Data service robot for intelligent railway inspection and inspection method thereof

Abstract

The invention discloses a data service robot for intelligent railway inspection and an inspection method thereof, and relates to the technical field of railway inspection, and the data service robot comprises a data acquisition module, a data processing module, a data storage module and a data quality evaluation module; the data acquisition module is used for acquiring railway track information in real time; the data processing module is used for processing the railway track information in real time and extracting the defect information of the track surface and the damage information in the track to obtain railway inspection information; the data storage module is used for storing and managing the railway track information and the railway inspection information; and the data quality evaluation module is used for evaluating the railway track information and the railway inspection information according to set railway quality indexes, and judging the accuracy, the integrity and the consistency of the railway track information and the railway inspection information. According to the invention, the efficiency and accuracy of railway inspection can be improved, the cost and risk of manual inspection are reduced, and powerful data support is provided for safe operation of the railway.

Description

Technical Field

[0001] The present invention relates to the technical field of railway inspection, and more particularly to a data service robot for railway intelligent inspection and an inspection method thereof. Background Art

[0002] Against the backdrop of the booming railway transportation industry, the scale of railway systems continues to expand and operating speeds continue to increase, placing higher demands on the safety and reliability of railway operations. Railway inspection, as a key link in ensuring safe railway operations, is of undeniable importance. Traditional manual inspection methods have numerous drawbacks. They are not only labor-intensive, material-intensive, and time-consuming, but also inefficient and difficult to guarantee data accuracy and integrity.

[0003] With technological advancements, railway inspection is gradually moving towards intelligent systems. This process has resulted in the accumulation of vast amounts of inspection data, covering a wide range of information along the railway line, including track conditions, equipment operation, and the surrounding environment. However, due to the complexity and diversity of railway inspection work, the quality of this data varies greatly. Furthermore, existing data management and utilization methods are relatively outdated, hindering the full realization of this data's value. Advanced technologies are needed to improve the quality and efficiency of inspection data collection. Furthermore, effective processing and analysis of this data are essential, along with the establishment of a comprehensive data management system to provide strong support for railway maintenance and management.

[0004] Therefore, how to provide a data service robot and its inspection method that can collect, process and evaluate railway inspection data with high quality is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a data service robot and an inspection method for railway intelligent inspection, which improves the efficiency and accuracy of railway inspection, reduces the cost and risk of manual inspection, and provides strong data support for the safe operation of railways.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a data service robot for railway intelligent inspection, comprising: a data acquisition module, a data processing module, a data storage module and a data quality assessment module;

[0007] The data acquisition module is used to collect railway track information in real time;

[0008] The data processing module is used to process the railway track information in real time, extract the defect information on the track surface and the damage information inside the track, and obtain railway inspection information;

[0009] The data storage module is used to store and manage the railway track information and the railway inspection information;

[0010] The data quality assessment module is used to assess the railway track information and the railway inspection information according to the set railway quality indicators to determine their accuracy, completeness and consistency.

[0011] Preferably, the data acquisition module includes an image sensor and a rail flaw detection sensor;

[0012] The image sensor is used to collect surface images of railway tracks;

[0013] The track flaw detection sensor is used to detect internal signals of railway tracks.

[0014] Preferably, the data processing module includes an image processing unit and a signal processing unit;

[0015] The image processing unit uses a machine learning model to identify the surface image of the railway track and obtain defect information on the track surface;

[0016] The signal processing unit performs waveform analysis on the internal signals of the railway track to obtain the internal damage information of the track.

[0017] Preferably, a training dataset with labeled images of real track surface defects is obtained;

[0018] Inputting the training data set into a machine learning model to be trained to obtain a multi-category probability of track defect estimation, wherein the multi-category probability of track defect estimation includes a probability of a pixel point in the training data set being classified into each of a plurality of preset track defect categories;

[0019] Generating a track defect depth estimation value and a confidence level for each pixel point in the training data set according to the track defect estimation multi-category probability;

[0020] The model parameters of the machine learning model to be trained are adjusted according to the loss function until the output of the loss function meets the preset optimization conditions, thereby obtaining a trained machine learning model.

[0021] Preferably, the loss function includes a first loss function and a second loss function.

[0022] The first loss function is used to measure the error of each pixel of the training data set as a whole, and the error of each pixel of the training data set is the difference between the track defect depth estimation value of the pixel and the actual track surface defect depth estimation value of the pixel in the actual track surface defect label;

[0023] When the output of the second loss function is minimized, the confidence of each pixel in the training data set is negatively correlated with the error of the pixel;

[0024] The second loss function is a confidence ranking loss function, which compares the errors or confidences of any two pixels in the training data set;

[0025] The confidence ranking loss function is used to compare the errors of any two pixels in the training data set. When the difference between the errors of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the errors of any two pixels are consistent; the confidence of any two pixels in the training data set is compared by the confidence ranking loss function. When the difference between the confidences of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the confidences of any two pixels are consistent.

[0026] Preferably, the signal processing unit performs waveform analysis on the internal signal of the railway track to obtain internal track damage information, including:

[0027] determining an angle domain according to a measuring point category of the rail flaw detection sensor;

[0028] Determine the credibility of signals inside the railway track and ignore signals with low credibility;

[0029] The retained internal railway track signal is clipped to obtain the angle domain waveform signal and extract the signal characteristic parameters;

[0030] The signal characteristic parameters are calculated, and the internal damage information of the track is obtained according to a preset threshold value and a basis for judging the internal damage of the track.

[0031] Preferably, the retained railway track internal signal is clipped to obtain an angle domain waveform signal, and the signal characteristic parameters are extracted, including: analyzing a reference signal, calculating an angle domain time period of each measuring point, and clipping the angle domain waveform signal based on the time period;

[0032] The reference signal is used to synchronize and locate the position and time information of each measuring point during the railway track flaw detection process.

[0033] Preferably, an inspection method for a data service robot for railway intelligent inspection includes: collecting railway track information in real time;

[0034] Processing the railway track information in real time, extracting track surface defect information and track internal damage information, and obtaining railway inspection information;

[0035] Storing and managing the railway track information and the railway inspection information;

[0036] According to the set railway quality indicators, the railway track information and the railway inspection information are evaluated to determine their accuracy, completeness and consistency.

[0037] Through the above technical solution, it can be seen that compared with the existing technology, the present invention discloses a data service robot for railway intelligent inspection and its inspection method, including: a data acquisition module, a data processing module, a data storage module and a data quality assessment module; the data acquisition module is used to collect railway track information in real time; the data processing module is used to process the railway track information in real time, extract the defect information on the track surface and the internal damage information of the track, and obtain railway inspection information; the data storage module is used to store and manage the railway track information and railway inspection information; the data quality assessment module is used to evaluate the railway track information and railway inspection information according to the set railway quality indicators, and judge their accuracy, completeness and consistency. The present invention can improve the efficiency and accuracy of railway inspection, reduce the cost and risk of manual inspection, and provide strong data support for the safe operation of the railway. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0039] Figure 1 This is a structural schematic diagram of a data service robot for railway intelligent inspection provided by the present invention.

[0040] Figure 2 A schematic flow chart of an inspection method for a data service robot for railway intelligent inspection provided by the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] The embodiment of the present invention discloses a data service robot for railway intelligent inspection, such as Figure 1As shown, it includes: data acquisition module, data processing module, data storage module and data quality assessment module;

[0043] The data acquisition module is used to collect railway track information in real time; the data acquisition module includes an image sensor, a track flaw detection sensor, etc. The railway track information includes data such as the surface condition, internal damage and geometric parameters of the track;

[0044] The data processing module is used to process the railway track information in real time, extract the defect information of the track surface and the damage information inside the track, and obtain railway inspection information; the data processing module processes the collected data in real time, performs image recognition and analysis on the images taken by the high-definition camera, extracts the defect information of the track surface; and performs signal processing on the data of the track flaw detection sensor to determine whether there is damage inside the track;

[0045] The data storage module is used to store and manage the railway track information and the railway inspection information to ensure the security and traceability of the data;

[0046] The data quality assessment module is used to evaluate the railway track information and railway inspection information based on the set railway quality indicators to determine their accuracy, completeness, and consistency. The data quality assessment module evaluates the collected and processed track inspection data based on the set railway quality indicators; evaluates the accuracy of the data by comparing it with known track standard parameters; checks whether the data covers all key track parts and parameters to evaluate the data integrity; and ensures the consistency of data collected at different times for the same track part.

[0047] Specifically, the data acquisition module includes an image sensor and a track flaw detection sensor;

[0048] The image sensor is used to collect surface images of railway tracks;

[0049] The track flaw detection sensor is used to detect internal signals of railway tracks.

[0050] Specifically, the data processing module includes an image processing unit and a signal processing unit;

[0051] The image processing unit uses a machine learning model to identify surface rail images and obtain information about rail surface defects. After training on a large amount of labeled rail surface image data, the system can accurately identify various defects in the images, such as wear, spalling, and wavy wear.

[0052] The signal processing unit performs waveform analysis on internal rail signals to obtain internal rail damage information. This analysis allows the user to determine whether there are any internal rail damage conditions, such as fatigue crack growth or internal material defects.

[0053] Specifically, a training dataset of images with labels of real track surface defects is obtained; these images are accurately labeled to clarify the actual situation of the track surface defects corresponding to each pixel point, determine the defect category of the area where a certain pixel point is located, and the specific value of the defect depth.

[0054] Inputting the training data set into a machine learning model to be trained to obtain a multi-category probability of track defect estimation, wherein the multi-category probability of track defect estimation includes a probability of a pixel point in the training data set being classified into each of a plurality of preset track defect categories;

[0055] Generating a track defect depth estimation value and a confidence level for each pixel point in the training data set based on the track defect estimation multi-category probability; the depth estimation value is an estimated value of the track defect depth in the area represented by the pixel point, and the confidence level reflects the reliability of the estimated value;

[0056] The model parameters of the machine learning model to be trained are adjusted according to the loss function until the output of the loss function meets the preset optimization conditions, thereby obtaining a trained machine learning model.

[0057] Specifically, the machine learning model adopts an improved convolutional neural network; the improved convolutional neural network is to add a batch normalization layer after the ResNet50 input layer of the convolutional neural network, add two convolutional layers after the five convolutional layers, use a 1x1 convolution kernel in the last convolutional layer, and add a dropout layer before the FC layer to obtain an improved convolutional neural network;

[0058] The convolution kernels of the two convolutional layers are 3x3 and 1x1 respectively.

[0059] Specifically, the loss function includes a first loss function and a second loss function.

[0060] The first loss function is used to measure the error of each pixel of the training data set as a whole, and the error of each pixel of the training data set is the difference between the track defect depth estimation value of the pixel and the actual track surface defect depth estimation value of the pixel in the actual track surface defect label;

[0061] When the output of the second loss function is minimized, the confidence of each pixel in the training data set is negatively correlated with the error of the pixel;

[0062] The second loss function is a confidence ranking loss function, which compares the errors or confidences of any two pixels in the training data set;

[0063] The confidence ranking loss function is used to compare the errors of any two pixels in the training data set. When the difference between the errors of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the errors of any two pixels are consistent; the confidence of any two pixels in the training data set is compared by the confidence ranking loss function. When the difference between the confidences of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the confidences of any two pixels are consistent.

[0064] Specifically, the loss function also includes a third loss function,

[0065] The third loss function is a region of interest confidence loss function, and a portion of the training data set is selected as the region of interest of the training data set.

[0066] Among them, compared with the case where the loss function does not include the third loss function, when the output of the loss function including the third loss function meets the preset optimization condition, the average value of the confidence of the pixel points in the area of ​​interest is higher.

[0067] Specifically, the interest region confidence loss function is determined according to the total number of pixels located in the interest region and the confidence of the pixels located in the interest region.

[0068] Specifically, the signal processing unit performs waveform analysis on the internal signal of the railway track to obtain internal track damage information, including:

[0069] The angle domain is determined based on the measuring point category of the rail flaw detection sensor. Rail flaw detection sensors are used to detect damage to the internal structure of railway tracks. They have different measuring point categories based on sensor type, installation location, and detection method. Based on the physical quantity detected, they are categorized as ultrasonic flaw detection sensors, magnetic particle flaw detection sensors, and other types. Based on the installation location, they are categorized as measuring points at different locations, such as the rail head, rail web, and rail base.

[0070] During flaw detection, the determination of the angle domain is closely related to the measurement point categories. The angle domain refers to the angular range of the flaw detection signal or the region associated with its angular characteristics. In this embodiment of the present invention, ultrasonic flaw detection is used as an example. Ultrasonic waves emitted by an ultrasonic probe propagate and reflect within the rail. Different measurement point categories (e.g., measurement points at different locations on the rail head) cause the ultrasonic waves to enter and reflect at different angles. Based on the location and detection direction of these measurement points, the angular range of the ultrasonic signal can be determined. This angular range is the so-called angle domain. For example, at a measurement point on the side of the rail head, the ultrasonic wave enters the rail structure at an angle of approximately 30-60 degrees for flaw detection. This angular range of 30-60 degrees is defined as the angle domain corresponding to this measurement point category. By determining this angle domain, the flaw detection signals received within this angle range can be more specifically analyzed, leading to a better assessment of the damage within the rail at that measurement point.

[0071] Determine the credibility of signals inside the railway track and ignore signals with low credibility;

[0072] The credibility of the internal signal of the railway track is judged, including comparing the similarity of multiple consecutive sensor data in the recent time. If there is no similarity at all, it is considered unreliable.

[0073] Compare the signals of multiple related measuring points on the railway track to determine whether the changes are reasonable. If not, they are considered unreliable.

[0074] The retained internal rail signals are cropped to obtain an angle-domain waveform signal, and the signal's characteristic parameters are extracted. In rail flaw detection, the initially acquired internal rail signals often contain complex and comprehensive information, including information from different angles and locations, and potentially affected by various interference factors. Cropping the retained internal rail signals involves selecting the portion of the signal within a specific angle domain, based on a defined angle domain, from these original, relatively complex signals, while removing irrelevant signal portions outside this range. This is analogous to cropping a large image to a specific region, retaining only the portion of interest and relevant to the target analysis. This cropping operation results in an angle-domain waveform signal, which focuses on the internal rail conditions within a specific angle range, making it easier to conduct subsequent targeted analysis.

[0075] The signal characteristic parameters are calculated, and the internal damage information of the track is obtained according to a preset threshold value and a basis for judging the internal damage of the track.

[0076] Specifically, the retained railway track internal signal is clipped to obtain an angle domain waveform signal, and the signal characteristic parameters are extracted, including: analyzing the reference signal, calculating the angle domain time period of each measuring point, and clipping the angle domain waveform signal based on the time period;

[0077] The reference signal is used to synchronize and locate the position and time information of each measuring point during the railway track flaw detection process.

[0078] In railway track flaw detection, by analyzing the reference signal, the position reference of the track flaw detection sensor during operation on the track and the time relationship of each measuring point relative to this reference can be determined;

[0079] Because the reference signal is a pulse signal or a periodic signal with distinct peaks, it is necessary to detect peaks in the signal. Use the scipy.signal.find_peaks function in Python to detect peaks. In this embodiment of the present invention, for discrete reference signal data (sample indices), peaks,_ = scipy.signal.find_peaks(S_d) returns an array of peak indices, peaks. These peaks correspond to specific locations or time markers of the rail flaw detection equipment.

[0080] The relationship between time and position is established based on the rail flaw detection sensor's operating speed v (unit: meters per second) and the known characteristics of the reference signal. In this embodiment of the present invention, the rail flaw detection sensor moves a distance L (calculated based on equipment parameters and track structure) within one reference signal cycle. Therefore, the relationship between time t (seconds) and position x (meters) can be expressed as x = v × t. Combined with the time stamps obtained from peak detection, the positional reference of each measuring point on the track is determined.

[0081] Based on the installation position and detection angle of the track flaw detection sensor, combined with the previously established time-position relationship model, the angular domain time period of each measurement point is calculated. In this embodiment of the present invention, for an ultrasonic flaw detection sensor, an ultrasonic wave is emitted into the track at an angle θ. The propagation speed of the ultrasonic wave in the track material is known to be c (unit: m / s), and the vertical distance between the sensor and the track surface is h (unit: m). The propagation path length of the ultrasonic wave inside the track can be calculated based on the trigonometric function relationship:

[0082]

[0083] The propagation time t=l / c, and combined with the position and time reference determined by the reference signal, the angle domain time period of the measuring point is obtained.

[0084] Assume that the original railway track internal signal is R(t) and the sampling frequency is f s First, the angle domain time period [tstart ,t end ] is converted to the corresponding sample index range [n start ,n end ],in, ( Indicates rounding down. Then, the angle domain waveform signal is clipped.

[0085] The embodiment of the present invention also discloses a patrol method for a data service robot for railway intelligent patrol, such as Figure 2 As shown, it includes: real-time collection of railway track information;

[0086] Processing the railway track information in real time, extracting track surface defect information and track internal damage information, and obtaining railway inspection information;

[0087] Storing and managing the railway track information and the railway inspection information;

[0088] According to the set railway quality indicators, the railway track information and the railway inspection information are evaluated to determine their accuracy, completeness and consistency.

[0089] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0090] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data service robot for railway intelligent inspection, characterized in that: include: Data acquisition module, data processing module, data storage module and data quality assessment module; The data acquisition module is used to collect railway track information in real time; The data processing module is used to process the railway track information in real time, extract the defect information on the track surface and the damage information inside the track, and obtain railway inspection information; The data storage module is used to store and manage the railway track information and the railway inspection information; The data quality assessment module is used to assess the railway track information and the railway inspection information according to the set railway quality indicators to determine their accuracy, completeness and consistency.

2. A data service robot for railway intelligent inspection according to claim 1, characterized in that: The data acquisition module includes an image sensor and a track flaw detection sensor; The image sensor is used to collect the surface image of the railway track; The track flaw detection sensor is used to detect internal signals of railway tracks.

3. A data service robot for railway intelligent inspection according to claim 2, characterized in that: The data processing module includes an image processing unit and a signal processing unit; The image processing unit identifies the surface image of the railway track through a machine learning model to obtain defect information on the track surface; The signal processing unit performs waveform analysis on the internal signals of the railway track to obtain the internal damage information of the track.

4. The data service robot for railway intelligent inspection according to claim 3 is characterized in that: Obtain a training dataset with labeled images of real track surface defects; Inputting the training data set into a machine learning model to be trained to obtain a multi-category probability of track defect estimation, wherein the multi-category probability of track defect estimation includes a probability that a pixel point of the training data set is classified into each track defect category of a plurality of preset track defect categories; Generating a track defect depth estimation value and a confidence level for each pixel point of the training data set according to the track defect estimation multi-category probability; The model parameters of the machine learning model to be trained are adjusted according to the loss function until the output of the loss function meets the preset optimization conditions, thereby obtaining a trained machine learning model.

5. The data service robot for railway intelligent inspection according to claim 4 is characterized in that: The loss function includes a first loss function and a second loss function, The first loss function is used to measure the error of each pixel of the training data set as a whole, and the error of each pixel of the training data set is the difference between the estimated depth value of the track defect of the pixel and the estimated depth value of the real track surface defect of the pixel in the real track surface defect label; Wherein, when the output of the second loss function is minimized, the confidence of each pixel of each pixel of the training data set is negatively correlated with the error of the pixel; The second loss function is a confidence ranking loss function, which compares the errors or confidences of any two pixels in the training data set; The confidence ranking loss function is used to compare the errors of any two pixels in the training data set. When the difference between the errors of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the errors of any two pixels are consistent. The confidence of any two pixels in the training data set is compared by the confidence ranking loss function. When the difference between the confidences of any two pixels is less than a preset threshold, the confidence ranking loss function determines that the confidences of any two pixels are consistent.

6. The data service robot for railway intelligent inspection according to claim 3 is characterized in that: The signal processing unit performs waveform analysis on the internal signals of the railway track to obtain the internal damage information of the track, including: Determining the angle domain according to the measuring point category of the rail flaw detection sensor; Judge the credibility of signals inside the railway track and ignore signals with low credibility; The retained internal signal of the railway track is clipped to obtain the angle domain waveform signal and extract the signal characteristic parameters; The signal characteristic parameters are calculated, and the internal damage information of the track is obtained according to a preset threshold value and a basis for judging the internal damage of the track.

7. The data service robot for railway intelligent inspection according to claim 6, characterized in that: The retained internal signal of the railway track is clipped to obtain an angle domain waveform signal, and the signal characteristic parameters are extracted, including: by analyzing the reference signal, calculating the angle domain time period of each measuring point, and clipping the angle domain waveform signal according to the time period; The reference signal is used to synchronize and locate the position and time information of each measuring point during the railway track flaw detection process.

8. A patrol method for a data service robot for railway intelligent patrol, characterized in that: include: Real-time collection of railway track information; Processing the railway track information in real time, extracting track surface defect information and track internal damage information, and obtaining railway inspection information; Storing and managing the railway track information and the railway inspection information; According to the set railway quality indicators, the railway track information and the railway inspection information are evaluated to determine their accuracy, completeness and consistency.

Images