flowchart of a multi-faceted monitoring method based on spot checks of power materials provided in an embodiment of the present invention;

[0066] Figure 8 A flowchart for determining the status of power materials using vibration time-series data, provided as an embodiment of the present invention;

[0067] Figure 9 A flowchart for determining the vibration characteristics of a first target provided in an embodiment of the present invention;

[0068] Figure 10 A flowchart for determining the status of power materials provided in an embodiment of the present invention;

[0069] Figure 11 A flowchart for determining the vibration analysis model provided in an embodiment of the present invention;

[0070] Figure 12 A diagram illustrating the architecture of the vibration analysis model provided in this embodiment of the invention;

[0071] Figure 13 The structural block diagram of the multi-directional monitoring device based on spot checks of power materials provided in the embodiments of the present invention is shown.

[0072] Among them, 201 is the data acquisition unit; and 202 is the data analysis unit. Detailed Implementation

[0073] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0074] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0075] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0076] In order to achieve monitoring at each stage of the power material sampling inspection process and improve the efficiency of power material sampling inspection, refer to Figure 1This invention provides a multi-directional monitoring terminal based on random inspection of power materials, including a controller module, a pressure switch, a mobile communication module, a positioning module, a light sensor, and a vibration sensor. The pressure switch, mobile communication module, positioning module, light sensor, and vibration sensor are all connected to the controller module. The pressure switch is used to acquire its status. The mobile communication module is used to establish a communication channel with the controller module. The positioning module is used to obtain the location information of the power materials and output positioning data. The light sensor is used to acquire light intensity data. The vibration sensor is used to acquire acceleration data and vibration time-series data. The controller module is used to acquire the pressure switch status, positioning data, light intensity data, acceleration data, and vibration time-series data, and executes the multi-directional monitoring method based on random inspection of power materials provided in this invention.

[0077] In a specific example, refer to Figure 2 The controller module uses a low-power STM32L151 microcontroller chip. The pressure switch in the multi-directional monitoring terminal is an XKB5858 pressure push-button switch, the light sensor is a GL5506 photoresistor, the vibration sensor is an ADXL345 triaxial accelerometer, the positioning module is a B305-5Q single-BeiDou, and the mobile communication module is an AIR780EG 4G mobile communication module. In other examples, to ensure the security of the communication channel established by the controller module through the mobile communication module, the multi-directional monitoring terminal uses an NRSEC3000 security encryption communication chip to encrypt the data transmitted through the mobile communication module's communication channel.

[0078] The controller module uses a GPIO interface to monitor the pressure switch status in real time, and an ADC interface to acquire the light intensity data detected by the light sensor in real time. 2 The C interface acquires triaxial impact acceleration and tilt angle values ​​from vibration sensors in real time. It connects to the positioning module via a UART interface to receive BeiDou satellite positioning latitude and longitude and movement speed information. It interacts with a secure encrypted communication chip via an SPI interface to encrypt and decrypt communication data between the multi-directional monitoring terminal and the monitoring management platform. It also interacts with the mobile communication module via a UART interface to achieve HTTP protocol communication between the multi-directional monitoring terminal and the monitoring management platform.

[0079] During the operation of the multi-directional monitoring terminal, the monitoring management platform switches the sampling task status according to the sampling inspection business process of power materials. The administrator of the monitoring management platform can manually control the termination of the sampling task at any time. Before each time the multi-directional monitoring terminal uploads information to the monitoring management platform, it requests the sampling task status via HTTP protocol and performs different logical processing on the detected data according to the current sampling task status.

[0080] In the example provided in this invention, refer to Figure 3 To facilitate the binding of the multi-directional monitoring terminal to power equipment, it is designed with a 60mm*50mm ABS material shell. The shell surface features an NFC identification area, a pressure switch, a power indicator light, and a photosensitive area. To ensure dust and water resistance, a dustproof and waterproof film is affixed to the shell surface. In other implementation examples, the material or size of the multi-directional monitoring terminal can be adjusted according to the specific needs of the power equipment; this is not limited.

[0081] The multi-directional monitoring terminal, combined with auxiliary tooling, is applicable to the installation and construction of all power distribution materials, including high and low voltage switchgear, JP cabinets, cable branch boxes, disconnect switches, ring main units, prefabricated substations, power metering boxes, distribution transformers, voltage transformers, current transformers, surge arresters, line insulators, overhead power cables, power cables, cable protection pipes, cement pillars, hardware, etc.

[0082] Reference Figure 4 For cabinet-type materials such as high and low voltage switchgear, JP cabinet, cable branch box, high disconnect switch, ring main unit, box-type substation, and power metering box, as well as power materials with flat mounting surfaces such as distribution transformer, voltage transformer, and current transformer, double-sided tape is used to directly attach multi-directional monitoring terminals to the samples to achieve monitoring of power materials during the sampling inspection process.

[0083] Reference Figure 5 For power materials with cylindrical surfaces such as surge arresters, line insulators, overhead power cables, power cables, cable protection pipes, and cement pillars, base-attached cable ties are used to bind them to the samples, and then 3M double-sided tape is used to attach the multi-directional monitoring terminal to the base of the cable tie.

[0084] Reference Figure 6 Customized transfer boxes are used to pack line fittings of various shapes and sizes. The box lid is then locked with cable ties with bases, and 3M double-sided tape is used to attach the multi-directional monitoring terminal to the cable tie base.

[0085] This invention provides a multi-directional monitoring terminal for power material sampling inspection. Applied during the power material sampling inspection process, the monitoring terminal integrates multiple monitoring technologies based on principles such as pressure switch status detection, light intensity measurement, triaxial impact vibration measurement, and high-precision BeiDou positioning, as well as IoT communication encryption technology. When the sample is sealed, the monitoring terminal is bound to the sample, enabling real-time monitoring and alarming of the sample's sealing status, position, and attitude after sealing, during sample delivery, and before sample reception. This prevents illegal acts such as intentional dismantling or substitution of samples, solving the problems of violations and quality risks in the power material sampling inspection process while also improving the efficiency of power material sampling inspection.

[0086] like Figure 7 As shown in the figure, this embodiment of the invention provides a multi-directional monitoring method based on random inspection of power materials, and the specific steps are as follows:

[0087] S101: Real-time acquisition of material sampling inspection data.

[0088] Specifically, the material sampling inspection data refers to the data collected by the multi-directional monitoring terminal during the sampling inspection of power materials, which are fixed on the multi-directional monitoring terminal during the sampling inspection.

[0089] S102: Based on the status of the power material sampling inspection task, analyze the material sampling inspection data and output the power material status and / or alarm information to complete the monitoring of power materials during the sampling inspection process.

[0090] In the first specific implementation, the material sampling inspection data includes pressure switch status and light intensity data. If the sampling inspection task status of the power materials is "sealing completed," the pressure switch status and light intensity data are assessed. If the pressure switch status is "raised" or the light intensity data reaches the light sensing threshold, it indicates that the multi-directional monitoring terminal may have been removed from the power material sample during the sealing process. A removal alarm message and the location information of the power materials are output to alert personnel of the anomaly and indicate the location requiring handling. If the pressure switch status is "lowered" and the light intensity data is less than the light sensing threshold, it indicates that the multi-directional monitoring terminal is working normally during the sealing process and no anomaly has occurred. Combined with the preset sealing time interval, the power material status is output. The aforementioned light sensing threshold is set based on factors such as the actual power materials being sampled and the sampling environment. The aforementioned sealing time interval is set according to the actual scenario and is not limited. In this example, the sealing time interval is 1 hour.

[0091] In the second specific implementation, if the sampling inspection task for power materials is in the sampling stage, the pressure switch status and light intensity data are assessed. If the pressure switch is raised or the light intensity data reaches the light sensing threshold, it indicates that the multi-directional monitoring terminal may have been removed from the power material sample during sampling. A removal alarm and the location information of the power material are output to alert personnel of the anomaly and indicate the location requiring handling. If the pressure switch is lowered and the light intensity data is less than the light sensing threshold, it indicates that the multi-directional monitoring terminal is working normally during sampling and no anomaly has occurred. The power material status is output based on the preset sampling time interval. The sampling time interval is set according to the actual scenario and is not limited thereto. In this example, the sealing time interval is 2 minutes.

[0092] In the third specific implementation, the material sampling inspection data includes acceleration data and positioning data. If the sampling inspection task status of the power materials is "sealing completed," the acceleration data and positioning data are evaluated. If the acceleration data exceeds the speed threshold or the positioning data exceeds the preset sealing area, it indicates that the power materials have deviated from the sealing area during the sealing process, which will have a certain impact on subsequent processes. A position deviation alarm message and the position information of the power materials are output to remind staff that the position of the power materials has shifted and needs adjustment, and to provide the necessary handling information. If the acceleration data is less than the speed threshold and the positioning data is within the preset sealing area, it indicates that the power materials have not deviated from the sealing area during the sealing process. Combined with the preset sealing time interval, the status of the power materials is output.

[0093] Understandably, the sampling task status sequentially progresses through: Idle, Sampling Start, Sealing Completed, Sample Delivery Preparation, Sample Delivery Started, Sample Delivery Completed, Sample Receiving Started, Sample Receiving Completed, and finally returns to the Idle sampling task status. When the sampling task status is "Idle," the multi-directional monitoring terminal uploads terminal information and location information every 24 hours. When the sampling task status is "Sampling Started," if the multi-directional terminal detects that the pressure switch has been pressed for 2 seconds and then rises, it is determined that the user has actively triggered the multi-directional monitoring terminal to upload relevant information. When the sampling task status is "Sealing Completed," the sample of the power equipment bound to the multi-directional monitoring terminal is stored in the sealing area. If the multi-directional monitoring terminal detects that the pressure switch has risen or the light intensity data exceeds the light sensing threshold, it is determined that the multi-directional monitoring terminal has been removed from the sample of the power equipment, and immediately triggers the upload of removal alarm information and the location information of the power equipment. If the multi-directional monitoring terminal detects that the impact acceleration amplitude or tilt angle exceeds the threshold, or that the positioning location is far from the sealing area, it determines that the sample has been moved. This means the acceleration data exceeds the velocity threshold or the positioning data exceeds the preset sealing area range, immediately triggering the upload of a position deviation alarm and the location information of the power materials. When the sampling task status is "Sealing Completed," the multi-directional monitoring terminal is attached to the power material sample and stored in the sealing area. If the multi-directional monitoring terminal does not detect any abnormalities, it uploads terminal information and positioning information every hour. When the sampling task status is "Sample Preparation," "Sample Start," "Sample Completed," or "Sample Receiving Start," the multi-directional monitoring terminal is attached to the power material sample. If the multi-directional monitoring terminal detects that the pressure switch has risen or the light intensity data exceeds the light sensing threshold, it determines that the multi-directional monitoring terminal has been removed from the power material sample, immediately triggering the upload of a removal alarm and positioning information. When the sampling task status is "Sample Preparation", "Sample Start", "Sample Completion", or "Sample Receiving Start", the multi-directional monitoring terminal is attached to the sample of the power material for transportation. If the terminal does not detect any abnormalities in the pressure switch and light intensity data, it will send terminal information and location information every 2 minutes. When the sampling task status is "Sample Receiving Complete", it will automatically switch to "Idle", indicating that the previous power material has been sampled.

[0094] To differentiate between normal vibrations (such as vibrations during transportation, sharp turns, and braking) and abnormal impacts (such as damage, replacement, drops, and violent collisions) of power materials during sampling and acceptance, improve identification accuracy, reduce false alarm rates at multi-directional monitoring terminals, and enhance reliability, an intelligent analysis module was deployed on the multi-directional monitoring terminal or monitoring management platform during the state analysis of power materials. This module works in conjunction with a time-series feature extraction module to analyze and process the vibration time-series data collected by the multi-directional monitoring terminal, achieving accurate identification of normal vibrations and abnormal impacts of the sampled materials. Figure 8 Specifically, it includes:

[0095] Based on the status of the random inspection of power materials, the vibration time series data is denoised to obtain standard vibration time series data.

[0096] Based on a preset silence threshold, the standard vibration time series data is evaluated, and the evaluation results are given.

[0097] Based on the judgment results, the standard vibration time series data is windowed to obtain multiple standard vibration time series sub-data.

[0098] By integrating the frequency domain and time domain features from multiple standard vibration time series data, the vibration characteristics of the first target are determined.

[0099] Based on a pre-built vibration analysis model, and combining the vibration characteristics of the first target and the vibration characteristics of the second target, the status of power materials is given. The vibration characteristics of the second target are obtained through a pre-built time-series analysis model.

[0100] Based on the analysis of the status of power equipment, alarm information is output.

[0101] In the example provided in this invention, the time series analysis model is a long short-term memory network model. Alternatively, the LSTM component of the vibration analysis model can be used to obtain the vibration characteristics of the second target; there is no limitation on this.

[0102] Furthermore, referring to Figure 9 The first target vibration characteristic includes at least one of waveform factor, zero-crossing rate, and maximum spectral amplitude;

[0103] By fusing frequency and time domain features from multiple standard vibration time series data, the vibration characteristics of the first target are determined, specifically including:

[0104] The fluctuations of multiple standard time series sub-data in the time domain are analyzed, and the time domain characteristics are given;

[0105] The proportion of multiple standard time series sub-data in different frequency bands in the frequency domain is analyzed, and frequency domain characteristics are given;

[0106] Based on the time-domain and frequency-domain characteristics, provide at least one of the waveform factor, zero-crossing rate, and maximum spectral amplitude.

[0107] In the fourth specific implementation, the material sampling inspection data includes vibration time-series data, which is acquired by a triaxial accelerometer (ADXL345) in a multi-directional monitoring terminal, collecting XYZ axis vibration time-series data. In the example provided by this invention, the sampling frequency of the aforementioned triaxial accelerometer is 200Hz, and the range is set to ±16g. By setting the sampling frequency, the triaxial accelerometer can balance high-frequency vibration capture with power consumption; by setting the range, it can cover the maximum possible impact during sample delivery and reception. The data generated by the triaxial accelerometer is a time series, requiring filtering and noise reduction of the vibration time-series data, sliding window segmentation, and then extraction of statistical features to obtain the first target vibration feature.

[0108] In the example provided by this invention, a low-pass filter is used to remove high-frequency noise, such as sensor jitter, from the vibration time-series data to obtain standard vibration time-series data. The standard vibration time-series data is divided into 5-second windows (500 sampling points / axis) with a 50% overlap rate to ensure that no continuous events are missed and to reduce computational resource consumption. Window segmentation is activated only when the peak acceleration exceeds a quiet threshold (e.g., 0.5g); in idle states, windows are not segmented to save computational resources. When the peak acceleration exceeds the quiet threshold, it indicates a possible abnormal situation, requiring more accurate data analysis. When the peak acceleration does not exceed the quiet threshold, it indicates that the sample of the electrical material is stationary or the transportation is relatively stable. The aforementioned quiet threshold can be set according to the actual scenario and is not limited thereto.

[0109] After windowing the standard vibration time series data, multiple standard vibration time series sub-data sets are obtained. For a specific standard vibration time series sub-data set, the following processing is performed:

[0110] For time-domain features, the mean, variance, peak value, peak-to-peak value, and root mean square are calculated. The X-axis, Y-axis, and Z-axis each have five dimensions, for a total of 15 dimensions.

[0111] For the frequency domain characteristics, the energy proportions of the low-frequency band and the high-frequency band are calculated after Fast Fourier Transform. The low-frequency energy proportion represents normal vibration, and the high-frequency energy proportion represents abnormal impact. Each axis is 2-dimensional, for a total of 6 dimensions. In this example, the range from 0 Hz to 50 Hz is defined as the low-frequency band, and the range from 50 Hz to 200 Hz is defined as the high-frequency band.

[0112] For the time-frequency joint characteristics, waveform factor, zero-crossing rate, and maximum spectral amplitude are calculated. Each axis is 3-dimensional, for a total of 9 dimensions. The waveform factor is the ratio of the mean to the root mean square, reflecting the vibration level of the electrical material sample. The zero-crossing rate (ZCR) is the number of times the signal crosses a zero point per unit time, i.e., the frequency at which the signal changes from positive to negative or vice versa. The maximum spectral amplitude is the maximum spectral amplitude after the standard vibration time series data has undergone a Fast Fourier Transform.

[0113] Specifically, it is expressed as follows:

[0114]

[0115] Peak = max(|X i |)

[0116] PtoP = max(X) i )-min(X i )

[0117]

[0118] Where μ is the mean, σ 2 X is the variance, Peak is the peak value, PtoP is the peak-to-peak value, RMS is the root mean square error, N is the total number of sampling points in each standard vibration time series subdata, and X is the root mean square error. i This represents the acceleration value at the i-th sampling point in each standard vibration time series subdata.

[0119] The peak value is the maximum absolute value of acceleration within each window. The peak-to-peak value is the difference between the maximum and minimum acceleration values ​​within each window. The root mean square (RMS) represents the average energy intensity of the signal.

[0120] Perform a Fast Fourier Transform on each standard vibration time series data segment to obtain the spectral amplitude |X(f)|, where f is the signal frequency.

[0121]

[0122] Wherein, LE represents the proportion of low-frequency energy corresponding to each standard vibration time series data, and HE represents the proportion of high-frequency energy corresponding to each standard vibration time series data.

[0123] The waveform factor ρ is specifically expressed as: ρ = μ / RMS

[0124] The zero-crossing rate ZCR is specifically expressed as:

[0125]

[0126] To eliminate the dimensional differences between various features, and considering the data characteristics of different features, different standardization methods are used to process each feature.

[0127] For mean, variance, peak value, peak-to-peak value, and root mean square, Min-Max standardization is used for data processing. For low-frequency energy proportion, high-frequency energy proportion, waveform factor, zero-crossing rate, and maximum spectral amplitude, Z-Score standardization is used for data processing.

[0128] Furthermore, referring to Figure 10 Determine the status of power equipment, specifically including:

[0129] The vibration characteristics of the first target and the vibration characteristics of the second target are spliced ​​and normalized to obtain the comprehensive vibration characteristics.

[0130] Based on a pre-built vibration analysis model, a time-series analysis of the comprehensive vibration characteristics is performed to obtain the time-series analysis results;

[0131] Based on the time series analysis results, the classification layer of the vibration analysis model is used to classify the comprehensive vibration characteristics and give the status of power materials.

[0132] In one specific implementation, the vibration pre-analysis model is a random forest model based on Long Short-Term Memory (LSTM) networks. LSTM is used to extract the temporal features of the vibration data, i.e., the second target vibration features, which are then fused with the first target vibration features and used as input to the random forest classifier for vibration mode classification and identification.

[0133] By setting a first memory network layer, a complete time-step sequence is returned; by setting a second memory network layer, the final time-step output is returned. A fully connected layer is set to output the second target vibration feature. The dropout layer in the fully connected layer sets a dropout rate to avoid overfitting. In the example provided in this invention, the first memory network layer has 64 LSTM units to return the complete time-step sequence. The second memory network layer has 128 LSTM units to return the final time-step output. The dropout layer has a dropout rate of 0.3 to prevent overfitting. The fully connected layer outputs the final temporal feature vector to characterize the temporal pattern of the vibration signal.

[0134] The LSTM component, comprising a first memory network layer, a second memory network layer, and a fully connected layer, can process time-series vibrational data and capture long-term dependencies. It performs a nonlinear transformation on synchronously input vibrational time-series data within a segmented window using multiple layers of neurons, outputting a fixed-length vector (128 dimensions), where each dimension represents an abstract time-series feature.

[0135] LSTM excels at processing time-series data, capturing long-term dependencies and dynamic changes. For example, in transportation, sudden braking might manifest as a continuous acceleration change, while a human collision might be a sudden high-frequency vibration event. LSTM can identify these temporal patterns through time-step processing, thus providing richer feature representations.

[0136] After concatenating the first and second vibration target features, Z-score is applied for standardization to obtain a comprehensive vibration feature, which serves as the input to the classification layer. Standardizing the concatenated first and second vibration target features ensures consistency across different feature scales, avoiding bias in random forests due to differences in feature dimensions.

[0137] By directly splicing the first vibration target features and the second vibration target features, the vibration analysis model becomes more dependent on the comprehensive vibration features, thereby improving its ability to distinguish similar events (such as sudden braking and human-caused collisions).

[0138] By setting parameters such as the number of trees in the classification layer, maximum depth, and class weights in a random forest classifier, and combining these parameters with comprehensive vibration characteristics, the classification probability is output to obtain the status of power materials.

[0139] Furthermore, referring to Figure 11 The pre-constructed vibration analysis model is determined in the following way:

[0140] Obtain the vibration training dataset, in which each data point is labeled with a vibration event label;

[0141] Construct the first memory network layer and the second memory network layer, and pre-train them using the vibration training dataset until convergence to obtain the vibration pre-analysis model;

[0142] The model parameters of the first and second memory network layers in the vibration pre-analysis model are frozen. Combined with the loss function, the fully connected layer and the classification layer are trained until convergence, thus obtaining the vibration analysis model.

[0143] In one specific implementation, historical data is used to construct the vibration training dataset required for model training. For example, raw vibration data collected by triaxial accelerometers are extracted from historical random inspection tasks of power materials. Samples containing typical events are selected first, including samples of smooth transportation, sharp turns, and braking during normal vibration, and samples of falling (large instantaneous acceleration in the vertical direction), collision (multi-axis high-frequency oscillation), and human-caused damage (abnormal continuous periodic bumps, instantaneous impacts, high-frequency vibrations, and violent vibrations during dismantling) during abnormal impacts.

[0144] The data in the vibration training dataset are labeled, with normal vibration labeled as "0" and abnormal impact labeled as "1". The vibration training dataset is then divided into training, validation, and test sets according to a certain proportion based on the time series.

[0145] The vibration analysis model adopts a phased training strategy. First, the first memory network layer, the second memory network layer, and the fully connected layer are pre-trained with LSTM, and then the classification layer is combined for joint fine-tuning training.

[0146] During the LSTM pre-training phase, hyperparameters, including batch size and training period, need to be set first. A portion of the vibration training dataset is used for LSTM pre-training to learn the long-term dependencies and dynamic patterns of vibration signals in the power material delivery and receiving process, providing a good initial state for the LSTM model parameters.

[0147] After completing the pre-training, the model parameters of the first and second memory network layers in the vibration pre-analysis model are frozen, that is, the model parameters of the first and second memory network layers are retained, and only the fully connected layer and the classification layer are trained, where the classification layer uses a random forest classifier.

[0148] The loss function for the vibration analysis model is constructed using Focal Loss. This reduces the loss weight of easily distinguishable vibration samples and increases the loss weight of difficult-to-distinguish vibration samples, making the model pay more attention to those difficult-to-distinguish samples and optimizing the model's ability to identify abnormal impacts.

[0149] Loss function FL(p) t Specifically, it is expressed as:

[0150] FL(p t )=-α t (1-p t ) γ log(p t )

[0151]

[0152] Where, p t This represents the predicted probability of the power equipment's condition by the vibration analysis model. y = 1 indicates an abnormal impact, y = 0 indicates normal vibration, p is the probability that the power equipment's condition is due to an abnormal impact, and α is the probability that the condition is due to an abnormal impact. t γ is the balancing factor, used to adjust the weights of positive and negative samples, while γ is the adjustment factor, used to control the weight difference between easy and difficult samples. In this example, the adjustment factor is preset to 2.

[0153] In other implementations, after training the vibration analysis model, the trained model is evaluated using data from the test set. In this example, the F1 score is used as the model evaluation metric, specifically expressed as follows:

[0154]

[0155] Among them, F1 _ Score is the F1 score, TP is the number of correctly identified abnormal impact samples, FP is the number of false alarms of normal vibration samples, and FN is the number of missed abnormal impact samples.

[0156] A higher F1 score indicates that the vibration analysis model achieves a better balance between precision and recall, and is more accurate in predicting positive categories.

[0157] If the F1 score is outside the preset model evaluation range, it indicates that the training effect of the vibration analysis model is limited and further optimization is needed. Optimization of the vibration analysis model can be achieved through multiple training iterations, or by combining the loss function and the F1 score, using Bayesian search to optimize the hyperparameters of the vibration analysis model. For example, optimizing the number of LSTM units (search range 32-128), the number of random forest trees (search range 100-300), and the maximum depth (search range 10-50), etc. Simultaneously, optimizing the optimal combination of the balance factor and adjustment factor in the loss function is also crucial.

[0158] Reference Figure 12 This invention obtains the first target vibration feature by performing noise reduction filtering, sliding window segmentation, and statistical feature extraction on the collected material sampling data. At the same time, an LSTM model is used to perform time series analysis on the material sampling data and extract time series features to give the second target vibration feature. The first target vibration feature and the second target vibration feature are fused to obtain the comprehensive vibration feature. Finally, the comprehensive vibration feature is input into a random forest classifier to obtain the corresponding power material status, that is, whether the power material corresponding to the material sampling data is in a normal vibration state or an abnormal impact state.

[0159] This invention employs a random forest model based on a long short-term memory network, combining LSTM and random forest. Simultaneously, it fuses a first target vibration feature and a second target vibration feature. The first target vibration feature is a statistical feature, sensitive to certain transient features, while the second target vibration feature is obtained through LSTM. LSTM can capture dynamic changes over time, i.e., the order or duration of events. The combination of the first and second target vibration features can more comprehensively describe the data, providing a more comprehensive basis for the random forest to make judgments and improving the accuracy of classification.

[0160] For example, abnormal impacts (such as falls) may manifest as sudden peaks (statistical features) and sustained patterns of high-frequency oscillations (LSTM temporal features), and combining the two can improve classification accuracy. By combining temporal and statistical features, multimodal feature fusion can be achieved, improving anomaly detection accuracy and more accurately distinguishing similar events such as sudden braking from human collisions.

[0161] Reference Figure 13 This invention provides a multi-directional monitoring device based on random inspection of power materials, comprising:

[0162] The data acquisition unit 201 is used to acquire material sampling inspection data in real time. The material sampling inspection data is the data collected by the multi-directional monitoring terminal during the sampling inspection of power materials. The power materials are fixed on the multi-directional monitoring terminal during the sampling inspection.

[0163] The data analysis unit 202 is used to analyze the sampling inspection data of power materials based on the sampling inspection task status, and output the status and / or alarm information of power materials to complete the monitoring of power materials during the sampling inspection process.

[0164] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0165] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A multi-dimensional monitoring method based on power material sampling inspection, characterized in that, include: Real-time acquisition of material sampling inspection data, which consists of data collected by multi-directional monitoring terminals during the sampling inspection of power materials, which are fixed on the multi-directional monitoring terminals during the sampling inspection; the material sampling inspection data includes vibration time series data; Based on the sampling inspection status of power materials, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, thereby completing the monitoring of power materials during the sampling inspection process. Specifically, this includes: denoising the vibration time-series data according to the sampling inspection status to obtain standard vibration time-series data; judging the standard vibration time-series data based on a preset silence threshold and providing the judgment result; segmenting the standard vibration time-series data into multiple standard vibration time-series sub-data based on the judgment result; fusing the frequency domain and time domain features from multiple standard vibration time-series sub-data to determine the first target vibration feature; and based on a pre-constructed vibration analysis model, combining the first target vibration feature and the second target vibration feature... The vibration characteristics of the target power materials are used to determine their status. The second target vibration characteristics are obtained through a pre-constructed time series analysis model. The vibration analysis model adopts a phased training strategy. First, the first memory network layer, the second memory network layer, and the fully connected layer are pre-trained with LSTM, and then the classification layer is combined for joint fine-tuning training. Based on the analysis of the power material status, alarm information is output. The sampling task status switches according to the power material sampling business process. The multi-directional monitoring terminal requests the sampling task status via HTTP protocol. The sampling task status is sequentially: idle, sampling start, sample sealing complete, sample delivery preparation, sample delivery start, sample delivery complete, sample reception start, sample reception complete, and finally returns to the idle sampling task status.

2. The multi-aspect monitoring method based on power material spot check according to claim 1, characterized in that, Material sampling inspection data includes pressure switch status and light intensity data; Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including: If the sampling inspection task status of power materials is "seal sealing completed", then the pressure switch status and light intensity data are judged. If the pressure switch is in the raised state or the light intensity data reaches the light sensing threshold, output the removal alarm information and the location information of the power materials. If the pressure switch is in a down state and the light intensity data is less than the light threshold, the status of the power materials will be output based on the preset sealing time interval.

3. The multi-angled monitoring method based on the spot check of electric power materials according to claim 2, characterized in that, Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including: If the sampling inspection task for power materials is in the sample delivery stage, then the status of the pressure switch and the light intensity data are used to make a judgment. If the pressure switch is in the raised state or the light intensity data reaches the light sensing threshold, output the removal alarm information and the location information of the power materials. If the pressure switch is in a down state and the light intensity data is less than the light threshold, the power material status is output based on the preset sampling time interval.

4. The multi-directional monitoring method based on random inspection of power materials as described in claim 1, characterized in that, Material sampling inspection data includes acceleration data and positioning data; Based on the status of the power material sampling inspection task, the sampling inspection data is analyzed to output the status and / or alarm information of the power materials, specifically including: If the sampling inspection task status of power materials is "seal sealing completed", then the acceleration data and positioning data are judged. If the acceleration data exceeds the velocity threshold or the positioning data exceeds the preset sealing area, the output will include a position deviation alarm and the location information of the power materials. If the acceleration data is less than the velocity threshold and the positioning data is within the preset sealing area, the status of the power materials is output based on the preset sealing time interval.

5. The multi-directional monitoring method based on random inspection of power materials as described in claim 1, characterized in that, The first target vibration characteristic includes at least one of waveform factor, zero-crossing rate, and maximum spectral amplitude; By fusing frequency and time domain features from multiple standard vibration time series data, the vibration characteristics of the first target are determined, specifically including: The fluctuations of multiple standard time series sub-data in the time domain are analyzed, and the time domain characteristics are given; The proportion of multiple standard time series sub-data in different frequency bands in the frequency domain is analyzed, and frequency domain characteristics are given; Based on the time-domain and frequency-domain characteristics, provide at least one of the waveform factor, zero-crossing rate, and maximum spectral amplitude.

6. The multi-directional monitoring method based on random inspection of power materials as described in claim 1, characterized in that, Based on a pre-constructed vibration analysis model, and combining the vibration characteristics of the first and second targets, the status of power materials is given, specifically including: The vibration characteristics of the first target and the vibration characteristics of the second target are spliced ​​and normalized to obtain the comprehensive vibration characteristics. Based on a pre-built vibration analysis model, a time-series analysis of the comprehensive vibration characteristics is performed to obtain the time-series analysis results; Based on the time series analysis results, the classification layer of the vibration analysis model is used to classify the comprehensive vibration characteristics and give the status of power materials.

7. The multi-directional monitoring method based on random inspection of power materials as described in claim 6, characterized in that, The pre-constructed vibration analysis model is determined in the following way: Obtain the vibration training dataset, in which each data point is labeled with a vibration event label; Construct the first memory network layer and the second memory network layer, and pre-train them using the vibration training dataset until convergence to obtain the vibration pre-analysis model; The model parameters of the first and second memory network layers in the vibration pre-analysis model are frozen. Combined with the loss function, the fully connected layer and the classification layer are trained until convergence, thus obtaining the vibration analysis model.

8. A multi-directional monitoring device based on random inspection of power materials, characterized in that, The multi-directional monitoring method based on random inspection of power materials as described in any one of claims 1-7 includes: The data acquisition unit is used to acquire material sampling inspection data in real time. The material sampling inspection data is the data collected by the multi-directional monitoring terminal during the sampling inspection of power materials. The power materials are fixed on the multi-directional monitoring terminal during the sampling inspection. The material sampling inspection data includes vibration time series data. The data analysis unit is used to analyze the sampling inspection data of power materials based on the sampling inspection task status, and output the status and / or alarm information of the power materials to complete the monitoring of power materials during the sampling inspection process. Specifically, it includes: denoising the vibration time-series data according to the sampling inspection task status of the power materials to obtain standard vibration time-series data; judging the standard vibration time-series data based on a preset silence threshold and providing the judgment result; segmenting the standard vibration time-series data into multiple standard vibration time-series sub-data based on the judgment result; fusing the frequency domain and time domain features of multiple standard vibration time-series sub-data to determine the first target vibration feature; and based on a pre-built vibration analysis model, combining the first target vibration feature... The vibration characteristics of the first and second targets are used to determine the status of power materials. The vibration characteristics of the second target are obtained through a pre-constructed time series analysis model. The vibration analysis model adopts a phased training strategy. First, the first memory network layer, the second memory network layer, and the fully connected layer are pre-trained with LSTM, and then the classification layer is combined for joint fine-tuning training. Based on the analysis of the status of power materials, alarm information is output. The sampling task status switches according to the sampling business process of power materials. The multi-directional monitoring terminal requests the sampling task status through the HTTP protocol. The sampling task status is sequentially: idle, sampling start, sample sealing complete, sample delivery preparation, sample delivery start, sample delivery complete, sample reception start, sample reception complete, and finally returns to the idle sampling task status.

9. A multi-directional monitoring terminal based on random inspection of power materials, characterized in that, It includes a controller module, a pressure switch, a mobile communication module, a positioning module, a light sensor, and a vibration sensor. The pressure switch, mobile communication module, positioning module, light sensor, and vibration sensor are all connected to the controller module. Pressure switch, used to obtain the pressure switch status; The mobile communication module is used to establish a communication channel with the controller module. The positioning module is used to obtain the location information of power materials and output positioning data; A light sensor is used to acquire light intensity data; Vibration sensors are used to acquire acceleration data and vibration time-series data. The controller module is used to acquire pressure switch status, positioning data, light intensity data, acceleration data and vibration timing data, and execute the multi-directional monitoring method based on spot checks of power materials as described in any one of claims 1-7.

Images