Track line quality evaluation method and device

Abstract

The invention provides a track line quality evaluation method and device. A data source is acquired and preprocessed; acquiring corresponding to-be-analyzed data of the equipment unit from the preprocessed data source; analyzing the to-be-analyzed data according to a dynamic and static detection score algorithm and the neural network model to obtain a dynamic and static detection score and a model detection score of the equipment unit; determining a comprehensive evaluation score of the equipment unit by using a preset weight model, the dynamic and static detection score and the model detection score; determining a plurality of over-limit levels based on a preset threshold table to determine a final over-limit level; and generating early warning information based on the final overrun grade and the first overrun grade of all the equipment units. The scheme combines a dynamic and static analysis algorithm and a neural network algorithm. Static detection ensures the accuracy of geometric parameters of the track, and dynamic detection reflects the real performance of the track in operation. After the two data are fused, the total score of the equipment unit is calculated, and mapping of the data and line quality evaluation is realized.

Description

Technical Field

[0001] The present invention relates to the field of rail transit technology, and in particular to a rail line quality evaluation method and device. Background Art

[0002] With the rapid development of urban rail transit, the quality assessment of rail lines has become a key link in ensuring the safety and comfort of train operation. The health of rail lines directly affects the operating efficiency of trains and the comfort experience of passengers.

[0003] However, traditional manual inspection methods usually involve professionals visually inspecting the track, manually measuring it, and using simple tools to assess the geometric state and overall safety of the track. Although this method can detect obvious defects, it has some disadvantages, such as low inspection efficiency, limited coverage, and is easily affected by the experience and subjective judgment of the inspectors, resulting in missed inspections or misjudgments. In addition, manual inspections often require a lot of manpower and time, and it is difficult to meet the high requirements of modern urban rail transit for real-time and accuracy.

[0004] Therefore, how to improve the accuracy and efficiency of track line quality assessment is an urgent problem that needs to be solved. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a track line quality evaluation method and device to solve the problem of low accuracy and efficiency in current track line quality evaluation.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention discloses a method for evaluating track line quality, the method comprising:

[0008] Acquire and preprocess a data source, wherein the data source is inspection data of multiple devices of each equipment unit in a track line, and each of the equipment units corresponds to a mileage section in the track line;

[0009] For each of the equipment units, obtaining the corresponding data to be analyzed of the equipment unit from the preprocessed data source;

[0010] Analyze the data to be analyzed according to a dynamic and static detection score algorithm to obtain a dynamic and static detection score of the equipment unit;

[0011] Analyzing the data to be analyzed based on a neural network model to obtain a model detection score of the device unit; the neural network model is pre-trained based on a historical data source;

[0012] Determine the comprehensive evaluation score of the equipment unit by using a preset weight model, the dynamic and static detection scores, and the model detection score;

[0013] Determine a first over-limit level of the comprehensive evaluation score and a second over-limit level of the dynamic and static detection score based on a preset threshold table, and determine a third over-limit level of the model detection score according to a preset over-limit standard;

[0014] Determining a final over-limit level of the equipment unit using the preset weight model, the second over-limit level, and the third over-limit level;

[0015] The warning information is generated based on the final over-limit levels of all equipment units and the first over-limit levels.

[0016] Preferably, the method further comprises:

[0017] For each of the equipment units, obtaining a corresponding equipment medical record of the equipment unit;

[0018] The dynamic and static detection scores of the equipment unit, the model detection score of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection score, the third over-limit level of the model detection score and the final over-limit level of the equipment unit are stored in the equipment medical record book corresponding to the equipment unit.

[0019] Preferably, the process of pre-training the neural network model based on the historical data source includes:

[0020] Acquire a historical data source, wherein the historical data source is historical inspection data of multiple devices of each device unit in the track line;

[0021] Inputting the historical data source into a neural network model, and outputting a detection score corresponding to the historical data source;

[0022] Determine, based on the loss function, a difference between a predicted value of the detection score corresponding to the historical data source and a true value of the detection score corresponding to the historical data source;

[0023] Determine whether the difference is less than a preset threshold;

[0024] If the difference is less than a preset threshold, it is determined that the final neural network model is obtained;

[0025] If the difference is not less than a preset threshold, the parameters of the neural network model are updated, and the step of inputting the historical data source into the neural network model and outputting the detection score corresponding to the historical data source is returned to be executed.

[0026] Preferably, analyzing the data to be analyzed according to a dynamic and static detection score algorithm to obtain the dynamic and static detection score of the equipment unit includes:

[0027] Obtaining an analysis method of a dynamic and static detection score algorithm, wherein the analysis method includes mean analysis or peak analysis;

[0028] According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score;

[0029] The sum of the manual static unit score, the track inspection instrument static unit score, the track inspection instrument dynamic unit score and the track inspection vehicle dynamic unit score is calculated to obtain the dynamic and static detection scores of the equipment unit.

[0030] Preferably, the analysis method is mean analysis;

[0031] According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score, including:

[0032] According to the analysis method, the number of first measuring points of each geometric dimension and the first measuring point data with values ​​are obtained from the manual static inspection data of the data to be analyzed;

[0033] Calculate the sum of the absolute values ​​of the first measuring point data with values ​​and the ratio of the sum to the number of the first measuring points to obtain the first mean values ​​of the various geometric dimensions;

[0034] Obtaining a first weight of each of the geometric dimensions from a preset unit management weight table;

[0035] Calculate the artificial static unit score according to the first mean of each geometric dimension and the first weight of each geometric dimension;

[0036] Acquire the number of second measuring points of various geometric dimensions and the second measuring point data with values ​​from the track inspection instrument static inspection data of the data to be analyzed;

[0037] Calculate the ratio of the sum of the absolute values ​​of the second measuring point data with values ​​to the number of the second measuring points to obtain the second mean values ​​of the various geometric dimensions;

[0038] Acquire the second weight of each geometric dimension from the unitized management weight table;

[0039] Calculating a static unit score of the track inspection instrument according to the second mean values ​​of the various geometric dimensions and the second weight values ​​of the various geometric dimensions;

[0040] From the unitized management weight table, obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed;

[0041] Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed;

[0042] If there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the multiplier of the data to be analyzed;

[0043] Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition;

[0044] If there are multiple measuring points in the dynamic inspection data of the instrument to be analyzed, obtain the third measuring point data with values ​​and the number of the third measuring points with values ​​from the dynamic inspection data of the instrument to be analyzed; wherein the third measuring point data with values ​​include water addition value and vertical addition value;

[0045] Calculate the dynamic unit score of the multiplier according to the number of the third measuring points with values, the third measuring point data with values, the water weighted value and the vertical weighted value;

[0046] Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed;

[0047] Obtaining the corresponding weight of the detection value of each device from the unit management weight table;

[0048] Based on the detection values ​​of each device and the corresponding weights, the dynamic unit score of the track inspection vehicle is calculated.

[0049] Preferably, the analysis method is peak analysis;

[0050] According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score, including:

[0051] According to the analysis method, from the manual static inspection data of the data to be analyzed, the measurement point data with the largest median value among the measurement point data with values ​​in each geometric dimension is obtained, and marked as the peak value of the manual static inspection data;

[0052] Obtaining the weight of the manual static inspection data from a preset unit management weight table;

[0053] Calculating an artificial static unit score according to a peak value of the artificial static inspection data and a weight of the artificial static inspection data;

[0054] From the track inspection instrument static inspection data to be analyzed, obtain the measuring point data with the largest median value among the measuring point data with values ​​in each geometric dimension, and mark it as the peak value of the track inspection instrument static inspection data;

[0055] Obtaining the weight of the static inspection data of the track inspection instrument from a preset unit management weight table;

[0056] Calculating a static unit score of the track inspection instrument according to a peak value of the static inspection data of the track inspection instrument and a weight of the static inspection data of the track inspection instrument;

[0057] From the unitized management weight table, obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed;

[0058] Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed;

[0059] If there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the multiplier of the data to be analyzed;

[0060] Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition;

[0061] If there are multiple measuring points in the dynamic inspection data of the instrument for the data to be analyzed, the water addition value with the largest value and the vertical addition value with the largest value are obtained from the dynamic inspection data of the instrument for the data to be analyzed;

[0062] Calculate the dynamic unit score of the multiplier based on the water added value with the largest value, the vertical added value with the largest value, the water weighted value and the vertical weighted value;

[0063] Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed;

[0064] Obtaining the corresponding weight of the detection value of each device from the unit management weight table;

[0065] Based on the detection values ​​of each device and the corresponding weights, the dynamic unit score of the track inspection vehicle is calculated.

[0066] A second aspect of the present invention discloses a track line quality evaluation device, the device comprising:

[0067] A first acquisition unit is used to acquire a data source and perform preprocessing, wherein the data source is inspection data of multiple devices of each equipment unit in a track line, and each of the equipment units corresponds to a mileage section in the track line;

[0068] A second acquisition unit is used to acquire, for each of the equipment units, the corresponding to-be-analyzed data of the equipment unit from the preprocessed data source;

[0069] A first analysis unit, configured to analyze the data to be analyzed according to a dynamic and static detection score algorithm to obtain a dynamic and static detection score of the device unit;

[0070] A second analysis unit is used to analyze the data to be analyzed based on a neural network model to obtain a model detection score of the device unit; the neural network model is pre-trained based on a historical data source;

[0071] A first determination unit, configured to determine a comprehensive evaluation score of the equipment unit by using a preset weight model, the dynamic and static detection scores, and the model detection score;

[0072] a second determination unit, configured to determine a first over-limit level of the comprehensive evaluation score and a second over-limit level of the dynamic and static detection score based on a preset threshold table, and to determine a third over-limit level of the model detection score according to a preset over-limit standard;

[0073] a third determining unit, configured to determine a final over-limit level of the equipment unit by using the preset weight model, the second over-limit level and the third over-limit level;

[0074] A generating unit is used to generate warning information based on the final over-limit levels of all equipment units and the first over-limit levels.

[0075] Preferably, the device further comprises:

[0076] A third acquisition unit is used to acquire, for each of the equipment units, a device medical record book corresponding to the equipment unit;

[0077] A storage unit is used to store the dynamic and static detection scores of the equipment unit, the model detection scores of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection scores, the third over-limit level of the model detection scores and the final over-limit level of the equipment unit in the equipment medical record book corresponding to the equipment unit.

[0078] Preferably, the device further comprises:

[0079] A fourth acquisition unit, configured to acquire a historical data source, wherein the historical data source is historical inspection data of a plurality of devices of each device unit in the track line;

[0080] An output unit, used to input the historical data source into a neural network model and output a detection score corresponding to the historical data source;

[0081] a fourth determining unit, configured to determine, based on a loss function, a difference between a predicted value of the detection score corresponding to the historical data source and a true value of the detection score corresponding to the historical data source;

[0082] A judging unit, used to judge whether the difference is less than a preset threshold;

[0083] A fifth determining unit, configured to determine that a final neural network model is obtained if the difference is less than a preset threshold;

[0084] The sixth determination unit is used to update the parameters of the neural network model and return to execute the output unit if the difference is not less than a preset threshold.

[0085] Preferably, the first analysis unit comprises:

[0086] An acquisition module, used to acquire an analysis method of a dynamic and static detection score algorithm, wherein the analysis method includes a mean analysis or a peak analysis;

[0087] An analysis module is used to analyze the manual static inspection data, the track inspection instrument static inspection data, the train instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed according to the analysis method, and obtain the manual static unit score, the track inspection instrument static unit score, the train instrument dynamic unit score and the track inspection vehicle dynamic unit score;

[0088] The calculation module is used to calculate the sum of the manual static unit score, the track inspection instrument static unit score, the added instrument dynamic unit score and the track inspection vehicle dynamic unit score to obtain the dynamic and static detection score of the equipment unit.

[0089] Based on the above-mentioned embodiment of the present invention, a track line quality evaluation method and device are provided, which obtains a data source and performs preprocessing; obtains the corresponding data to be analyzed of the equipment unit from the preprocessed data source; analyzes the data to be analyzed according to the dynamic and static detection score algorithm and the neural network model to obtain the dynamic and static detection score and model detection score of the equipment unit; determines the comprehensive evaluation score of the equipment unit using the preset weight model, dynamic and static detection score and model detection score; determines multiple over-limit levels based on the preset threshold table to determine the final over-limit level; generates warning information based on the final over-limit level and the first over-limit level of all equipment units. This scheme combines the dynamic and static analysis algorithm with the neural network algorithm. Static detection ensures the accuracy of the track geometry parameters, while dynamic detection reflects the actual performance of the track in operation. After the data of the two are fused, the total score of the equipment unit is calculated to achieve the mapping of data and line quality evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0091] Figure 1 A flow chart of a track line quality evaluation method provided by an embodiment of the present invention;

[0092] Figure 2 A flow chart of analyzing data to be analyzed using a mean analysis method provided in an embodiment of the present invention;

[0093] Figure 3 A flow chart of analyzing data to be analyzed using a peak value analysis method provided in an embodiment of the present invention;

[0094] Figure 4 A flowchart of a neural network model training provided by an embodiment of the present invention;

[0095] Figure 5 A schematic diagram of a neural network model provided by an embodiment of the present invention;

[0096] Figure 6 A structural block diagram of a track line quality evaluation device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0097] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0098] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0099] As can be seen from the background technology, the traditional manual inspection method has low inspection efficiency, limited coverage, and is easily affected by the experience and subjective judgment of the inspectors, resulting in missed inspections or misjudgments. In addition, manual inspections often require a lot of manpower and time, which is difficult to meet the high requirements of modern urban rail transit for real-time and accuracy.

[0100] Therefore, an embodiment of the present invention provides a track line quality evaluation method and device, which obtains a data source and performs preprocessing; obtains the corresponding data to be analyzed of the equipment unit from the preprocessed data source; analyzes the data to be analyzed according to the dynamic and static detection score algorithm and the neural network model to obtain the dynamic and static detection score and model detection score of the equipment unit; determines the comprehensive evaluation score of the equipment unit using the preset weight model, dynamic and static detection score and model detection score; determines multiple over-limit levels based on a preset threshold table to determine the final over-limit level; generates warning information based on the final over-limit level and the first over-limit level of all equipment units. This solution combines the dynamic and static analysis algorithm with the neural network algorithm. Static detection ensures the accuracy of the track geometry parameters, while dynamic detection reflects the actual performance of the track in operation. After the data of the two are fused, the total score of the equipment unit is calculated to achieve the mapping of data and line quality evaluation.

[0101] See also Figure 1 , shows a flow chart of a track line quality evaluation method provided by an embodiment of the present invention. The method comprises:

[0102] Step S101: Acquire data source and perform preprocessing.

[0103] It should be noted that the data source is the inspection data of multiple devices in each equipment unit in a rail line, and each equipment unit corresponds to a mileage section in the rail line.

[0104] It can be understood that a rail line is divided into multiple specific mileage sections in advance (for example, the rails, curves, vertical curves, switches and other equipment contained in the rail line are divided into sections every 25 meters), and multiple devices are set in each mileage section. Therefore, the inspection data of multiple devices in each equipment unit are collected in advance in a preset manner to obtain the data source.

[0105] In actual applications, data sources include manual static inspection data, track inspection instrument static inspection data, train-adding instrument dynamic inspection data, and track inspection vehicle dynamic inspection data. Among them, manual static inspection data includes seven data items: gauge, level, triangular pit, direction, height, structure, and rectification. Each 25-meter equipment unit (i.e., rail unit) is divided into 8 sections, with a total of 9 points (actually recording data from 8 measuring points). Manual static inspection data supports both mean analysis and peak analysis.

[0106] The static inspection data of the track detector includes track gauge, level, triangular pit, track direction, height and other data. Each equipment unit has 10 measuring points (actually recording data from 8 measuring points). The static inspection data of the track detector supports both mean analysis and peak analysis.

[0107] The dynamic inspection data of the train instrument includes mileage, water addition, water addition level, vertical addition, vertical addition level, manual mark, speed, line name, engineering section, workshop, work area, mark, locomotive number, train number, date, time, vehicle model, equipment number, a total of 18 fields. Among them, the important fields are mileage, water addition, water addition level, vertical addition, and vertical addition level. The dynamic inspection data of the train instrument supports two methods: mean analysis and peak analysis.

[0108] The dynamic inspection data of the track inspection vehicle includes the following: the tolerance management value of the dynamic geometric dimensions of the track, the irregular quality index (TQI) management standard of the 200-meter section track (unit: mm), the TQI kilometer status assessment standard, the defect record report form (divided into level 1 to level 4), the TQI summary report form, the TQI kilometer summary report form, the kilometer summary report form, and the section summary report form. The dynamic inspection data of the track inspection vehicle supports both mean analysis and peak analysis.

[0109] In the specific implementation of step S101, the pre-collected data source is obtained and pre-processed, such as cleaning, denoising and normalizing the data source, to ensure that the pre-processed data source meets the input requirements of the subsequent neural network model.

[0110] Step S102: for each device unit, obtaining the corresponding to-be-analyzed data of the device unit from the preprocessed data source.

[0111] In the process of implementing step S102, for each equipment unit, the manual static inspection data, track inspection instrument static inspection data, train instrument dynamic inspection data and track inspection vehicle dynamic inspection data corresponding to the current equipment unit are obtained from the preprocessed data source and marked as data to be analyzed.

[0112] Step S103: Analyze the data to be analyzed according to the dynamic and static detection score algorithm to obtain the dynamic and static detection score of the equipment unit.

[0113] In the process of implementing step S103, the analysis method of the dynamic and static detection score algorithm is obtained. According to the analysis method, the manual static inspection data, the static inspection data of the track inspection instrument, the dynamic inspection data of the vehicle and the dynamic inspection data of the vehicle are analyzed respectively to obtain the manual static unit score, the static unit score of the track inspection instrument, the dynamic unit score of the vehicle and the dynamic unit score of the vehicle. Thus, the sum of the manual static unit score, the static unit score of the track inspection instrument, the dynamic unit score of the vehicle and the dynamic unit score of the vehicle is calculated to obtain the dynamic and static detection score of the equipment unit.

[0114] It should be noted that the analysis methods include mean analysis or peak analysis. Mean analysis refers to summing the absolute values ​​of the measured point data with values, dividing it by the number of measured points to obtain the mean, and then calculating the scores of each geometric dimension according to the preset weight table, and finally summing them up to obtain the equipment unit score. Peak analysis refers to selecting the maximum value of the measured points with values ​​in each geometric dimension, calculating the scores according to the weight table and summing them up.

[0115] Specifically, geometric dimensions refer to track moment, level, triangular pit, direction, and height.

[0116] It can be understood that the specific implementation method of analyzing the data to be analyzed according to the mean analysis method to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score is detailed in the embodiments of the present invention. Figure 2 According to the peak analysis method to be analyzed data analysis, get the artificial static unit score, track inspection instrument static unit score, add ride instrument dynamic unit score and track inspection vehicle dynamic unit score specific implementation of the present invention see the embodiment of the present invention Figure 3 The content in.

[0117] Step S104: Analyze the data to be analyzed based on the neural network model to obtain the model detection score of the equipment unit.

[0118] It should be noted that the neural network model is pre-trained based on historical data sources. For the specific training process, see Figure 4 The content shown in .

[0119] In the process of specifically implementing step S104, the pre-trained neural network model is used to analyze the data to be analyzed of the current device unit to obtain the model detection score of the current device unit.

[0120] Step S105: Determine the comprehensive evaluation score of the equipment unit using the preset weight model, the dynamic and static detection scores, and the model detection score.

[0121] In the process of specifically implementing step S105, the dynamic and static detection scores and model detection scores of the current equipment unit are input into the preset weight model, so as to calculate using the preset dynamic and static analysis weights and the preset neural network analysis weights in the preset weight model to obtain a comprehensive evaluation score of the current equipment unit.

[0122] The specific calculation principle is shown in formula (1).

[0123] S 总得分 =S1×K1+S2×K2(1)

[0124] Among them, S 总得分 is the comprehensive evaluation score of the current equipment unit; S1 is the dynamic and static analysis algorithm score; K1 is the preset dynamic and static analysis weight; S2 is the neural network analysis algorithm score; K2 is the preset neural network analysis weight.

[0125] Step S106: determining a first over-limit level of the comprehensive evaluation score and a second over-limit level of the dynamic and static detection score based on a preset threshold table, and determining a third over-limit level of the model detection score according to a preset over-limit standard.

[0126] In the specific implementation of step S106, a preset threshold table is obtained, and a comprehensive evaluation score (S 总得分 ) and determine the second over-limit level of the dynamic and static detection scores; obtain a preset over-limit standard, and determine the third over-limit level of the model detection score according to the preset over-limit standard.

[0127] It should be noted that since the types of equipment units include rail type, curve type, vertical curve type and turnout type, the preset threshold table for each type is different. On the other hand, the speed level corresponding to each track is different. Therefore, the speed level is also one of the factors in the preset threshold table.

[0128] It is understandable that the speed level of a track is fixed at the stage of completion. In other words, the speed level of each equipment unit on the track is consistent. For example: the maximum speed of high-speed rail tracks is usually designed to be between 230 km / h and 320 km / h; the speed of EMU tracks is usually supported by 150 km / h to 220 km / h; the speed of green train tracks is usually designed to be 80 km / h.

[0129] It should be noted that different speed levels have different quality requirements for equipment units. The higher the speed level, the higher the quality requirements for equipment units, and the lower the speed level, the lower the quality requirements for equipment units.

[0130] For example, (1) the preset threshold value table for type rail is as follows (Table 1):

[0131] Table 1 Over-limit speed level 0<V≤80 80<V≤120 120<V≤160 160<V≤200 excellent <![CDATA[0≤S 总得分 <18]]> <![CDATA[0≤S 总得分 <15]]> <![CDATA[0≤S 总得分 <15]]> <![CDATA[0≤S 总得分 <3]]> qualified <![CDATA[18≤S 总得分 <20]]> <![CDATA[15≤S 总得分 <20]]> <![CDATA[15≤S 总得分 <20]]> <![CDATA[3≤S 总得分 <4]]> Disqualified <![CDATA[20≤S 总得分 <99999]]> <![CDATA[20≤S 总得分 <99999]]> <![CDATA[20≤S 总得分 <99999]]> <![CDATA[4≤S 总得分 <99999]]>

[0132] (2) The preset threshold values ​​for curve, vertical curve and turnout are as follows (Table 2):

[0133] Table 2 Over-limit speed level 0<V≤80 80<V≤120 120<V≤160 160<V≤200 excellent <![CDATA[0≤S 总得分 <18]]> <![CDATA[0≤S 总得分 <15]]> <![CDATA[0≤S 总得分 <15]]> <![CDATA[0≤S 总得分 <15]]> qualified <![CDATA[18≤S 总得分 <20]]> <![CDATA[15≤S 总得分 <20]]> <![CDATA[15≤S 总得分 <20]]> <![CDATA[15≤S 总得分 <20]]> Disqualified <![CDATA[20≤S 总得分 <99999]]> <![CDATA[20≤S 总得分 <99999]]> <![CDATA[20≤S 总得分 <99999]]> <![CDATA[20≤S 总得分 <99999]]>

[0134] For another example, the preset over-limit standards are as follows: if the model detection score is between 0.8-1, the equipment unit quality evaluation is defined as good, and the third over-limit level is evaluated as 1; if the model detection score is between 0.6-0.8 (including 0.8), the equipment unit quality evaluation is defined as qualified, and the third over-limit level is evaluated as 2; if the model detection score is below 0.6 (including 0.6), the equipment unit quality evaluation is defined as unqualified, and the third over-limit level is evaluated as 3.

[0135] Step S107: Determine the final over-limit level of the equipment unit by using the preset weight model, the second over-limit level and the third over-limit level.

[0136] In the process of specifically implementing step S107, the second over-limit level of the dynamic and static detection scores and the third over-limit level of the model detection score are input into the preset weight model, so as to calculate using the preset dynamic and static analysis weights and the preset neural network analysis weights in the preset weight model to obtain the final over-limit level of the current equipment unit.

[0137] The specific calculation principle is shown in formula (2).

[0138] (2)

[0139] in, is the final over-limit level of the current equipment unit; The second level of over-limit score for dynamic and static tests; To preset dynamic and static analysis weights; The third level of excess limit for model detection score; Analyze weights for a preset neural network.

[0140] Step S108: generating warning information based on the first over-limit level and the final over-limit level of all equipment units.

[0141] In the specific implementation of step S108, warning information is generated according to the final over-limit level and the first over-limit level of all equipment units and the preset final over-limit level priority.

[0142] It should be noted that the preset final over-limit level priority, for example: the priority of disqualified is higher than the priority of qualified which is higher than the priority of excellent.

[0143] That is to say, when generating warning information, priority is given to generating warning information for equipment units whose final over-limit level is disqualified. When generating warning information, warning information can also be generated based on the data source, final over-limit level and comprehensive evaluation score of the current equipment unit.

[0144] Specifically, the early warning information includes but is not limited to the location of the equipment unit that needs to be repaired and the rectification requirements of the equipment unit, etc. Among them, the rectification requirements can be understood as repairing the equipment unit with a disqualified comprehensive evaluation level to a qualified or excellent comprehensive evaluation level.

[0145] It is understandable that accurate positioning of the damaged position in the equipment unit can be achieved through early warning information, making it easier to assign the direct defects of the equipment unit to the corresponding operation and maintenance personnel for repair, effectively reducing 80% of the quality analysis work of line engineers and 60% of the damage inspection work of team members.

[0146] It can be understood that after the early warning information is generated, the equipment medical record of the equipment unit is used to store the dynamic and static detection scores of the equipment unit, the model detection score of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection score, the third over-limit level of the model detection score and the final over-limit level of the equipment unit.

[0147] Specifically, for each equipment unit, obtain the equipment medical record book corresponding to the equipment unit; store the dynamic and static detection scores of the equipment unit, the model detection score of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection score, the third over-limit level of the model detection score and the final over-limit level of the equipment unit in the equipment medical record book corresponding to the equipment unit.

[0148] It should be noted that this process not only helps to establish detailed equipment operation and maintenance files, but also provides data support for subsequent fault analysis, trend monitoring and performance evaluation. Storing equipment medical records can ensure the traceability of information, facilitate historical comparison and early warning, and also improve the scientific nature of maintenance decisions, reduce the risk of equipment failure, extend the service life of equipment, and ultimately improve overall operational efficiency.

[0149] In an embodiment of the present invention, a dynamic and static analysis algorithm is combined with a neural network algorithm. Static detection ensures the accuracy of track geometry parameters, while dynamic detection reflects the actual performance of the track in operation. After the data of the two are fused, the total score of the equipment unit is calculated, and the data is mapped to the line quality evaluation through dynamic and static algorithm analysis and combined with the neural network algorithm. This method integrates the accuracy of static detection, the real-time nature of dynamic detection, and the intelligent analysis advantages of neural networks, and can comprehensively evaluate the line quality. After training and optimization, the model can accurately identify problem areas and guide maintenance strategies.

[0150] The above embodiments of the present invention Figure 1 For the specific implementation method of analyzing the data to be analyzed according to the mean analysis method to obtain the manual static unit score, the track inspection instrument static unit score, the train instrument dynamic unit score and the track inspection vehicle dynamic unit score, see Figure 2 , showing a flow chart of analyzing the data to be analyzed by the mean analysis method provided by an embodiment of the present invention, including:

[0151] It should be noted that the calculation order of the following manual static unit score, track inspection instrument static unit score, vehicle dynamic unit score and track inspection vehicle dynamic unit score is only for illustrative purposes and does not limit the order of calculating the scores in actual applications.

[0152] Step S201: according to the analysis method, the number of first measuring points of each geometric dimension and the first measuring point data with values ​​are obtained from the manual static inspection data of the data to be analyzed.

[0153] In the specific implementation of step S201, the first measuring point data and the number of first measuring points having values ​​for each geometric dimension are obtained from the manual static inspection data of the to-be-analyzed data of the current equipment unit according to the mean analysis method.

[0154] Step S202: Calculate the ratio of the sum of the absolute values ​​of the first measuring point data with values ​​to the number of the first measuring points to obtain the first mean values ​​of various geometric dimensions.

[0155] In the specific implementation of step S202, the absolute values ​​of the first measuring point data of a certain geometric dimension are calculated and added to obtain the sum value, and then the ratio of the sum value to the number of the first measuring points of the geometric dimension is calculated to obtain the first mean value of each geometric dimension.

[0156] Step S203: Obtain the first weight of each geometric dimension from a preset unit management weight table.

[0157] For example: from the preset unitized management weight table, according to the inspection type is "manual inspection of the line"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the overlimit type is the type of the corresponding geometric dimension measurement point data (gauge, level, triangular pit, direction, high and low); the speed level is the index path of the corresponding speed level, and the first weight of each geometric dimension is obtained.

[0158] Step S204: Calculate the artificial static unit score according to the first mean value of each geometric dimension and the first weight value of each geometric dimension.

[0159] In the specific implementation of step S204, for each geometric dimension, the product of the first mean value of the geometric dimension and the first weight value of the geometric dimension is calculated to obtain the score of the geometric dimension; the scores of each geometric dimension are summed to obtain the artificial static unit score.

[0160] Step S205: acquiring the number of second measuring points of various geometric dimensions and second measuring point data with values ​​from the track inspection instrument static inspection data of the data to be analyzed.

[0161] In the specific implementation of step S205, the second measuring point data and the number of the second measuring points having values ​​for each geometric dimension are obtained from the track inspection instrument static inspection data to be analyzed of the current equipment unit.

[0162] Step S206: Calculate the ratio of the sum of the absolute values ​​of the second measuring point data with values ​​to the number of the second measuring points to obtain the second mean values ​​of various geometric dimensions.

[0163] In the specific implementation of step S206, the absolute values ​​of the second measuring point data of a certain geometric dimension are calculated and added to obtain the sum value, and then the ratio of the sum value to the number of second measuring points of the geometric dimension is calculated to obtain the second mean value of each geometric dimension.

[0164] Step S207: Obtain the second weight of each geometric dimension from the unit management weight table.

[0165] For example: from the preset unit management weight table, according to the inspection type is "static inspection by track inspection instrument"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the overlimit type is the type of the corresponding geometric dimension measurement point data (gauge, level, triangular pit, direction, height); the speed level is the index path of the corresponding speed level, and the second weight of each geometric dimension is obtained.

[0166] Step S208: Calculate the track inspection instrument static unit score according to the second mean value of each geometric dimension and the second weight value of each geometric dimension.

[0167] In the specific implementation of step S208, for each geometric dimension, the product of the second mean value of the geometric dimension and the second weight value of the geometric dimension is calculated to obtain the score of the geometric dimension; the scores of each geometric dimension are summed to obtain the static unit score of the track inspection instrument.

[0168] Step S209: Obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed from the unitized management weight table.

[0169] For example: from the preset unitized management weight table, according to the inspection type of "adding instrument to check over-limit weight"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the over-limit type is the corresponding vertical acceleration, lateral acceleration (horizontal acceleration); the speed level is the index path of the corresponding speed level, obtain the water weighted value and vertical weighted value.

[0170] Step S210: Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed.

[0171] It should be noted that determining the number of measuring points in the dynamic inspection data of the multiplier to be analyzed is to determine how many inspection points of the multiplier are in the current equipment unit.

[0172] In the process of implementing step S210, it is determined whether there is only one measuring point in the dynamic inspection data of the added-on instrument of the data to be analyzed. If there is only one measuring point in the dynamic inspection data of the added-on instrument of the data to be analyzed, step S211 is executed; if there are multiple measuring points in the dynamic inspection data of the added-on instrument of the data to be analyzed, step S213 is executed.

[0173] Step S211: If there is only one measuring point in the dynamic inspection data of the instrument for analyzing the data, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the instrument for analyzing the data.

[0174] In the specific implementation of step S211, if there is only one measuring point in the dynamic inspection data of the instrument to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition corresponding to the measuring point are obtained from the dynamic inspection data of the instrument to be analyzed.

[0175] Step S212: Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition.

[0176] In the process of implementing step S212, the product of the water weighted value and the absolute value of the water addition is calculated to obtain the water addition score; the product of the vertical weighted value and the absolute value of the vertical addition is calculated to obtain the vertical addition score, and the sum of the water addition score and the vertical addition score is recorded as the dynamic unit score of the multiplier; then step S215 is executed.

[0177] Step S213: if there are multiple measuring points in the dynamic inspection data of the instrument to be analyzed, obtain the third measuring point data with values ​​and the number of the third measuring points with values ​​from the dynamic inspection data of the instrument to be analyzed.

[0178] It should be noted that the third measuring point data with values ​​include water addition value and vertical addition value.

[0179] In the process of implementing step S213, when there are multiple measuring points in the dynamic inspection data of the added-on instrument for the data to be analyzed, the third measuring point data with values ​​and the number of the third measuring points are obtained from the dynamic inspection data of the added-on instrument for the data to be analyzed.

[0180] Step S214: Calculate the dynamic unit score of the multiplier according to the number of third measuring points with values, the third measuring point data with values, the water weighted value and the vertical weighted value.

[0181] In the specific implementation of step S214, the sum of the absolute values ​​of all the water addition values ​​with values ​​is calculated, and the ratio between the sum and the number of the third measuring points with values ​​is calculated, and the product of the ratio and the water weighted value is recorded as the water addition score. The sum of the absolute values ​​of all the vertical addition values ​​with values ​​is calculated, and the ratio between the sum and the number of the third measuring points with values ​​is calculated, and the product of the ratio and the vertical addition weighted value is recorded as the vertical addition score. The sum of the water addition score and the vertical addition score is recorded as the dynamic unit score of the multiplier.

[0182] For example:

[0183] Water-added score = (sum of absolute values ​​of water-added with values / number of measuring points with values) × water-added weighted value;

[0184] Vertical addition score = (sum of the absolute values ​​of vertical additions with values / number of measuring points with values) × vertical addition weight.

[0185] Step S215: Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed.

[0186] It should be noted that since the data in the dynamic inspection data of the track inspection vehicle is in the form of a track inspection report, such as a TQI summary report. Specifically, the TQI summary report is based on a measurement point for every 200-meter interval, so the analysis method will not affect the score, that is, whether it is the mean analysis method or the peak analysis method, the dynamic unit score of the track inspection vehicle is the same.

[0187] In the specific implementation of step S215, the detection value of each device in the equipment unit is obtained from the track inspection vehicle dynamic inspection data of the equipment unit to be analyzed.

[0188] Step S216: Obtain the corresponding weight of the detection value of each device from the unit management weight table.

[0189] For example: from the preset unitized management weight table, the inspection type is "TQI", the overrun type is the relevant inspection item of the "TQI summary report", the speed level is the "inspection standard" of the "TQI summary report", and the unit type determines the corresponding weight of the inspection value of each device according to the corresponding kilometer mark of the current equipment unit in the "TQI summary report".

[0190] Step S217: Calculate the dynamic unit score of the track inspection vehicle based on the detection value of each device and the corresponding weight.

[0191] In the specific implementation of step S217, the product between the detection value of each device and the corresponding weight is calculated to obtain the score of each device, and the sum of the scores of all devices is recorded as the track inspection vehicle dynamic unit score.

[0192] In the embodiment of the present invention, the accuracy of static detection and the real-time nature of dynamic detection are effectively integrated to provide a more comprehensive track status assessment; secondly, through mean analysis, data fluctuations can be smoothed, trend recognition capabilities can be enhanced, and potential problems can be discovered; finally, analyzing the data to be analyzed according to the mean analysis method helps to optimize maintenance strategies and improve track safety and operational efficiency.

[0193] The above embodiments of the present invention Figure 1 For the specific implementation method of analyzing the data to be analyzed according to the peak analysis method to obtain the manual static unit score, the track inspection instrument static unit score, the train instrument dynamic unit score and the track inspection vehicle dynamic unit score, see Figure 3 , showing a flow chart of analyzing the data to be analyzed by the peak analysis method provided by an embodiment of the present invention, including:

[0194] It should be noted that the calculation order of the following manual static unit score, track inspection instrument static unit score, vehicle dynamic unit score and track inspection vehicle dynamic unit score is only for illustrative purposes and does not limit the order of calculating the scores in actual applications.

[0195] Step S301: According to the analysis method, the measured point data with the largest median value in the measured point data having values ​​in each geometric dimension is obtained from the manual static inspection data of the data to be analyzed, and marked as the peak value of the manual static inspection data.

[0196] In the process of implementing step S301, according to the peak analysis method, the measuring point data with values ​​in each geometric dimension are obtained from the manual static inspection data to be analyzed of the current equipment unit, and the measuring point data with the largest value are extracted from these measuring point data with values, and the measuring point data is marked as the peak value of the manual static inspection data.

[0197] Step S302: Obtain the weight of the manual static inspection data from a preset unit management weight table.

[0198] For example: from the preset unit management weight table, according to the inspection type is "manual inspection of the line"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the overrun type is the type of the corresponding geometric dimension measurement point data (gauge, level, triangular pit, direction, high and low); the speed level is the index path of the corresponding speed level, obtain the weight of the peak value, and obtain the weight of the manual static inspection data.

[0199] Step S303: Calculate the artificial static unit score according to the peak value of the artificial static inspection data and the weight of the artificial static inspection data.

[0200] In the specific implementation of step S303, the product of the peak value of the artificial static inspection data and the weight of the artificial static inspection data is recorded as the artificial static unit score.

[0201] Step S304: obtaining the maximum measuring point data of the measuring point data having values ​​in each geometric dimension from the track inspection instrument static inspection data to be analyzed, and marking it as the peak value of the track inspection instrument static inspection data.

[0202] In the process of implementing step S304, the measuring point data with values ​​in each geometric dimension are obtained from the track inspection instrument static inspection data to be analyzed of the current equipment unit, and the measuring point data with the largest value is extracted from these measuring point data with values, and the measuring point data is marked as the peak value of the track inspection instrument static inspection data.

[0203] Step S305: Obtain the weight of the static inspection data of the track inspection instrument from the preset unit management weight table.

[0204] For example: from the preset unit management weight table, according to the inspection type is "track inspection instrument static inspection"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the overrun type is the type of the corresponding geometric dimension measurement point data (gauge, level, triangular pit, direction, high and low); the speed level is the index path of the corresponding speed level, obtain the weight of the peak value, and obtain the weight of the track inspection instrument static inspection data.

[0205] Step S306: Calculate the track inspection instrument static unit score according to the peak value of the track inspection instrument static inspection data and the weight of the track inspection instrument static inspection data.

[0206] In the specific implementation of step S306, the product between the peak value of the static inspection data of the track inspection instrument and the weight of the static inspection data of the track inspection instrument is calculated, and the product is recorded as the static unit score of the track inspection instrument.

[0207] Step S307: Obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed from the unitized management weight table.

[0208] For example: from the preset unitized management weight table, according to the inspection type of "adding instrument to check over-limit weight"; the equipment type is the type of the current equipment unit (turnout, rail, curve or vertical curve); the over-limit type is the corresponding vertical acceleration, lateral acceleration (horizontal acceleration); the speed level is the index path of the corresponding speed level, obtain the water weighted value and vertical weighted value.

[0209] Step S308: Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed.

[0210] In the process of implementing step S308, it is determined whether there is only one measuring point in the dynamic inspection data of the added-on instrument of the data to be analyzed. If there is only one measuring point in the dynamic inspection data of the added-on instrument of the data to be analyzed, step S309 is executed; if there are multiple measuring points in the dynamic inspection data of the added-on instrument of the data to be analyzed, step S311 is executed.

[0211] Step S309: If there is only one measuring point in the dynamic inspection data of the instrument for analyzing the data, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the instrument for analyzing the data.

[0212] In the specific implementation of step S309, if there is only one measuring point in the dynamic inspection data of the instrument to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition corresponding to the measuring point are obtained from the dynamic inspection data of the instrument to be analyzed.

[0213] Step S310: Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition.

[0214] In the specific implementation process of step S310, the product of the water weighted value and the absolute value of the water addition is calculated to obtain the water addition score; the product of the vertical weighted value and the absolute value of the vertical addition is calculated to obtain the vertical addition score, and the sum of the water addition score and the vertical addition score is recorded as the dynamic unit score of the multiplier; then step S313 is executed.

[0215] Step S311: if there are multiple measuring points in the dynamic inspection data of the instrument to be analyzed, the largest water addition value and the largest vertical addition value are obtained from the dynamic inspection data of the instrument to be analyzed.

[0216] In the process of implementing step S311, when there are multiple measuring points in the dynamic inspection data of the added-in instrument for the data to be analyzed, the valuable measuring point data are obtained from the dynamic inspection data of the added-in instrument for the data to be analyzed, and the largest water added value and the largest vertical added value are extracted from these valuable measuring point data.

[0217] Step S312: Calculate the dynamic unit score of the multiplier based on the largest water added value, the largest vertical added value, the water weighted value and the vertical weighted value.

[0218] In the process of implementing step S312, the product of the largest water addition value (also called the maximum water addition value) and the water weighted value is calculated to obtain the water addition score; the product of the largest vertical addition value (also called the maximum vertical addition value) and the vertical weighted value is calculated to obtain the vertical addition score; the sum of the water addition score and the vertical addition score is recorded as the dynamic unit score of the multiplier.

[0219] For example:

[0220] Water plus score = maximum water plus value × water weighted value;

[0221] Vertical addition score = maximum vertical addition value × vertical addition weight.

[0222] Step S313: Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed.

[0223] Step S314: Obtain the corresponding weight of the detection value of each device from the unit management weight table.

[0224] Step S315: Calculate the dynamic unit score of the track inspection vehicle based on the detection value of each device and the corresponding weight.

[0225] It should be noted that the specific implementation of steps S313 to S315 is the same as Figure 2 The specific implementation methods of steps S215 to S217 are the same and will not be repeated here.

[0226] In the embodiment of the present invention, the peak analysis method helps to quickly identify abnormal points and timely discover potential problems or faults. It has high sensitivity and can effectively capture drastic changes in a short period of time. In addition, by analyzing peak changes, future track conditions can be predicted, maintenance decisions can be assisted, resource allocation can be optimized, and key areas can be ensured to receive timely attention. This method helps to improve overall operational safety, reduce accident risks, and maintain keen monitoring of track conditions in complex operating environments.

[0227] The above embodiments of the present invention Figure 1 For the specific implementation method of pre-training the neural network model based on historical data sources, see Figure 4 , shows a flowchart of a neural network model training provided by an embodiment of the present invention, including:

[0228] It should be noted that the embodiment of the present invention selects the classic recurrent neural network model (Recurrent Neural Network, RNN) and the long short-term memory network model (Long Short-Term Memory, LSTM), as well as the naive Bayes classifier to construct the neural network model.

[0229] It can be understood that in the input layer of the neural network model, multiple input channels are set according to the data type corresponding to the equipment in the equipment unit (such as gauge, level, triangular pit, direction, height, etc.) to receive the preprocessed data.

[0230] In the hidden layer of the neural network model, multiple hidden neural units are designed to extract the spatial features of the data; at the same time, the neural network model introduces a gating mechanism (such as LSTM or GRU (Gated Recurrent Unit)) to solve the gradient vanishing and explosion problems encountered by the traditional RNN model when processing long sequence data, which helps the network maintain the stability of information flow in long sequences.

[0231] In the output layer of the neural network model, multiple dimensions of output are set to predict the results corresponding to different quality evaluation indicators (such as equipment unit scores or over-limit levels, etc.).

[0232] It should be noted that the output of the neural network model is converted into a probability value through the naive Bayes classifier, that is, Figure 1 The "Neural Network Analysis Algorithm Score".

[0233] It is understandable that the neural network model is trained using historical data sources. The specific training process is as follows:

[0234] Step S401: Acquire historical data source.

[0235] It should be noted that the historical data source is the historical inspection data of multiple devices in each equipment unit in the track line.

[0236] In the specific implementation of step S401, a historical data source is obtained, and preprocessing operations such as cleaning, denoising, normalization, and key feature extraction are performed on the historical data source to ensure that the preprocessed historical data source meets the input requirements of the neural network model.

[0237] Step S402: input the historical data source into the neural network model, and output the detection score corresponding to the historical data source.

[0238] In the specific implementation process of step S402, the preprocessed historical data source is input into the neural network model, and the detection score corresponding to the preprocessed historical data source is output.

[0239] It should be noted that sigmoid (or Tanh) is used as the activation function in the neural network model.

[0240] Step S403: Based on the loss function, determine the difference between the predicted value of the detection score corresponding to the historical data source and the true value of the detection score corresponding to the historical data source.

[0241] It should be noted that during the training process of the neural network model, the cross-validation method is used to verify the neural network model.

[0242] In the specific implementation of step S403, based on the loss function, the difference between the predicted value of the detection score corresponding to the historical data source and the true value of the detection score corresponding to the historical data source is calculated.

[0243] Step S404: Determine whether the difference is less than a preset threshold value. If the difference is less than the preset threshold value, execute step S405; if the difference is not less than the preset threshold value, execute step S406.

[0244] It can be understood that if the difference is less than the preset threshold, it indicates that the deviation between the predicted value of the neural network model and the true value is small; if the difference is not less than the preset threshold, it indicates that the deviation between the predicted value of the neural network model and the true value is large, and the neural network model still needs to be trained.

[0245] Step S405: If the difference is less than the preset threshold, it is determined that the final neural network model is obtained.

[0246] In the specific implementation of step S405, when the difference between the predicted value of the detection score corresponding to the historical data source and the true value of the detection score corresponding to the historical data source is less than a preset threshold, it is determined that the final neural network model is obtained.

[0247] Step S406: If the difference is not less than the preset threshold, update the parameters of the neural network model and return to execute step S402.

[0248] In the process of implementing step S406, when the difference between the predicted value of the detection score corresponding to the historical data source and the true value of the detection score corresponding to the historical data source is not less than a preset threshold, the learning rate, batch size, number of iterations and other hyperparameters of the neural network model are adjusted through grid search, random search or Bayesian optimization and other methods, and the process returns to step S402.

[0249] Combination Figure 5 The schematic diagram of the neural network model shown in the figure is divided into three stages. The first stage is the forget stage, which is mainly implemented by the forget gate, that is, filtering out some useless information and discarding it from the current cell state. This step determines the memory of the cell state at the past moment. The Sigmoid layer of the forget gate looks at the output h of the previous unit hidden state. t-1 and the input x at the current time step t And through matrix operation, output a value between 0 and 1 for each cell state information to obtain f t , used to adjust the cell information of the previous time step. The process is as shown in formula (3):

[0250] (3)

[0251] Among them, W represents the weight during the training process; b represents the bias during the training process.

[0252] The second stage is the selective memory stage, which is to decide which part of the information input at the current time step needs to be stored in the cell state. This stage consists of two parts. The first step is to determine the update content through the Sigmoid layer of the input gate and based on the current time step input x t and the hidden state h of the previous time step t-1 Calculate i t .

[0253] The second tanh layer creates a new vector that is added to the cell state.

[0254] The third step combines these two vectors to create the updated value for the current moment , and select and memorize the cell state of the previous time step to obtain the input cell state of the next time step. The process is as shown in formula (4) to formula (6):

[0255] (4)

[0256] (5)

[0257] (6)

[0258] Among them, W represents the weight during the training process; b represents the bias during the training process.

[0259] The third stage is the output stage. This stage needs to determine the output content of the model, run the Sigmoid layer of the output gate and input x according to the current time step t and the hidden state h of the previous time step t-1 Calculate o t , which determines which parts of the cell state to output. The cell state is then processed through the tanh layer to obtain a value between -1 and 1 and combined with the output of the Sigmoid gate to obtain the final output h t , so far the forward propagation ends. The process of the output stage is as shown in formula (7) to formula (8):

[0260] (7)

[0261] (8)

[0262] Among them, x t represents the input at the current moment; W represents the weight during training; b represents the bias during training; h t represents the hidden state at time t; C t represents the cell state at time t; sigmoid and tanh represent activation functions.

[0263] It should be noted that the back propagation process is divided into three stages. The first stage is to obtain the output value of each LSTM unit in the network forward propagation process, and map the output prediction value of the current sequence to the new value range through the sigmoid normalized exponential function. and the true value y i The residual E can be calculated. The process is as shown in formula (9) and formula (10):

[0264] (9)

[0265] (10)

[0266] The second stage is to calculate the error of each unit during the back propagation of the network , consistent with the recurrent neural network, the LSTM error term back propagation also includes two directions, mainly from the following two sources: one is generated during the error term propagation to the upper layer, and the other is generated during the back propagation along time, calculated as shown in formula (11):

[0267] (11)

[0268] in, is the error term at time t; E is the residual between the predicted value and the true value; h t represents the hidden state at time t.

[0269] The third stage is to calculate the gradient of each weight based on the calculated error term, obtain the weight matrix of the network, and adjust and update the various network parameters of the LSTM model.

[0270] In the embodiment of the present invention, a neural network algorithm is used to achieve efficient and accurate evaluation of line quality. At the same time, through adaptive learning, the model can automatically identify the features in the detection data to meet the evaluation requirements of different lines and equipment. In addition, the trained neural network model supports real-time input and evaluation of detection data, providing timely and effective support for line maintenance, and comprehensively considering a variety of detection data and evaluation indicators, thereby achieving a comprehensive evaluation of line quality.

[0271] Corresponding to the track line quality evaluation method provided by the above embodiment of the present invention, see Figure 6 , shows a structural block diagram of a track line quality evaluation device provided in an embodiment of the present invention.

[0272] The device includes: a first acquisition unit 601 , a second acquisition unit 602 , a first analysis unit 603 , a second analysis unit 604 , a first determination unit 605 , a second determination unit 606 , a third determination unit 607 and a generation unit 608 .

[0273] The first acquisition unit 601 is used to acquire a data source and perform preprocessing, where the data source is inspection data of multiple devices of each equipment unit in a track line, and each equipment unit corresponds to a mileage section in the track line.

[0274] The second acquisition unit 602 is used to acquire, for each device unit, the to-be-analyzed data corresponding to the device unit from the preprocessed data source.

[0275] The first analysis unit 603 is used to analyze the data to be analyzed according to the dynamic and static detection score algorithm to obtain the dynamic and static detection score of the equipment unit.

[0276] The second analysis unit 604 is used to analyze the data to be analyzed based on the neural network model to obtain the model detection score of the equipment unit; the neural network model is pre-trained based on the historical data source.

[0277] The first determination unit 605 is used to determine the comprehensive evaluation score of the equipment unit by using a preset weight model, dynamic and static detection scores, and a model detection score.

[0278] The second determination unit 606 is used to determine the first over-limit level of the comprehensive evaluation score and the second over-limit level of the dynamic and static detection score based on a preset threshold table, and determine the third over-limit level of the model detection score according to a preset over-limit standard.

[0279] The third determining unit 607 is used to determine the final over-limit level of the equipment unit by using the preset weight model, the second over-limit level and the third over-limit level.

[0280] The generating unit 608 is configured to generate warning information based on the final over-limit level and the first over-limit level of all equipment units.

[0281] In the embodiment of the present invention, the dynamic and static analysis algorithms are combined with the neural network algorithm. Static detection ensures the accuracy of the track geometry parameters, while dynamic detection reflects the actual performance of the track in operation. After the data of the two are fused, the total score of the equipment unit is calculated, and the data is mapped to the line quality evaluation through dynamic and static algorithm analysis and combined with the neural network algorithm. It integrates the accuracy of static detection, the real-time nature of dynamic detection and the intelligent analysis advantages of neural networks, and can comprehensively evaluate the line quality. After training and optimization, the model can accurately identify problem areas and guide maintenance strategies.

[0282] Combination Figure 6 The content shown, the device also includes: a third acquisition unit and a storage unit.

[0283] The third acquisition unit is used to acquire the device medical record book corresponding to each device unit.

[0284] A storage unit is used to store the dynamic and static detection scores of the equipment unit, the model detection scores of the equipment unit, the comprehensive evaluation scores of the equipment unit, the first over-limit level of the comprehensive evaluation scores, the second over-limit level of the dynamic and static detection scores, the third over-limit level of the model detection scores and the final over-limit level of the equipment unit in the equipment medical record book corresponding to the equipment unit.

[0285] Combination Figure 6 According to the content shown, the device also includes: a fourth acquisition unit, an output unit, a fourth determination unit, a judgment unit, a fifth determination unit and a sixth determination unit.

[0286] The fourth acquisition unit is used to acquire a historical data source, wherein the historical data source is historical inspection data of multiple devices of each equipment unit in the track line.

[0287] The output unit is used to input the historical data source into the neural network model and output the detection score corresponding to the historical data source.

[0288] The fourth determination unit is used to determine, based on the loss function, a difference between a predicted value of the detection score corresponding to the historical data source and a true value of the detection score corresponding to the historical data source.

[0289] The judging unit is used to judge whether the difference is less than a preset threshold.

[0290] The fifth determining unit is used to determine the final neural network model if the difference is less than a preset threshold.

[0291] The sixth determination unit is used to update the parameters of the neural network model if the difference is not less than a preset threshold value, and return to the execution output unit.

[0292] Combination Figure 6 As shown in the figure, the first analysis unit 603 includes: an acquisition module, an analysis module and a calculation module.

[0293] The acquisition module is used to obtain the analysis method of the dynamic and static detection score algorithm, and the analysis method includes mean analysis or peak analysis.

[0294] The analysis module is used to analyze the manual static inspection data, track inspection instrument static inspection data, added vehicle dynamic inspection data and track inspection vehicle dynamic inspection data in the data to be analyzed according to the analysis method, and obtain the manual static unit score, track inspection instrument static unit score, added vehicle dynamic unit score and track inspection vehicle dynamic unit score.

[0295] The calculation module is used to calculate the sum of the manual static unit score, the track inspection instrument static unit score, the added instrument dynamic unit score and the track inspection vehicle dynamic unit score to obtain the dynamic and static detection scores of the equipment unit.

[0296] Combination Figure 6 When the analysis method is mean analysis, the analysis module includes:

[0297] The first artificial static unit score calculation submodule is used to obtain the first measurement point number and the first measurement point data with values ​​of each geometric dimension from the artificial static inspection data of the data to be analyzed according to the analysis method; calculate the sum of the absolute values ​​of the first measurement point data with values ​​and the ratio of the first measurement point number to obtain the first mean value of each geometric dimension; obtain the first weight of each geometric dimension from the preset unit management weight table; calculate the artificial static unit score according to the first mean value of each geometric dimension and the first weight of each geometric dimension.

[0298] The first track inspection instrument static unit score calculation submodule is used to obtain the number of second measuring points of each geometric dimension and the second measuring point data with values ​​from the track inspection instrument static inspection data of the data to be analyzed; calculate the sum of the absolute values ​​of the second measuring point data with values ​​and the ratio of the second measuring point number to obtain the second mean of each geometric dimension; obtain the second weight of each geometric dimension from the unitized management weight table; and calculate the static unit score of the track inspection instrument according to the second mean of each geometric dimension and the second weight of each geometric dimension.

[0299] The first calculation submodule of the dynamic unit score of the multiplier is used to obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed from the unitized management weight table; determine whether there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed; if there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, obtain the absolute value of the water addition and the absolute value of the vertical addition from the dynamic inspection data of the multiplier of the data to be analyzed; calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition.

[0300] The second calculation submodule of the dynamic unit score of the multiplier is used to obtain the third measuring point data with value and the number of the third measuring points with value from the dynamic inspection data of the multiplier of the data to be analyzed if there are multiple measuring points in the dynamic inspection data of the multiplier; wherein the third measuring point data with value include water added value and vertical added value; and calculate the dynamic unit score of the multiplier according to the number of the third measuring points with value, the third measuring point data with value, the water weighted value and the vertical weighted value.

[0301] The first track inspection vehicle dynamic unit score calculation submodule is used to obtain the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed; obtain the corresponding weight of the detection value of each device from the unit management weight table; and calculate the dynamic unit score of the track inspection vehicle based on the detection value of each device and the corresponding weight.

[0302] Combination Figure 6 The content shown, when the analysis mode is peak analysis, the analysis module includes:

[0303] The second artificial static unit score calculation submodule is used to obtain the measuring point data with the largest median value of the measuring point data with values ​​in various geometric dimensions from the artificial static inspection data to be analyzed according to the analysis method, and mark it as the peak value of the artificial static inspection data; obtain the weight of the artificial static inspection data from the preset unit management weight table; calculate the artificial static unit score according to the peak value of the artificial static inspection data and the weight of the artificial static inspection data.

[0304] The second track inspection instrument static unit score calculation submodule is used to obtain the measuring point data with the largest median value of the measuring point data with values ​​in various geometric dimensions from the track inspection instrument static inspection data to be analyzed, and mark it as the peak value of the track inspection instrument static inspection data; obtain the weight of the track inspection instrument static inspection data from a preset unit management weight table; and calculate the track inspection instrument static unit score according to the peak value of the track inspection instrument static inspection data and the weight of the track inspection instrument static inspection data.

[0305] The third calculation submodule of the dynamic unit score of the multiplier is used to obtain the water-weighted value and the vertical-weighted value of the dynamic inspection data of the multiplier of the data to be analyzed from the unitized management weight table; determine whether there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed; if there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, obtain the absolute value of the water addition and the absolute value of the vertical addition from the dynamic inspection data of the multiplier of the data to be analyzed; calculate the dynamic unit score of the multiplier according to the water-weighted value, the vertical-weighted value, the absolute value of the water addition and the absolute value of the vertical addition.

[0306] The fourth calculation submodule of the dynamic unit score of the multiplier is used to obtain the water added value and the vertical added value with the largest value from the dynamic inspection data of the multiplier to be analyzed if there are multiple measuring points in the dynamic inspection data of the multiplier to be analyzed; and calculate the dynamic unit score of the multiplier based on the water added value with the largest value, the vertical added value with the largest value, the water weighted value and the vertical weighted value.

[0307] The second track inspection vehicle dynamic unit score calculation submodule is used to obtain the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed; obtain the corresponding weight of the detection value of each device from the unit management weight table; and calculate the dynamic unit score of the track inspection vehicle based on the detection value of each device and the corresponding weight.

[0308] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0309] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0310] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating track line quality, characterized in that: The method comprises: Acquire and preprocess a data source, wherein the data source is inspection data of multiple devices of each equipment unit in a track line, and each of the equipment units corresponds to a mileage section in the track line; For each of the equipment units, obtaining the corresponding data to be analyzed of the equipment unit from the preprocessed data source; Analyze the data to be analyzed according to a dynamic and static detection score algorithm to obtain a dynamic and static detection score of the equipment unit; Analyzing the data to be analyzed based on a neural network model to obtain a model detection score of the device unit; the neural network model is pre-trained based on a historical data source; Determine the comprehensive evaluation score of the equipment unit by using a preset weight model, the dynamic and static detection scores, and the model detection score; Determine a first over-limit level of the comprehensive evaluation score and a second over-limit level of the dynamic and static detection score based on a preset threshold table, and determine a third over-limit level of the model detection score according to a preset over-limit standard; Determining a final over-limit level of the equipment unit using the preset weight model, the second over-limit level, and the third over-limit level; The warning information is generated based on the final over-limit levels of all equipment units and the first over-limit levels.

2. The method according to claim 1, characterized in that The method further comprises: For each of the equipment units, obtaining the equipment medical record book corresponding to the equipment unit; The dynamic and static detection scores of the equipment unit, the model detection score of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection score, the third over-limit level of the model detection score and the final over-limit level of the equipment unit are stored in the equipment medical record book corresponding to the equipment unit.

3. The method according to claim 1, characterized in that The process of pre-training the neural network model based on historical data sources includes: Acquire a historical data source, wherein the historical data source is historical inspection data of multiple devices of each device unit in the track line; Inputting the historical data source into a neural network model, and outputting a detection score corresponding to the historical data source; Determine, based on the loss function, a difference between a predicted value of the detection score corresponding to the historical data source and a true value of the detection score corresponding to the historical data source; Determine whether the difference is less than a preset threshold; If the difference is less than a preset threshold, it is determined that the final neural network model is obtained; If the difference is not less than a preset threshold, the parameters of the neural network model are updated, and the step of inputting the historical data source into the neural network model and outputting the detection score corresponding to the historical data source is returned to be executed.

4. The method according to claim 1, characterized in that: The step of analyzing the data to be analyzed according to the dynamic and static detection score algorithm to obtain the dynamic and static detection score of the equipment unit includes: Obtaining an analysis method of a dynamic and static detection score algorithm, wherein the analysis method includes mean analysis or peak analysis; According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score; The sum of the manual static unit score, the track inspection instrument static unit score, the track inspection instrument dynamic unit score and the track inspection vehicle dynamic unit score is calculated to obtain the dynamic and static detection scores of the equipment unit.

5. The method according to claim 4, characterized in that The analysis method is mean analysis; According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score, including: According to the analysis method, the number of first measuring points of each geometric dimension and the first measuring point data with values ​​are obtained from the manual static inspection data of the data to be analyzed; Calculate the sum of the absolute values ​​of the first measuring point data with values ​​and the ratio of the sum to the number of the first measuring points to obtain the first mean values ​​of the various geometric dimensions; Obtaining a first weight of each of the geometric dimensions from a preset unit management weight table; Calculate the artificial static unit score according to the first mean of each geometric dimension and the first weight of each geometric dimension; Acquire the number of second measuring points of various geometric dimensions and the second measuring point data with values ​​from the track inspection instrument static inspection data of the data to be analyzed; Calculate the ratio of the sum of the absolute values ​​of the second measuring point data with values ​​to the number of the second measuring points to obtain the second mean values ​​of the various geometric dimensions; Acquire the second weight of each geometric dimension from the unitized management weight table; Calculating a static unit score of the track inspection instrument according to the second mean values ​​of the various geometric dimensions and the second weight values ​​of the various geometric dimensions; From the unitized management weight table, obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed; Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed; If there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the multiplier of the data to be analyzed; Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition; If there are multiple measuring points in the dynamic inspection data of the instrument to be analyzed, obtain the third measuring point data with values ​​and the number of the third measuring points with values ​​from the dynamic inspection data of the instrument to be analyzed; wherein the third measuring point data with values ​​include water addition value and vertical addition value; Calculate the dynamic unit score of the multiplier according to the number of the third measuring points with values, the third measuring point data with values, the water weighted value and the vertical weighted value; Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed; Obtaining the corresponding weight of the detection value of each device from the unit management weight table; Based on the detection values ​​of each device and the corresponding weights, the dynamic unit score of the track inspection vehicle is calculated.

6. The method according to claim 4, characterized in that The analysis method is peak analysis; According to the analysis method, the manual static inspection data, the track inspection instrument static inspection data, the ride instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed are analyzed respectively to obtain the manual static unit score, the track inspection instrument static unit score, the ride instrument dynamic unit score and the track inspection vehicle dynamic unit score, including: According to the analysis method, from the manual static inspection data of the data to be analyzed, the measurement point data with the largest median value among the measurement point data with values ​​in each geometric dimension is obtained, and marked as the peak value of the manual static inspection data; Obtaining the weight of the manual static inspection data from a preset unit management weight table; Calculating an artificial static unit score according to a peak value of the artificial static inspection data and a weight of the artificial static inspection data; From the track inspection instrument static inspection data to be analyzed, obtain the measuring point data with the largest median value among the measuring point data with values ​​in each geometric dimension, and mark it as the peak value of the track inspection instrument static inspection data; Obtaining the weight of the static inspection data of the track inspection instrument from a preset unit management weight table; Calculating a static unit score of the track inspection instrument according to a peak value of the static inspection data of the track inspection instrument and a weight of the static inspection data of the track inspection instrument; From the unitized management weight table, obtain the water weighted value and the vertical weighted value of the dynamic inspection data of the multiplier of the data to be analyzed; Determine whether there is only one measuring point in the dynamic inspection data of the instrument to be analyzed; If there is only one measuring point in the dynamic inspection data of the multiplier of the data to be analyzed, the absolute value of the water addition and the absolute value of the vertical addition are obtained from the dynamic inspection data of the multiplier of the data to be analyzed; Calculate the dynamic unit score of the multiplier according to the water weighted value, the vertical weighted value, the absolute value of the water addition and the absolute value of the vertical addition; If there are multiple measuring points in the dynamic inspection data of the instrument for the data to be analyzed, the water addition value with the largest value and the vertical addition value with the largest value are obtained from the dynamic inspection data of the instrument for the data to be analyzed; Calculate the dynamic unit score of the multiplier based on the water added value with the largest value, the vertical added value with the largest value, the water weighted value and the vertical weighted value; Acquire the detection value of each device from the track inspection vehicle dynamic inspection data to be analyzed; Obtaining the corresponding weight of the detection value of each device from the unit management weight table; Based on the detection values ​​of each device and the corresponding weights, the dynamic unit score of the track inspection vehicle is calculated.

7. A track line quality evaluation device, characterized in that: The device comprises: A first acquisition unit is used to acquire a data source and perform preprocessing, wherein the data source is inspection data of multiple devices of each equipment unit in a track line, and each of the equipment units corresponds to a mileage section in the track line; A second acquisition unit is used to acquire, for each of the equipment units, the corresponding to-be-analyzed data of the equipment unit from the preprocessed data source; A first analysis unit, configured to analyze the data to be analyzed according to a dynamic and static detection score algorithm to obtain a dynamic and static detection score of the device unit; A second analysis unit is used to analyze the data to be analyzed based on a neural network model to obtain a model detection score of the device unit; the neural network model is pre-trained based on a historical data source; A first determination unit, configured to determine a comprehensive evaluation score of the equipment unit by using a preset weight model, the dynamic and static detection scores, and the model detection score; a second determination unit, configured to determine a first over-limit level of the comprehensive evaluation score and a second over-limit level of the dynamic and static detection score based on a preset threshold table, and to determine a third over-limit level of the model detection score according to a preset over-limit standard; a third determining unit, configured to determine a final over-limit level of the equipment unit by using the preset weight model, the second over-limit level and the third over-limit level; A generating unit is used to generate warning information based on the final over-limit levels of all equipment units and the first over-limit levels.

8. The device according to claim 7, characterized in that The device also includes: A third acquisition unit is used to acquire, for each of the equipment units, a device medical record book corresponding to the equipment unit; A storage unit is used to store the dynamic and static detection scores of the equipment unit, the model detection scores of the equipment unit, the comprehensive evaluation score of the equipment unit, the first over-limit level of the comprehensive evaluation score, the second over-limit level of the dynamic and static detection scores, the third over-limit level of the model detection scores and the final over-limit level of the equipment unit in the equipment medical record book corresponding to the equipment unit.

9. The device according to claim 7, characterized in that The device also includes: A fourth acquisition unit, configured to acquire a historical data source, wherein the historical data source is historical inspection data of a plurality of devices of each device unit in the track line; An output unit, used to input the historical data source into a neural network model and output a detection score corresponding to the historical data source; a fourth determining unit, configured to determine, based on a loss function, a difference between a predicted value of the detection score corresponding to the historical data source and a true value of the detection score corresponding to the historical data source; A judging unit, used to judge whether the difference is less than a preset threshold; A fifth determining unit, configured to determine that a final neural network model is obtained if the difference is less than a preset threshold; The sixth determination unit is used to update the parameters of the neural network model and return to execute the output unit if the difference is not less than a preset threshold.

10. The device according to claim 7, characterized in that The first analysis unit comprises: An acquisition module, used to acquire an analysis method of a dynamic and static detection score algorithm, wherein the analysis method includes a mean analysis or a peak analysis; An analysis module is used to analyze the manual static inspection data, the track inspection instrument static inspection data, the train instrument dynamic inspection data and the track inspection vehicle dynamic inspection data in the data to be analyzed according to the analysis method, and obtain the manual static unit score, the track inspection instrument static unit score, the train instrument dynamic unit score and the track inspection vehicle dynamic unit score; The calculation module is used to calculate the sum of the manual static unit score, the track inspection instrument static unit score, the added instrument dynamic unit score and the track inspection vehicle dynamic unit score to obtain the dynamic and static detection score of the equipment unit.

Images