Intelligent inspection method and device of track type inspection robot

Abstract

The present application relates to the technical field of robot inspection, in particular to an intelligent inspection method and device of a track type inspection robot, the method comprising: obtaining various monitoring data of a secondary node when the distance between the inspection robot and the secondary node is within a first preset range; obtaining predicted monitoring data of the secondary node through a preset distance attenuation model when the monitoring data meets a first preset abnormal early warning condition; obtaining the stability of the predicted monitoring data when the predicted monitoring data meets a second preset abnormal early warning condition; inputting the stability of various predicted monitoring data into multiple risk models to obtain multiple risk values, and performing risk early warning on the secondary node according to the maximum risk value, thereby avoiding missed inspection of the secondary node, improving inspection safety, improving overall inspection efficiency and monitoring flexibility, avoiding unnecessary energy consumption, and improving economic efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot inspection, in particular to an intelligent inspection method and device of a track-type inspection robot. BACKGROUND

[0002] When the track-type inspection robot is performing inspection, an electronic map is first established according to the Radio Frequency Identification (RFID) or visual label calibration of the to-be-measured task nodes (such as transformers, switch cabinets, contact network poles, etc.) on the track; thereafter, the background task system comprehensively considers the node priority, inspection cycle, shortest path and obstacle avoidance constraints to generate an optimal time route. In the operation of the track-type inspection robot, the robot stops at each point according to the planned path, collects relevant data of the corresponding node through the relevant sensors arranged, and sends the data to the background for monitoring. At present, when each to-be-measured task node in the planned path is inspected, the division rule of the to-be-measured task node is relatively single, and only two types of nodes, i.e., the nodes that need to be inspected and the nodes that do not need to be inspected, are divided. The nodes that need to be inspected are defined as main nodes, and the remaining nodes are defined as secondary nodes. Then, the existing inspection does not inspect the secondary nodes, which may ignore the risks of the secondary nodes, thereby affecting the safety of the inspection. SUMMARY

[0003] In order to solve the technical problem of low safety of the existing track-type inspection robot inspection method, the purpose of the present application is to provide an intelligent inspection method and device of a track-type inspection robot, and the technical solution adopted is as follows:

[0004] In the first aspect of the present application, an intelligent inspection method of a track-type inspection robot is provided, which comprises:

[0005] When the distance between the inspection robot and the secondary node is within a first preset range, various monitoring data of the secondary node are obtained;

[0006] When the monitoring data meets a first preset abnormality warning condition, a preset distance attenuation model is used to obtain predicted monitoring data of the secondary node when the distance between the inspection robot and the secondary node is within a second preset range;

[0007] When the predicted monitoring data meets a second preset abnormality warning condition, the stability of the predicted monitoring data is obtained;

[0008] The stability of various predicted monitoring data is input into a plurality of risk models to obtain a plurality of risk values, and a maximum risk value is determined;

[0009] The secondary node is warned according to the maximum risk value.

[0010] In an exemplary embodiment, after the acquisition of the various monitoring data of the secondary node, the intelligent inspection method of the track inspection robot further comprises:

[0011] The monitoring data is filtered through a low-pass filter to obtain an initial filtered signal;

[0012] The initial filtered signal is filtered using a pre-trained adaptive filter to obtain a regular component signal;

[0013] The initial filtered signal is subtracted from the regular component signal to obtain a target component signal.

[0014] In an exemplary embodiment, the monitoring data satisfies a first preset abnormality early warning condition, including that the monitoring data is outside a preset normal fluctuation range.

[0015] In an exemplary embodiment, the preset distance attenuation model includes an inverse square law model, a diffusion theory model, and an exponential attenuation model.

[0016] In an exemplary embodiment, the intelligent inspection method of the track inspection robot further comprises: acquiring an abnormality degree of the predicted monitoring data;

[0017] The predicted monitoring data satisfies a second preset abnormality early warning condition, including that the abnormality degree is greater than or equal to a preset abnormality degree threshold.

[0018] In an exemplary embodiment, the abnormality degree is obtained from the difference between the predicted monitoring data and a preset standard value.

[0019] In an exemplary embodiment, the stability of the predicted monitoring data is obtained from a negative correlation of a standard deviation of the predicted monitoring data.

[0020] In an exemplary embodiment, the risk model includes risk weights of various monitoring data;

[0021] The risk value acquisition process includes: based on the risk weights of various monitoring data of the risk model, weighting and summing the stability of various predicted monitoring data to obtain a risk value corresponding to the risk model.

[0022] In an exemplary embodiment, the risk early warning of the secondary node according to the maximum risk value comprises:

[0023] Comparing the maximum risk value with a preset risk threshold, if the maximum risk value is greater than or equal to the preset risk threshold, outputting a risk early warning signal of the secondary node.

[0024] In a second aspect of the present application, an intelligent inspection device of a track inspection robot is provided, comprising a memory and a processor; the memory is connected with the processor; the memory is used for storing program instructions; and the processor is used for implementing the intelligent inspection method of the track inspection robot when the program instructions are executed.

[0025] The present application has the following beneficial effects: in the operation process of the inspection robot approaching the secondary node, the secondary node is pre-inspected when the distance between them is a certain distance, and when the preset abnormal early warning condition is met, the prediction monitoring data when the distance to the secondary node is closer is obtained through the preset distance attenuation model, and a plurality of risk values are obtained based on the prediction monitoring data and in combination with a plurality of risk models, and the real risk of the secondary node is determined by the maximum risk value, so as to realize the risk early warning of the secondary node, the present application does not monitor the risk of the secondary node, avoids missing inspection, improves the inspection safety, and does not inspect the secondary node according to the inspection mode of the primary node, significantly improves the overall inspection efficiency, improves the monitoring flexibility, avoids unnecessary energy consumption, and improves the economy. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flowchart of an intelligent inspection method of a track inspection robot provided by an embodiment of the present application;

[0027] Figure 2 is a filtering flowchart of monitoring data provided by an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific implementation, structure, features and effects of the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The data information collected in the present application is obtained with the authorization of the data owner.

[0030] As a specific scenario, according to the physical layout of the inspection scene (such as the top of the pipe gallery, the top of the machine room, the two sides of the equipment passage, etc.), the rigid track system is accurately installed. In the construction of the rigid track system, the layout of straight sections, curves, slopes, and turnouts needs to be considered to ensure smooth and non-shaking operation of the track inspection robot (hereinafter referred to as the inspection robot). In this embodiment, RFID tags or UltraWideBand (UWB) positioning base stations can be installed every certain distance (such as 20 meters) along the track as the reference point of the absolute position of the inspection robot, used to correct the cumulative error of the odometer and achieve accurate positioning. The inspection robot is manually controlled or automatically operated to make the laser radar carried by it scan the environment in all directions, and the sensor data is processed in real time through the SLAM (Simultaneous Localization and Mapping) algorithm to generate a high-precision two-dimensional or three-dimensional map, and the details in the map are optimized through manual editing.

[0031] In the generated map, all to-be-detected task nodes such as transformers, switch cabinets, and contact net poles are marked. The priority of each to-be-detected task node is determined manually, and a path planning is performed through a graph search algorithm, and the inspection robot runs according to the planned path. During the operation of the inspection robot, it stops at each point according to the planned path, collects relevant data of the corresponding to-be-detected task node through the relevant sensors arranged, and sends the data to the background for monitoring. The relevant sensors arranged include visible light cameras, infrared thermal images, sound pickups, and various types of gas sensors.

[0032] In the pre-setting, the staff divides each to-be-detected task node into two categories, namely main nodes and secondary nodes, according to the importance level of each to-be-detected task node. The main node is an important monitoring object, and the inspection robot needs to perform intensive monitoring on each main node on the planned path. After the inspection robot arrives at the specified position of the main node, it stops or slowly performs in-depth inspection, and collects multi-aspect data of the main node through multiple sensors. For the main node, this embodiment still performs inspection in the conventional inspection manner, which will not be described in detail. The secondary node is each node other than the main node, and it is not necessary to perform in-depth monitoring on each secondary node, but to perform pre-inspection on each secondary node and judge whether in-depth monitoring is needed according to the pre-inspection result.

[0033] In the process of running according to the planned path, the inspection robot will pass each secondary node on the planned path in turn. Based on the current position of the inspection robot and the position of each secondary node, the positional relationship between the two is determined, so that in the process of running according to the planned path, the secondary node to be passed will be taken as the object of data analysis to execute the intelligent inspection method of the track inspection robot provided in the embodiment. For each secondary node on the planned path, the intelligent inspection method of the track inspection robot provided in the embodiment is used for inspection.

[0034] As shown in Figure 1 The intelligent inspection method of the track inspection robot provided in the embodiment includes the following steps:

[0035] Step S1: When the distance between the inspection robot and the secondary node is within a first preset range, obtaining various monitoring data of the secondary node;

[0036] Step S2: When the monitoring data meets a first preset abnormality warning condition, obtaining predicted monitoring data of the distance between the inspection robot and the secondary node within a second preset range through a preset distance decay model;

[0037] Step S3: When the predicted monitoring data meets a second preset abnormality warning condition, obtaining the stability of the predicted monitoring data;

[0038] Step S4: Inputting the stability of various predicted monitoring data into a plurality of risk models to obtain a plurality of risk values and determining a maximum risk value;

[0039] Step S5: Risk warning for the secondary node according to the maximum risk value.

[0040] Each step will be described in detail as follows.

[0041] Step S1: When the distance between the inspection robot and the secondary node is within a first preset range, obtaining various monitoring data of the secondary node.

[0042] As a specific example, in the embodiment, the sensors provided on the inspection robot include a vibration sensor, a sound wave sensor, a gas sensor, and a temperature and humidity sensor. The collected monitoring data includes vibration data, sound wave data (i.e. sound data), gas concentration data, temperature data, and humidity data.

[0043] If the secondary node has a security risk, it will affect its surrounding environment and spread to a certain range through the medium. For example, if the secondary node has a gas leak due to equipment abnormalities, the leaked gas will spread to a certain range through air circulation. Therefore, the detection data of the gas sensor will reflect the detection concentration value of the leaked gas, such as carbon monoxide gas emitted when the transformer insulation is damaged. For another example, if the secondary node emits abnormal sound due to equipment abnormalities, the abnormal sound will spread to a certain range. Therefore, the detection data of the sound wave sensor will reflect the abnormal sound wave.

[0044] During the operation of the inspection robot, the distance to the secondary node to be passed through becomes closer and closer. The embodiment predefines a first preset range, which represents a certain distance from the secondary node. The upper limit value of the first preset range is used to control the start of the sensors on the inspection robot. Therefore, when the distance between the inspection robot and the secondary node becomes closer and closer until the upper limit value of the first preset range is reached, the inspection robot starts various sensors to perform data detection. The lower limit value of the first preset range is used to control the shutdown of the sensors on the inspection robot. Therefore, when the distance between the inspection robot and the secondary node is within the first preset range, the various sensors of the inspection robot perform data detection. When the distance between the inspection robot and the secondary node reaches the lower limit value of the first preset range, the sensors on the inspection robot are controlled to be turned off, and data detection is no longer performed. The specific range of the first preset range and the specific values of the upper and lower limits are determined by the actual application scenario. As a specific example, the upper limit value of the first preset range is 10 meters, and the lower limit value is 5 meters.

[0045] Therefore, during the operation of the inspection robot on the track, when the distance between the inspection robot and the secondary node reaches the upper limit value of the first preset range, the various sensors on the inspection robot are started to obtain various monitoring data of the secondary node. When the distance between the inspection robot and the secondary node reaches the lower limit value of the first preset range, the various sensors on the inspection robot are turned off, and the various monitoring data of the secondary node is no longer obtained. Since the inspection robot needs to consume a certain amount of time when it operates within the first preset range, the running time of the inspection robot within the first preset range is defined as a data collection time period, and the various monitoring data collected by the inspection robot is essentially the various monitoring data collected by the various sensors within the data collection time period. Based on the sampling frequency of the various sensors, the various monitoring data is a sequence of various monitoring data, and the monitoring data sequence includes data at each time within the data collection time period.

[0046] It should be understood that the above data collection process is performed during the movement of the inspection robot, and therefore in the various monitoring data collected, there may be a large amount of noise in the data due to the movement of the inspection robot, resulting in inaccurate detection results. Therefore, in order to improve the reliability of the data, the embodiment performs noise reduction processing on the collected various monitoring data. Then, as shown in Figure 2 the intelligent inspection method of the track inspection robot provided by the embodiment further comprises:

[0047] Step S11: filtering the monitoring data through a low-pass filter to obtain an initial filtered signal.

[0048] In the embodiment, the temperature, humidity and gas concentration data in the monitoring data are scalar data, that is, they are obtained at a specific sampling frequency, and the three types of data can be analyzed separately. Moreover, the sampling frequency of the three types of data can be set to a lower frequency, such as 10 Hz, and a moving average filter is used for filtering to eliminate transient fluctuations.

[0049] For vibration and acoustic wave data, the sampling frequency is high, and the individual data points do not have specific meaning, so the complete data of the frequency and waveform are meaningful. The sampling frequency of the vibration and acoustic wave data is set to 10 kHz.

[0050] In order to facilitate data processing, the embodiment sets a time window, the length of which is set according to actual needs, such as 0.5 s, and the timestamps of the various monitoring data in the window are aligned. Therefore, with 0.5 s as the time window, the various monitoring data are subjected to data noise reduction.

[0051] For any one of the vibration and acoustic wave data, a low-pass filter is used to filter out the high-frequency noise components, achieving preliminary high-frequency noise filtering and obtaining an initial filtered signal.

[0052] Step S12: filtering the initial filtered signal using a pre-trained adaptive filter to obtain a regular component signal.

[0053] The adaptive filter is pre-trained, and the training process of the adaptive filter is a prior art, which is briefly introduced as follows: collect normal monitoring data, perform low-pass filtering to obtain standard data, use the standard data as target data to train the adaptive filter until the weight converges.

[0054] The adaptive filter is pre-trained, and the training process of the adaptive filter is a prior art, which is briefly introduced as follows: collect normal monitoring data, perform low-pass filtering to obtain standard data, use the standard data as target data to train the adaptive filter until the weight converges.

[0055] Step S13: subtracting the regular component signal from the initial filtered signal to obtain a target component signal.

[0056] Subtracting the initial filtering signal from the regular component signal obtains a target component signal. Thus, filtering of the vibration and acoustic wave data is realized.

[0057] It should be understood that, as other embodiments, the present embodiment can also filter various monitoring data by using other existing data filtering algorithms, or not filter various monitoring data. If various monitoring signals are filtered, the monitoring data in the following steps is filtered data, and if various monitoring signals are not filtered, the monitoring data in the following steps is unfiltered data.

[0058] Step S2: When the monitoring data meets the first preset abnormal early warning condition, the predicted monitoring data in which the distance between the inspection robot and the secondary node is within the second preset range is obtained by using a preset distance attenuation model.

[0059] For various monitoring data, the first preset abnormal early warning condition is set, such as when the temperature deviates from the normal temperature fluctuation range. Different monitoring data correspond to different first preset abnormal early warning conditions. It should be understood that, since the distance apart from the secondary node is collected, the first preset range of distance needs to be considered when setting the first preset abnormal early warning condition. Moreover, since the monitoring data at multiple time points in the data collection time period is obtained, for any kind of monitoring data, the first preset abnormal early warning condition can be: calculating the average value of the amplitude of the monitoring data at each time point in the data collection time period, if the average value is within the preset normal fluctuation range, it indicates that the monitoring data is normal and does not meet the first preset abnormal early warning condition; if it is not within the preset normal fluctuation range, it is determined that the monitoring data is abnormal and meets the first preset abnormal early warning condition. Correspondingly, the preset normal fluctuation range needs to consider the influence of the first preset range of distance. Other various monitoring data are the same as this.

[0060] It should be understood that different types of monitoring data will have different degrees of attenuation in space propagation. During the inspection process of the inspection robot, since the inspection robot is moving, the monitoring data collected by the inspection robot will change correspondingly due to the different distances from the secondary node, that is, as the distance gradually decreases, the signal strength gradually increases.

[0061] The role of the distance attenuation model is to predict the monitoring data at or close to the secondary node based on the monitoring data of the inspection robot within the first preset range using the distance attenuation model. The core idea is to establish a mathematical model between the monitoring data and the distance, and use the model to extrapolate to different distances. The distance attenuation model describes the law that the intensity of physical signals (such as sound, light, electromagnetic wave, magnetic field strength, etc.) decreases with the increase of the distance from the source. The most common distance attenuation model is the inverse square law model, which is suitable for point sources of sound, light, wireless signals, etc. in a uniform medium. The present embodiment presets several distance attenuation models, in addition to the inverse square law model, including the diffusion theory model and the exponential decay model. Various monitoring data select the appropriate distance attenuation model according to the actual situation. In an exemplary embodiment, the distance attenuation model corresponding to sound waves and temperature is the inverse square law model, the distance attenuation model corresponding to gas concentration and vibration is the diffusion theory model, and the distance attenuation model corresponding to humidity is the exponential decay model. It should be understood that the various distance attenuation models involved in the present embodiment are all existing models, and their principles and specific data processing processes are all prior art, which will not be described in detail.

[0062] It is determined whether the various monitoring data meet the corresponding first preset abnormal warning condition. When at least one monitoring data meets its corresponding first preset abnormal warning condition, it is determined that the monitoring data meets the first preset abnormal warning condition. Then, for any kind of monitoring data, the monitoring data of the inspection robot within the first preset range is obtained by the corresponding preset distance attenuation model, and the predicted monitoring data is obtained when the distance between the inspection robot and the secondary node is within the second preset range. The value corresponding to the second preset range is less than the value corresponding to the first preset range, and the second preset range represents that the inspection robot is very close to the secondary node. In an exemplary embodiment, the numerical range of the second preset range can be greater than 0 and less than or equal to 1 meter, that is, the second preset range represents that the distance between the inspection robot and the secondary node is less than or equal to 1 meter. Thus, the monitoring data within the inspection robot moving to a distance less than or equal to 1 meter from the secondary node is obtained, which is defined as predicted monitoring data. Through the distance attenuation model, the influence of the distance variable during the movement of the inspection robot can be eliminated, and the comparability of the data is enhanced.

[0063] Step S3: When the predicted monitoring data meets the second preset abnormal warning condition, the stability of the predicted monitoring data is obtained.

[0064] For any kind of prediction monitoring data, the more abnormal the kind of prediction monitoring data is, the higher the risk of the secondary node is, therefore, the embodiment also obtains the abnormality degree of the kind of prediction monitoring data, the abnormality degree represents the risk level of the kind of prediction monitoring data, the higher the abnormality degree is, the higher the risk level is. In an exemplary embodiment, for the kind of prediction monitoring data, the embodiment sets a preset standard value, the preset standard value represents the numerical value of the kind of prediction monitoring data in the most normal state, if the kind of prediction monitoring data is in a normal numerical value range in the normal state, the preset standard value can be the numerical value corresponding to the center position of the normal numerical value range. Since the actual numerical values of the kind of prediction monitoring data at several time points in the second preset range are obtained, the average value of the actual numerical values of the kind of prediction monitoring data at each time point in the second preset range is calculated, and the abnormality degree of the kind of prediction monitoring data is obtained by the difference between the average value of the actual numerical values of the kind of prediction monitoring data and the preset standard value, the larger the difference is, the greater the abnormality degree is. Wherein, the difference is the absolute value of the difference between the average value of the actual numerical values of the kind of prediction monitoring data and the preset standard value, and then the absolute value of the difference is normalized, such as using tanh function for normalization, and the result after normalization is taken as the abnormality degree of the kind of prediction monitoring data, so as to obtain the abnormality degrees of other kinds of prediction monitoring data.

[0065] For the kind of prediction monitoring data, the embodiment presets an abnormality degree threshold, the preset abnormality degree threshold is used to determine whether the abnormality degree of the kind of prediction monitoring data is high. The numerical value range of the preset abnormality degree threshold is 0-1, and the specific numerical value is set by actual judgment needs, for example, if a safer judgment mechanism is needed, the preset abnormality degree threshold can be set smaller. Then, the kind of prediction monitoring data satisfies the second preset abnormality warning condition, specifically: the abnormality degree of the kind of prediction monitoring data is greater than or equal to the preset abnormality degree threshold corresponding to the kind of prediction monitoring data.

[0066] Moreover, the more stable the predicted monitoring data within the second preset range is, the lower the possibility of error and distortion of the predicted monitoring data is, and the higher the risk of the secondary node is with respect to the predicted monitoring data. Therefore, when the predicted monitoring data meets the second preset abnormal early warning condition, the stability of the predicted monitoring data is obtained, which represents the data stability of the predicted monitoring data. In an exemplary embodiment, the standard deviation of the numerical value of the actual data of each time within the second preset range of the predicted monitoring data is obtained, and the standard deviation is used to represent the data fluctuation. The greater the standard deviation is, the higher the data fluctuation is, and the worse the data stability is. Therefore, the stability of the predicted monitoring data is obtained by negatively correlating the standard deviation of the predicted monitoring data, for example, the negative correlation can be normalized by exp(-x), where x represents the object to be normalized, and exp represents the exponential function with the natural constant as the base. The result of the negative correlation normalization is the stability of the predicted monitoring data. Thus, the stability of various predicted monitoring data is obtained.

[0067] Step S4: inputting the stability of various predicted monitoring data into a plurality of risk models to obtain a plurality of risk values, and determining a maximum risk value.

[0068] Since there can be multiple sources and types of risks, and different types of risks have different data representations, the embodiment constructs a risk library, which includes a plurality of risk models. The actual number of risk models is set according to actual needs. Each risk model includes the risk weight of various monitoring data, and the risk weight of various monitoring data in different risk models is different, thereby ensuring that the risk library covers more risk types. It should be understood that the various risk models included in the risk library can be determined by domain experts (such as senior operation and maintenance engineers) corresponding to the application scenario, specifically, the data weight of various monitoring data in each risk model is determined, for example, the data weight of various monitoring data in the transformer overheating risk is set as: the temperature weight is 0.7, the sound wave weight is 0.2, the vibration weight is 0.1, and the weight of other types of monitoring data is 0. It should be understood that for any risk model, the sum of the data weights of various monitoring data is 1.

[0069] For any risk model, the stability of various predicted monitoring data is input into the risk model, and the stability of various predicted monitoring data is weighted and summed according to the risk weight of various monitoring data in the risk model, and the result is the risk value corresponding to the risk model. Thus, the risk value of the secondary node with respect to each risk model is obtained. The greater the risk value is, the greater the risk of the secondary node with respect to the corresponding risk model is. Therefore, the maximum risk value, i.e., the risk value with the largest numerical value, is determined from the risk values corresponding to various risk models.

[0070] Step S5: risk warning of the secondary node according to the maximum risk value.

[0071] The maximum risk value represents the real risk of the secondary node, and thus the risk warning of the secondary node is performed according to the maximum risk value. In an exemplary embodiment, a preset risk threshold is provided, which is used to determine whether the maximum risk value is high. The preset risk threshold has a value range of 0-1, and the specific value is set according to actual judgment needs, for example, if a safer judgment mechanism is needed, the preset risk threshold can be set to be smaller. Then, the maximum risk value is compared with the preset risk threshold, if the maximum risk value is greater than or equal to the preset risk threshold, it is determined that the secondary node has a large risk, and a risk warning signal of the secondary node is output. In subsequent control, the risk warning signal of the secondary node is output at the same time, and a speed reduction instruction is also output to control the inspection robot to perform a speed reduction operation, to slow down or stop at the secondary node, so as to perform in-depth inspection on the secondary node. If the maximum risk value is less than the preset risk threshold, it indicates that the risk of the secondary node is low, and the inspection robot passes the secondary node at the original speed.

[0072] By using the above process, the secondary nodes in the inspection path are pre-inspected, when it is determined that there is a large risk, the risk warning signal of the secondary node is output, and at the same time, in-depth inspection is performed, so as to realize the inspection of each secondary node.

[0073] The embodiment also provides an intelligent inspection device of a track inspection robot, which comprises a memory and a processor, the memory is connected with the processor, and the memory is used to store program instructions; the processor is used to realize the steps in the above-mentioned intelligent inspection method of the track inspection robot when the program instructions are executed.

[0074] In an exemplary embodiment, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps in the above-mentioned intelligent inspection method of the track inspection robot.

[0075] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0076] Each embodiment in the specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other, and each embodiment mainly describes the differences from other embodiments.

Claims

1. An intelligent inspection method of a track inspection robot, characterized in that, The method comprises the following steps: When the distance between the inspection robot and the secondary node is within a first preset range, obtaining various monitoring data of the secondary node; The first preset range represents a certain distance from the secondary node. When the distance between the inspection robot and the secondary node is less than the upper limit value of the first preset range, the inspection robot starts various sensors for data detection. When the distance between the inspection robot and the secondary node reaches the lower limit value of the first preset range, the sensors on the inspection robot are controlled to be turned off, and data detection is no longer performed; When the monitoring data meets a first preset abnormal early warning condition, through a preset distance attenuation model, predicted monitoring data of the inspection robot and the secondary node within a second preset range is obtained; the second preset range is smaller than the first preset range; The input of the preset distance attenuation model is different types of monitoring data and the distance between the inspection robot and the secondary node, and the output is predicted monitoring data; the different types of monitoring data include: sound wave, temperature, gas concentration, vibration, humidity; the distance attenuation model corresponding to sound wave and temperature is a square inverse ratio model; the distance attenuation model corresponding to gas concentration and vibration is a diffusion theory model; the distance attenuation model corresponding to humidity is an exponential attenuation model; When the predicted monitoring data meets a second preset abnormal early warning condition, the stability of the predicted monitoring data is obtained; the stability is obtained by negatively correlating the standard deviation of the predicted monitoring data; The stability of various predicted monitoring data is input into a plurality of risk models to obtain a plurality of risk values, and the maximum risk value is determined; According to the maximum risk value, a risk warning is given to the secondary node.

2. The intelligent inspection method of the track-type inspection robot according to claim 1, characterized in that, After obtaining the various monitoring data of the secondary node, the method further comprises the following steps: The monitoring data is filtered through a low-pass filter to obtain an initial filtered signal; An initially trained adaptive filter is used to filter the initial filtered signal to obtain a regular component signal; The initial filtered signal is subtracted from the regular component signal to obtain a target component signal.

3. The intelligent inspection method of the track-type inspection robot according to claim 1, wherein, The monitoring data meets the first preset abnormal early warning condition, including that the monitoring data is outside a preset normal fluctuation range.

4. The intelligent inspection method of the track-type inspection robot according to claim 1, further characterized by The method comprises the following steps: Obtaining the abnormality degree of the predicted monitoring data; The predicted monitoring data meets the second preset abnormal early warning condition, including that the abnormality degree is greater than or equal to a preset abnormality degree threshold.

5. The intelligent inspection method of the track-type inspection robot according to claim 4, characterized in that, The abnormality degree is obtained from the difference between the predicted monitoring data and a preset standard value.

6. The intelligent inspection method of the track-type inspection robot according to claim 1, wherein, The risk model includes the risk weight of various monitoring data; The risk value acquisition process comprises: based on the risk weight of various monitoring data of the risk model, the stability of various predicted monitoring data is weighted and summed to obtain the risk value corresponding to the risk model.

7. The intelligent inspection method of the track-type inspection robot according to claim 1, wherein, According to the maximum risk value, a risk warning is given to the secondary node. A memory and a processor; 8. An intelligent inspection device of a track inspection robot, characterized in that it comprises: The memory is connected with the processor; The memory is used for storing program instructions; ​ The processor is configured to implement the intelligent inspection method of the track inspection robot according to any one of claims 1-7 when program instructions are executed.

Images