Rail robot inspection control method and system based on internet of things

Abstract

The application discloses a track robot inspection control method and system based on the Internet of Things, relates to the technical field of robot inspection, and comprises the following steps: collecting data based on a track robot, calculating accumulated cost value by comprehensively considering track width and track slope data, dynamically planning an inspection track, and collecting vibration signals in track operation to calculate vibration compensation. The method can ensure the stability of robot operation and the track keeping ability through dynamic balance load difference calculation, can construct a comprehensive fault feature map through weighted fusion of crack fault distribution probability curve and wear trend diagram, can effectively unify multiple types of fault features, can detect cracks and wear alone, can calculate the feature mean value and standard deviation of cracks, wear and composite faults respectively according to fault points extracted from the comprehensive fault feature map, can generate fault classification probability based on a classification Gaussian model, and can enhance the accurate classification ability of multiple fault areas.

Description

Technical Field

[0001] This invention relates to the field of robot inspection technology, and in particular to an Internet of Things-based track-mounted robot inspection control method and system. Background Technology

[0002] With the development of modern railway, subway and industrial transportation track systems, the importance of daily track maintenance and safety inspection has become increasingly prominent. In recent years, with the development of Internet of Things (IoT) technology, track robots have gradually become one of the core technical means of track inspection. Track robots combine multiple sensors (such as track width sensors, sound sensors, electromagnetic sensors) and IoT communication technology, which can not only collect track operation data in real time, but also achieve efficient automated inspection through autonomous path planning and dynamic trajectory adjustment.

[0003] However, existing track-based robot inspection technologies face several limitations. For example, they fail to fully consider the multidimensional characteristics of the track environment (such as slope, width, and vibration characteristics). In environments where track geometry and slope have a significant impact, robots struggle to effectively handle dynamic load differences caused by track imbalances, which may lead to problems with inspection path planning and positioning accuracy. Secondly, in the fault detection stage, existing technologies often fail to fully utilize the comprehensive perception capabilities of multi-sensor fusion for complex faults such as cracks and wear. In the fault diagnosis stage, current crack and wear characteristic analysis lacks a unified mathematical model, making it impossible to effectively correlate fault distribution with path information. It can only make preliminary judgments for single fault types, making it difficult to achieve collaborative planning and priority processing of multiple faults. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an IoT-based method and system for controlling the inspection of a track-mounted robot. This addresses the challenges in environments where track geometry and slope significantly influence performance. Robots struggle to effectively handle dynamic load differences caused by track imbalances, potentially leading to issues with inspection path planning and positioning accuracy. Furthermore, in the fault detection stage, existing technologies often fail to fully utilize the comprehensive perception capabilities of multi-sensor fusion for complex faults such as cracks and wear. In the fault diagnosis stage, current crack and wear characteristic analysis lacks a unified mathematical model, hindering the effective correlation between fault distribution and path information. Only preliminary judgments can be made for single fault types, making it difficult to achieve collaborative planning and prioritization of multiple faults.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an Internet of Things-based method for controlling the inspection of a track-mounted robot, comprising:

[0008] Data collection is performed using a track-mounted robot. The cumulative cost is calculated by combining track width and track slope data. The inspection trajectory is dynamically planned. Vibration signals during track operation are collected to calculate vibration compensation. The dynamic balance load difference caused by track slope is analyzed, and the speed of the left and right wheel sets is adjusted accordingly.

[0009] Sound signals are collected and vibration correction is performed by combining the vibration compensation amount caused by track operation. Noise reduction and signal power calculation are performed using fast Fourier transform. A probability description function of propagation path and crack location is constructed to generate crack fault distribution probability curve. Magnetic flux change signals are collected and corrected based on track width data, and converted into power spectrum using fast Fourier transform. Frequency features are screened and the spectral intensity is integrated over different track lengths to fit the wear trend and generate a comprehensive fault feature map by combining it with the crack fault distribution probability curve. A Gaussian model is constructed for fault type classification.

[0010] A bimodal propagation model of track segment fault points is constructed, the propagation probability curve of track faults is analyzed, the comprehensive probability distribution of track faults is calculated by combining the comprehensive fault feature map, the fault track area is analyzed, a priority list is generated, and the comprehensive weight is calculated to control the robot to perform fault inspection.

[0011] As a preferred embodiment of the IoT-based track-mounted robot inspection control method of the present invention, the following steps are included: dynamically planning the inspection trajectory, collecting vibration signals during track operation to calculate vibration compensation, and analyzing the dynamic balance load difference caused by track slope.

[0012] The A-star algorithm is used to perform dynamic trajectory planning by combining data on track width and track slope. The cumulative cost is calculated based on the slope and horizontal length of the track segment, and the trajectory cost is calculated based on the lowest estimated cost from the current node to the target node.

[0013] The output trajectory path consists of several nodes, including the coordinate data and slope data of each node;

[0014] Vibration signals during track operation are collected in real time by accelerometers and inertial sensors. The vibration signals are decomposed into frequency domain signals using fast Fourier transform, and the maximum amplitude value of the frequency corresponding to the maximum amplitude value is extracted as the vibration amplitude spectrum to calculate the vibration compensation amount.

[0015] Based on the slope data, robot type, and wheel set data, the dynamic balance load difference caused by the track slope is determined, and the adjustment speed of the left and right wheel sets is determined by combining the vibration amplitude spectrum.

[0016] As a preferred embodiment of the IoT-based track-mounted robot inspection and control method of the present invention, the following steps are included: constructing a probability description function for the propagation path and crack location, generating a crack fault distribution probability curve, collecting magnetic flux change signals and correcting them based on track width data, converting them into a power spectrum using a fast Fourier transform, integrating the frequency characteristics over the spectral intensity of different track lengths, fitting the wear trend, and combining it with the crack fault distribution probability curve to generate a comprehensive fault feature map.

[0017] The sound signals generated during track operation are recorded in real time by acoustic sensors, and vibration correction is performed by combining the vibration compensation amount caused by track operation.

[0018] Based on the pre-trained denoising autoencoder (DCAE) model, the corrected audio signal is denoised.

[0019] Historical acoustic data during track operation are collected, including acoustic propagation changes when passing through cracked areas and the denoised signal. The denoised signal is processed using Fast Fourier Transform to obtain the frequency domain signal, and the signal power is calculated.

[0020] A window function is used to segment and average the frequency domain signal, and the average is compared with the historical track crack power spectral density. The location corresponding to the signal power value that is greater than the crack power spectral density is judged as the high probability fault location.

[0021] A probability description function for propagation path and crack location is constructed, and the acoustic wave propagation data output by the multi-track sensor is fitted to generate a crack fault distribution probability curve.

[0022] Electromagnetic sensors are used to acquire magnetic flux change signals in real time, and corrections are made based on the track width.

[0023] The magnetic flux change signal is sampled uniformly over time and obtained by fast Fourier transform. The signal is then converted into a power spectrum. The fault frequency range is determined based on historical track wear fault data. Frequency features are selected and the wear trend is fitted by integrating the spectral intensity over different track lengths based on the frequency features.

[0024] Based on wear trends and crack fault distribution, a weighted superposition is performed to generate a comprehensive fault feature map;

[0025] Simultaneously, feature value distributions including crack failure, wear failure, and a combination of the two failures are extracted. The mean and standard deviation data of the three feature value distributions are determined respectively. A Gaussian model is constructed and the failure type is classified based on the comprehensive feature pairs.

[0026] The locations of predicted faults and their corresponding fault types are statistically analyzed for each track.

[0027] As a preferred embodiment of the IoT-based track-mounted robot inspection and control method of the present invention, wherein: the propagation probability curve of track faults is analyzed, and the comprehensive probability distribution of track faults is calculated by combining the comprehensive fault feature map, and the fault track region is analyzed, including...

[0028] By utilizing the attenuation characteristics of acoustic and electromagnetic signals, a bimodal propagation model of track fault points is constructed. Based on the peak values ​​of the crack fault distribution probability curve and wear trend diagram, the propagation probability curve of track faults is analyzed, and the comprehensive probability distribution of track faults is calculated by integrating the fault feature diagram and the propagation probability curve.

[0029] The maximum value is extracted as the center point of the predicted track fault, and the track segment to which it belongs is marked as the predicted fault area.

[0030] As a preferred embodiment of the IoT-based track-mounted robot inspection and control method of the present invention, wherein: the generation of the priority list includes,

[0031] Based on the distribution of fault classification probabilities at each track x, the center point of the predicted fault is labeled with the fault type, and the priority is labeled according to the magnitude of the fault classification probability value to form a priority list.

[0032] As a preferred embodiment of the IoT-based track-mounted robot inspection and control method of the present invention, wherein: the step of calculating comprehensive weights to control the robot for fault inspection includes,

[0033] Based on the center point of the predicted track fault, the response distance from the robot's current position to the center point of different predicted track faults is calculated, and a comprehensive weight is calculated.

[0034] The fault inspection trajectory is generated based on the comprehensive weight, and the robot is controlled to perform fault inspection operations.

[0035] As a preferred embodiment of the IoT-based track-mounted robot inspection and control method of the present invention, wherein: the data acquisition based on the track-mounted robot includes,

[0036] The track-based robot uses laser rangefinders to obtain the coordinates of the left and right track edges and determine the track width. It also uses height sensors to measure the height of the two ends of the track and determines the slope of the track segments based on the horizontal length of different track segments between track nodes.

[0037] Secondly, the present invention provides an Internet of Things-based track-mounted robot inspection and control system, comprising,

[0038] Track data acquisition module: Real-time acquisition of track width, gradient, vibration, sound, and magnetic flux change signals;

[0039] Wheelset dynamic adjustment module: Analyzes dynamic balance load difference and adjusts wheelset speed in combination with vibration compensation;

[0040] The sound signal processing module performs vibration correction and noise reduction on the sound signal, extracts frequency features based on fast Fourier transform and calculates signal power, and constructs a probability description of crack fault distribution.

[0041] Magnetic flux change signal processing module: Corrects the acquired magnetic flux change signal according to the track width, extracts frequency features through fast Fourier transform, and fits the track wear trend;

[0042] Fault Feature Analysis Module: Integrates crack fault distribution probability curves and wear trend maps to generate a comprehensive fault feature map of the track and uses a Gaussian model to classify fault types;

[0043] The bimodal propagation analysis module constructs a bimodal propagation model of the track segment based on acoustic and electromagnetic signals, analyzes the fault propagation probability curve, and further calculates the comprehensive probability distribution of track faults.

[0044] Fault Area Priority Module: Based on the comprehensive probability distribution of track faults, a priority list is generated and a comprehensive weight calculation is performed to generate fault inspection paths.

[0045] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the Internet of Things-based track robot inspection control method as described in the first aspect of the present invention.

[0046] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the Internet of Things-based track robot inspection control method as described in the first aspect of the present invention.

[0047] The beneficial effects of this invention are as follows: By calculating the dynamic balance load difference, the stability of robot operation and trajectory maintenance are ensured; by weighted fusion of the crack fault distribution probability curve and the wear trend map, a comprehensive fault feature map is constructed, effectively unifying the characteristics of multiple types of faults. It can not only detect cracks and wear separately, but also calculate the feature mean and standard deviation of cracks, wear and compound faults for the fault points extracted from the comprehensive fault feature map. Based on the classification Gaussian model, the fault classification probability is generated, which enhances the accurate classification ability of multiple fault regions. Through the dual-characteristic correlation analysis of the peak values ​​of the crack fault distribution probability curve and the wear trend map, the fault propagation probability curve is constructed. By utilizing the cooperative propagation characteristics between cracks and wear, the perception improvement from local probability to regional probability is achieved. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating the IoT-based track-mounted robot inspection control method in Example 1.

[0050] Figure 2 This is a schematic diagram of the IoT-based track-type robot inspection and control system in Example 1. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] Example 1, referring to Figures 1 to 2 This is the first embodiment of the present invention, which provides an Internet of Things-based track-mounted robot inspection control method, including the following steps:

[0055] S1, based on the track-type robot, collects data, calculates the cumulative cost value by integrating track width and track slope data, dynamically plans the inspection trajectory, collects vibration signals during track operation to calculate vibration compensation, analyzes the dynamic balance load difference caused by track slope, and adjusts the speed of the left and right wheel sets.

[0056] Preferably, data acquisition is based on a track-mounted robot, including:

[0057] The track-based robot uses laser rangefinders to obtain the coordinates of the left and right track edges and determine the track width. It also uses height sensors to measure the height of the two ends of the track and determines the slope of the track segments based on the horizontal length of different track segments between track nodes.

[0058] Furthermore, the inspection trajectory is dynamically planned, vibration signals during track operation are collected to calculate vibration compensation, and the dynamic balance load difference caused by track gradient is analyzed, including...

[0059] The A-Star algorithm is used to perform dynamic trajectory planning by integrating data on track width and track slope. The cumulative cost is calculated based on the slope and horizontal length of the track segment, and the trajectory cost is calculated based on the lowest estimated cost from the current node to the target node, expressed as:

[0060] f(n) = g(n) + h(n);

[0061]

[0062] Where f(n) represents the total generation value of node n, g(n) represents the cumulative generation value from the starting point to the current node n, h(n) represents the minimum estimated cost from the current node to the target node, calculated based on Euclidean distance, n represents the total number of orbital nodes, and S i L represents the gradient value of the i-th track segment. i Let x represent the length of the i-th track segment. m and y m Represents the target node coordinates, x n and y n Indicates the coordinates of the current node;

[0063] The output trajectory path consists of several nodes, including the coordinate data and slope data of each node;

[0064] Vibration signals (time-domain acceleration information) during track operation are collected in real time using accelerometers and inertial sensors. The vibration signals are decomposed into frequency-domain signals using a fast Fourier transform, and the maximum amplitude value corresponding to the frequency of the maximum amplitude is extracted as the vibration amplitude spectrum. The vibration compensation amount is then calculated and expressed as follows:

[0065] φ = 2πf1·Δt;

[0066] X c =-A z sin(2πf1t+φ);

[0067] Among them, A z Represents the vibration amplitude spectrum, X cThe value represents the vibration compensation amount, used to correct the robot's deviation caused by vibration. f1 represents the frequency corresponding to the maximum amplitude, t represents time, φ represents the oscillation phase difference, which is calculated by comparing the trajectory node position with the reference trajectory, and Δt represents the time delay of the vibration period.

[0068] Based on the slope data, robot type, and wheel set data, the dynamic balance load difference caused by the track slope is determined. Combined with the vibration amplitude spectrum, the adjustment speed of the left and right wheel sets is calculated, expressed as:

[0069] ΔT=S i ·g·W zh ·R lz ;

[0070]

[0071] Among them, R lz V represents the radius of the robot's wheel assembly, ΔT represents the dynamic balance load difference caused by the track slope, and V L and V R V′ represents the current linear velocity of the left and right wheel sets respectively, Δ′t represents the sampling interval, I represents the rotational inertia of the wheel set, determined based on the robot wheel set parameters, and V′ represents the rotational inertia of the wheel set. L and V′ R represents the speed of the left and right wheel sets after adjustment, and g represents the acceleration due to gravity.

[0072] In conventional path planning methods, the influence of slope is usually not considered, and only Euclidean distance is used as the planning basis. By calculating the cumulative cost value based on the track slope and length, the cost value is dynamically adjusted so that the robot prioritizes the selection of lower slope and shorter track segments. This significantly improves the energy consumption optimization capability of the trajectory planning process and avoids uneven load or climbing power loss caused by excessive selection of short Euclidean distance paths. Considering the dynamic load of slope can significantly improve the trajectory execution efficiency of the robot in complex track environments (such as mountainous or steep slope areas), while reducing kinetic energy loss and wheel slippage risk.

[0073] By acquiring acceleration signals from accelerometers and inertial sensors, and extracting the vibration amplitude spectrum using fast Fourier transform, the vibration dynamics characteristics of local track areas can be accurately identified, enabling intelligent compensation. Integrating the vibration compensation into the path planning system can effectively correct deviations caused by track vibration during robot operation. Through dynamic adjustment, high-precision path maintenance during trajectory planning and execution is achieved, making it particularly suitable for environments with poor track quality or abnormal vibration. By calculating the dynamic load difference, the slope data is combined with the kinematic characteristics of the wheel set (including wheel set rotational inertia, radius, etc.), and the speed of the left and right wheel sets is adjusted in real time to ensure the robot's smooth operation and trajectory maintenance capability.

[0074] S2. Acquire sound signals and combine them with vibration compensation caused by track operation for vibration correction. Then, perform noise reduction and use Fast Fourier Transform to calculate signal power. Construct a probability description function of propagation path and crack location, generate crack fault distribution probability curve, collect magnetic flux change signals based on track width data for correction, and use Fast Fourier Transform to convert them into power spectrum. Filter frequency features and integrate the spectral intensity on different track lengths to fit wear trends and combine with crack fault distribution probability curve to generate a comprehensive fault feature map. Construct a Gaussian model for fault type classification.

[0075] Preferably, a probability description function for the propagation path and crack location is constructed to generate a crack fault distribution probability curve. Magnetic flux change signals are collected and corrected based on track width data, then converted to a power spectrum using a fast Fourier transform. Frequency features are selected, and the spectral intensity is integrated over different track lengths to fit the wear trend. This is combined with the crack fault distribution probability curve to generate a comprehensive fault feature map, including...

[0076] The sound signals generated by the track during operation are recorded in real time by acoustic sensors, and vibration correction is performed by combining the vibration compensation amount caused by the track operation. This is expressed as follows:

[0077] s′(t)=s(t)-X c sin(2πf1t+φ);

[0078] Where s′(t) represents the corrected sound signal, and s(t) represents the acquired sound signal;

[0079] The corrected audio signal is denoised based on a pre-trained denoising autoencoder (DCAE) model.

[0080] Historical acoustic data during track operation is collected, including acoustic propagation changes when passing through the crack region and the denoised signal. The denoised signal is processed using Fast Fourier Transform to obtain the frequency domain signal, and the signal power is calculated, expressed as:

[0081]

[0082] Where P(f) represents the signal power value, |S(f)| 2 The square of the amplitude of the frequency domain signal is represented by , the energy intensity of the signal at the corresponding frequency is represented by , and N represents the total number of sampling points of the signal;

[0083] A window function is used to segment and average the frequency domain signal, and the average is compared with the historical track crack power spectral density. The location corresponding to the signal power value that is greater than the crack power spectral density is judged as the high probability fault location.

[0084] A probability description function for the propagation path and crack location is constructed, and the acoustic wave propagation data output by the multi-track sensor is fitted to generate a crack fault distribution probability curve, which is expressed as:

[0085] C(x) = s′(t)·exp(-εD);

[0086] Where D represents the distance between the sensor and the high-probability fault location, ε represents the acoustic energy attenuation coefficient, and the attenuation coefficient is adjusted experimentally based on the comparison of the actual crack location distribution and acoustic signal propagation changes. C(x) represents the crack fault distribution probability curve.

[0087] The electromagnetic flux change signal is acquired in real time using an electromagnetic sensor and corrected based on the track width, as shown below:

[0088] B′(t)=B(t)-S i ·g·W zh ;

[0089] Where B′(t) represents the corrected magnetic flux change signal, B(t) represents the acquired magnetic flux change signal, and W zh Indicates the total weight of the robot;

[0090] In track wear detection, electromagnetic sensors are usually placed on the bottom of the robot, close to the track. The sensors will detect changes in magnetic flux caused by track wear, cracks or other defects in real time. Theoretically, changes in magnetic flux of track material mainly reflect the actual situation of track wear.

[0091] During the sensing process, external interference factors in the robot's operating environment (such as changes in load distribution caused by slope) can cause the sensing signal B(t) to deviate from its true value. The effect of slope on load distribution includes the following: when the robot travels along a sloped track, gravity causes changes in the normal force applied to the track. This changes the contact characteristics of the track, which in turn leads to a shift in magnetic flux. When going uphill, the magnetic flux change signal may weaken because the normal force decreases, and when going downhill, the magnetic flux change signal may strengthen because the normal force increases. Therefore, the slope can indirectly interfere with the magnetic flux signal through changes in load and normal force.

[0092] The magnetic flux change signal is sampled uniformly over time, and the frequency domain signal is obtained through Fast Fourier Transform (FFT). This is then converted into a power spectrum. Based on historical track wear fault data, the fault frequency range is determined, frequency characteristics are selected, and the wear trend is fitted by integrating the spectral intensity over different track lengths based on these frequency characteristics. The result is expressed as:

[0093]

[0094] Where T(x) represents the wear trend map, T0 represents the initial reference value of the wear-free track, and F′B (ω) represents the frequency characteristic of the magnetic flux change signal, x represents the index of the orbital position, and L s Indicates the total number of orbital positions;

[0095] Based on wear trends and crack failure distribution, a weighted superposition is performed to generate a comprehensive failure feature map, represented as follows:

[0096] F(x) = C(x) + τ·T(x);

[0097] Where F(x) represents the comprehensive fault feature map, and τ represents the fusion weight, which is used to adjust the relative importance of acoustic signals and electromagnetic signals. It is empirically calibrated based on historical test data and can be determined by the ratio of the mean square error of the acoustic signal to the mean square error of the electromagnetic signal.

[0098] Simultaneously, feature value distributions including crack failure, wear failure, and a combination of both failures are extracted. The mean and standard deviation data of the three types of feature value distributions are determined respectively. A Gaussian model is constructed, and failure type classification is performed based on the comprehensive feature pairs, as shown below:

[0099]

[0100] Where G(x) represents the fault classification probability at track x, σ represents the standard deviation of the eigenvalue distribution, and μ represents the peak point of the comprehensive probability distribution (i.e. the position with the highest fault probability), which is used to describe the center position of the track fault.

[0101] Statistically determine the location of the predicted fault and the corresponding fault type at each track x.

[0102] By correcting the vibration compensation amount, the sound signal is modified and then denoised by the DCAE model, which can significantly reduce the signal noise interference from non-acoustic fault sources and enhance the sensitivity of crack location feature extraction. By comparing the acoustic signal power spectrum analysis with the crack power spectrum density, high-probability fault locations can be accurately extracted. Compared with only identifying the track points corresponding to signals above the crack power spectrum threshold as high-probability fault location points, this effectively avoids the false alarm problem of small amplitude signals and reduces false detections caused by the complexity of the track environment.

[0103] By fitting the acoustic wave energy propagation formula with actual crack data and dynamically adjusting the attenuation coefficient, a reliable crack distribution probability curve is constructed, realizing the extension of crack detection from single point detection to spatial distribution modeling. Especially in large-scale track detection, the long-distance fitting characteristics of acoustic wave propagation fully amplify the sensor coverage capability and reduce the risk of decreased detection accuracy.

[0104] By correcting magnetic flux change signals and extracting frequency domain features, accurate modeling of wear trends is achieved. By using a track width correction model, magnetic signal offsets caused by gravity are eliminated, ensuring that the magnetic signal can truly reflect the track wear characteristics. Combined with the spectral features extracted by fast Fourier transform, the random magnetic noise between equipment is further reduced by filtering based on the historical track wear frequency range. The corrected magnetic flux data, combined with the spectral intensity integral, can dynamically fit the wear trends of different track lengths, so that even in scenarios with uneven track thickness or slope changes, the wear distribution pattern can be fully reflected.

[0105] By weighted fusion of crack fault distribution probability curves and wear trend maps, a comprehensive fault feature map is constructed, effectively unifying the features of multiple fault types. It can not only detect cracks and wear separately, but also effectively label complex scenarios in real-world environments where multiple fault types are combined, improving the comprehensiveness and practicality of fault classification and perception results. For fault points extracted from the comprehensive fault feature map, the feature mean and standard deviation of cracks, wear, and combined faults are calculated separately. Based on a classification Gaussian model, fault classification probabilities are generated, enhancing the ability to accurately classify multiple fault regions. It can not only identify single fault types, but also quantify the probability and distribution characteristics of combined fault regions.

[0106] S3. Construct a bimodal propagation model of track segment fault points, analyze the propagation probability curve of track faults, calculate the comprehensive probability distribution of track faults by combining the comprehensive fault feature map, analyze the fault track area, generate a priority list, and calculate the comprehensive weight to control the robot to perform fault inspection.

[0107] Preferably, the propagation probability curve of the track fault is analyzed, and the comprehensive probability distribution of the track fault is calculated by combining the comprehensive fault characteristic map, and the faulty track region is analyzed, including...

[0108] Utilizing the attenuation characteristics of acoustic and electromagnetic signals, a bimodal propagation model for track fault points is constructed. Based on the peak values ​​of the crack fault distribution probability curve and wear trend diagram, the propagation probability curve of track faults is analyzed. Finally, the comprehensive probability distribution of track faults is calculated using the combined fault characteristic map and propagation probability curve, expressed as:

[0109]

[0110] F * (x)=F(x)·Po(x);

[0111] Where Po(x) represents the track fault propagation probability curve, A and C represent the initial intensity of the acoustic signal and the initial intensity of the electromagnetic signal, respectively, determined by the peak values ​​of C(x) and T(x), B and M represent the attenuation coefficients of the acoustic signal and the electromagnetic signal, respectively, which can be calibrated by the track material, and F *(x) represents the overall probability distribution of a fault at track x;

[0112] The maximum value is extracted as the center point of the predicted track fault, and the track segment to which it belongs is marked as the predicted fault area.

[0113] By constructing a bimodal propagation model based on the attenuation characteristics of acoustic and electromagnetic signals, the ability to perceive and predict fault range is enhanced. The attenuation characteristics of acoustic and electromagnetic signals are used to quantitatively describe the high-probability area around the fault point. The bimodal propagation model can take into account the characteristics of multiple signal sources when propagating over long distances, effectively making up for the shortcomings of single-signal mode in terms of complex fault characteristics and remote fault areas.

[0114] By using the dual-characteristic correlation analysis of the crack fault distribution probability curve and the peak value of the wear trend map, the fault propagation probability curve is constructed. By utilizing the synergistic propagation characteristics between cracks and wear (such as the attenuation changing with material and track load), the perception is improved from local probability to regional probability, dynamically reflecting the distribution range of the propagation probability curve. This allows the detection of complex fault distribution areas to go beyond single-point characteristics and accurately describe the overall shape of the fault distribution.

[0115] By extracting the maximum value of the comprehensive probability distribution of faults, the fault prediction center point is located and the track segment is marked, achieving high-precision regional fault identification. By fusing the comprehensive fault feature map and the propagation probability curve, the advantages of both signals are not only quantitatively combined, but also a dynamic description of the fault center point and the overall characteristics distribution of the track is realized. The dynamic nature of fault center point extraction combined with track segment marking provides a clear distribution range for subsequent track maintenance plans. At the same time, in the case of multi-point prediction, the fault area coverage is automatically optimized, reducing the risk of false detection.

[0116] Furthermore, a priority list is generated, including,

[0117] Based on the distribution of fault classification probabilities at each track x, the center point of the predicted fault is labeled with the fault type, and the priority is labeled according to the magnitude of the fault classification probability value to form a priority list.

[0118] Furthermore, the robot is controlled to perform fault inspection by calculating comprehensive weights, including:

[0119] Based on the center point of the predicted track fault, the response distance from the robot's current position to the center point of different predicted track faults is calculated, and a comprehensive weight is calculated, expressed as:

[0120]

[0121] Where W(L) a P represents the comprehensive weight of the predicted fault center point. yxj (L aD represents the fault priority at the predicted fault center point. xy (L a ) represents the response distance to the predicted fault center point, and ∈ represents the minimum value to avoid the denominator being zero;

[0122] The fault inspection trajectory is generated based on the comprehensive weight, and the robot is controlled to perform fault inspection operations.

[0123] By calculating the response distance from the robot's current position to the predicted fault center point on the track, and combining it with fault priority to generate a comprehensive weight, the agility of inspection path planning is improved. Through the dynamic calculation of the comprehensive weight, the coverage efficiency of inspection in densely faulty areas on the track is optimized.

[0124] This embodiment also provides an Internet of Things-based track-mounted robot inspection and control system, including,

[0125] Track data acquisition module: Real-time acquisition of track width, gradient, vibration, sound, and magnetic flux change signals;

[0126] Wheelset dynamic adjustment module: Analyzes dynamic balance load difference and adjusts wheelset speed in combination with vibration compensation;

[0127] The sound signal processing module performs vibration correction and noise reduction on the sound signal, extracts frequency features based on fast Fourier transform and calculates signal power, and constructs a probability description of crack fault distribution.

[0128] Magnetic flux change signal processing module: Corrects the acquired magnetic flux change signal according to the track width, extracts frequency features through fast Fourier transform, and fits the track wear trend;

[0129] Fault Feature Analysis Module: Integrates crack fault distribution probability curves and wear trend maps to generate a comprehensive fault feature map of the track and uses a Gaussian model to classify fault types;

[0130] The bimodal propagation analysis module constructs a bimodal propagation model of the track segment based on acoustic and electromagnetic signals, analyzes the fault propagation probability curve, and further calculates the comprehensive probability distribution of track faults.

[0131] Fault Area Priority Module: Based on the comprehensive probability distribution of track faults, a priority list is generated and a comprehensive weight calculation is performed to generate fault inspection paths.

[0132] This embodiment also provides a computer device applicable to the Internet of Things (IoT) based track-mounted robot inspection control method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the IoT-based track-mounted robot inspection control method proposed in the above embodiment.

[0133] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0134] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the IoT-based track-type robot inspection control method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0135] In summary, this invention ensures the stability of robot operation and trajectory maintenance by calculating the dynamic balance load difference. It constructs a comprehensive fault feature map by weighted fusion of the crack fault distribution probability curve and the wear trend map, effectively unifying the characteristics of multiple fault types. This not only allows for the separate detection of cracks and wear, but also calculates the feature mean and standard deviation of cracks, wear, and composite faults for the fault points extracted from the comprehensive fault feature map. Based on a classification Gaussian model, it generates fault classification probabilities, enhancing the accurate classification capability for multiple fault regions. Furthermore, through dual-characteristic correlation analysis of the peak values ​​of the crack fault distribution probability curve and the wear trend map, it constructs a fault propagation probability curve. Utilizing the collaborative propagation characteristics between cracks and wear, it achieves a perception improvement from local probability to regional probability.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for inspection and control of a track-mounted robot based on the Internet of Things, characterized in that, include: Data collection is performed using a track-based robot. The cumulative cost is calculated by combining track width and track slope data. The inspection trajectory is dynamically planned. Vibration signals during track operation are collected to calculate vibration compensation. The dynamic balance load difference caused by track slope is analyzed, and the speed of the left and right wheel sets is adjusted accordingly. Sound signals are collected and vibration correction is performed by combining the vibration compensation amount caused by track operation. Noise reduction and signal power calculation are performed using fast Fourier transform. A probability description function of propagation path and crack location is constructed to generate crack fault distribution probability curve. Magnetic flux change signals are collected and corrected based on track width data, and converted into power spectrum using fast Fourier transform. Frequency features are screened and the spectral intensity is integrated over different track lengths to fit the wear trend and generate a comprehensive fault feature map by combining it with the crack fault distribution probability curve. A Gaussian model is constructed for fault type classification. A bimodal propagation model of track segment fault points is constructed, the propagation probability curve of track faults is analyzed, the comprehensive probability distribution of track faults is calculated by combining the comprehensive fault feature map, the fault track area is analyzed, a priority list is generated, and the comprehensive weight is calculated to control the robot to perform fault inspection. The dynamic planning inspection trajectory collects vibration signals during track operation, calculates vibration compensation, and analyzes the dynamic balance load difference caused by track gradient, including... The A-star algorithm is used to perform dynamic trajectory planning by combining data on track width and track slope. The cumulative cost is calculated based on the slope and horizontal length of the track segment, and the trajectory cost is calculated based on the lowest estimated cost from the current node to the target node. The output trajectory path consists of several nodes, including the coordinate data and slope data of each node; Vibration signals during track operation are collected in real time by accelerometers and inertial sensors. The vibration signals are decomposed into frequency domain signals using fast Fourier transform, and the maximum amplitude value of the frequency corresponding to the maximum amplitude value is extracted as the vibration amplitude spectrum to calculate the vibration compensation amount. Based on the slope data, robot type, and wheel set data, the dynamic balance load difference caused by the track slope is determined, and the adjustment speed of the left and right wheel sets is determined by combining the vibration amplitude spectrum.

2. The IoT-based track-mounted robot inspection control method as described in claim 1, characterized in that: The process involves constructing a probability description function for the propagation path and crack location, generating a crack fault distribution probability curve, collecting magnetic flux change signals for correction based on track width data, converting them into a power spectrum using a Fast Fourier Transform, integrating the spectral intensity over different track lengths using frequency features, fitting the wear trend, and combining this with the crack fault distribution probability curve to generate a comprehensive fault feature map. The sound signals generated during track operation are recorded in real time by acoustic sensors, and vibration correction is performed by combining the vibration compensation amount caused by track operation. Based on the pre-trained denoising autoencoder (DCAE) model, the corrected audio signal is denoised. Historical acoustic data during track operation are collected, including acoustic propagation changes when passing through cracked areas and the denoised signal. The denoised signal is processed using Fast Fourier Transform to obtain the frequency domain signal, and the signal power is calculated. A window function is used to segment and average the frequency domain signal, and the average is compared with the historical track crack power spectral density. The location corresponding to the signal power value that is greater than the crack power spectral density is judged as the high probability fault location. A probability description function for propagation path and crack location is constructed, and the acoustic wave propagation data output by the multi-track sensor is fitted to generate a crack fault distribution probability curve. Electromagnetic sensors are used to acquire magnetic flux change signals in real time, and corrections are made based on the track width. The magnetic flux change signal is sampled uniformly over time and obtained by fast Fourier transform. The signal is then converted into a power spectrum. The fault frequency range is determined based on historical track wear fault data. Frequency features are selected and the wear trend is fitted by integrating the spectral intensity over different track lengths based on the frequency features. Based on wear trends and crack fault distribution, a weighted superposition is performed to generate a comprehensive fault feature map; Simultaneously, feature value distributions including crack failure, wear failure, and a combination of the two failures are extracted. The mean and standard deviation data of the three feature value distributions are determined respectively. A Gaussian model is constructed and the failure type is classified based on the comprehensive feature pairs. The locations of predicted faults and their corresponding fault types are statistically analyzed for each track.

3. The IoT-based track-mounted robot inspection control method as described in claim 2, characterized in that: The analysis involves analyzing the propagation probability curve of track faults, combining it with a comprehensive fault characteristic map to calculate the comprehensive probability distribution of track faults, and analyzing the faulty track region, including... By utilizing the attenuation characteristics of acoustic and electromagnetic signals, a bimodal propagation model of track fault points is constructed. Based on the peak values ​​of the crack fault distribution probability curve and wear trend diagram, the propagation probability curve of track faults is analyzed, and the comprehensive probability distribution of track faults is calculated by integrating the fault feature diagram and the propagation probability curve. The maximum value is extracted as the center point of the predicted track fault, and the track segment to which it belongs is marked as the predicted fault area.

4. The IoT-based track-mounted robot inspection control method as described in claim 3, characterized in that: The generation priority list includes, Based on the distribution of fault classification probabilities at each track x, the center point of the predicted fault is labeled with the fault type, and the priority is labeled according to the magnitude of the fault classification probability value to form a priority list.

5. The IoT-based track-mounted robot inspection control method as described in claim 4, characterized in that: The calculation of comprehensive weights and control of the robot for fault inspection includes, Based on the center point of the predicted track fault, the response distance from the robot's current position to the center point of different predicted track faults is calculated, and a comprehensive weight is calculated. The fault inspection trajectory is generated based on the comprehensive weight, and the robot is controlled to perform fault inspection operations.

6. The IoT-based track-mounted robot inspection control method as described in claim 1, characterized in that: The data collection based on the orbital robot includes, The track-based robot uses laser rangefinders to obtain the coordinates of the left and right track edges and determine the track width data. It also uses height sensors to measure the height data of the two ends of the track and determines the slope data of the track segments based on the horizontal length of different track segments between track nodes.

7. An Internet of Things (IoT)-based track-mounted robot inspection and control system, based on the IoT-based track-mounted robot inspection and control method according to any one of claims 1 to 6, characterized in that: include, Track data acquisition module: Real-time acquisition of track width, gradient, vibration, sound, and magnetic flux change signals; Wheelset dynamic adjustment module: Analyzes dynamic balance load difference and adjusts wheelset speed in combination with vibration compensation; The sound signal processing module performs vibration correction and noise reduction on the sound signal, extracts frequency features based on fast Fourier transform and calculates signal power, and constructs a probability description of crack fault distribution. Magnetic flux change signal processing module: Corrects the acquired magnetic flux change signal according to the track width, and extracts frequency features through fast Fourier transform to fit the track wear trend; Fault Feature Analysis Module: Integrates crack fault distribution probability curves and wear trend maps to generate a comprehensive fault feature map of the track and uses a Gaussian model to classify fault types; The bimodal propagation analysis module constructs a bimodal propagation model of the track segment based on acoustic and electromagnetic signals, analyzes the fault propagation probability curve, and further calculates the comprehensive probability distribution of track faults. Fault Area Priority Module: Based on the comprehensive probability distribution of track faults, a priority list is generated and a comprehensive weight calculation is performed to generate fault inspection paths.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the Internet of Things-based track-mounted robot inspection control method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the Internet of Things-based track-mounted robot inspection control method as described in any one of claims 1 to 6.

Images