Multi-module coupled double-unit traction turnout fault intelligent diagnosis method and device

Abstract

The invention discloses a multi-module coupled double-unit traction turnout fault intelligent diagnosis method and device. The method comprises the following steps: determining predicted fault types corresponding to a first extension set and a second extension set by adopting an integrated rank correlation metric algorithm; a predicted fault type combination is obtained, a preset conflict detection rule base is inquired based on the predicted fault type combination, and conflict detection rules meeting conditions are obtained; calculating to obtain a conflict detection value; if the conflict detection value is 1, a preset fault cause screening rule base is inquired based on predicted fault types corresponding to the first extension set and the second extension set, and a corresponding fault cause factor set is obtained; performing non-null logic judgment and cascade coding based on the fault-induced factor sets corresponding to the first extension set and the second extension set to obtain a coding result; querying a preset fault positioning coding table to obtain a corresponding fault cause range; and querying a preset maintenance support rule base, and outputting a fault processing suggestion. The accuracy and diagnosis depth of dual-computer fault analysis can be improved.

Description

Technical Field

[0001] This invention relates to the field of rail transit signal processing technology, and in particular to a method and device for intelligent diagnosis of faults in dual-motor traction turnouts with multi-module coupling. Background Technology

[0002] Turnouts (usually including switch machines) are the most basic operating equipment in rail transportation, used to change the direction of train traffic and enable train switching and reversing. Due to the complex structure of turnout systems, they have become a weak link in on-site equipment maintenance, and their failure rate remains consistently high. With the continuous acceleration of rail transit in my country, subway mainlines and large railway stations have generally installed dual-traction turnouts (i.e., using two switch machines simultaneously to pull the turnout) to ensure switching efficiency and reliability.

[0003] In addition to a single-machine control circuit for each traction point, dual-machine traction turnouts also include circuits for sequential start-up, fault protection, and series indication to ensure the synchronicity of dual-machine traction. Therefore, if one switch machine fails to start for any reason, the other switch machine will be protected by the control circuit even if it is functioning normally. Thus, when a dual-machine traction turnout malfunctions, it is common for each machine to show abnormal operation on the centralized signal monitoring system, and there are as many as four possible causes for the actual fault: (1) Machine 1 malfunctions but Machine 2 is functioning normally; (2) Machine 2 malfunctions but Machine 1 is functioning normally; (3) The common part of the circuits of Machine 1 and Machine 2 malfunctions; (4) Machine 1 and Machine 2 each have their own independent fault points. This complex situation undoubtedly exacerbates the difficulty of emergency handling of faults on site. Therefore, it is of great significance to develop reliable and efficient intelligent methods to assist in the on-site analysis of dual-machine traction turnout faults. Summary of the Invention

[0004] In order to achieve integrated functions of single-machine fault type identification, dual-machine fault root cause location, fault cause range confirmation, and fault handling strategy construction, so as to improve the accuracy and diagnostic depth of dual-machine fault analysis, this invention provides a multi-module coupled intelligent diagnostic method and device for dual-machine traction turnout faults.

[0005] In a first aspect, embodiments of the present invention provide a multi-module coupled intelligent fault diagnosis method for dual-machine traction turnouts, comprising: Based on the fault power curves corresponding to the first and second extensions and the preset fault template curve set, the predicted fault types corresponding to the first and second extensions are determined by the integrated rank correlation metric algorithm. Based on the predicted fault types corresponding to the first and second extensions, a combination of predicted fault types is obtained. Based on the combination of predicted fault types, a preset conflict detection rule library is queried to filter out conflict detection rules that meet the conditions. Each conflict detection rule includes a combination conflict flag corresponding to the combination of fault types of the two machines and the theoretical time range for monitoring corresponding to the first and second extensions to continue. Based on the actual monitoring durations of the first and second extensions, and in conjunction with the conflict detection rules that meet the conditions, the conflict detection value is calculated. If the conflict detection value is 1, then based on the predicted fault types corresponding to the first extension and the second extension, the preset fault cause filtering rule library is queried to obtain the fault cause set corresponding to the first extension and the second extension. Based on the fault cause set corresponding to the first and second extensions, non-empty logic judgment and cascaded coding are performed to obtain the coding result. Based on the encoding result, a preset fault location encoding table is queried to obtain the corresponding fault cause range; Based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data, the preset maintenance support rule base is queried, and fault handling suggestions are output.

[0006] Optionally, the fault template curve set includes multiple single template test groups, wherein each single template test group includes multiple fault template curves and a fault type label corresponding to each fault template curve; Based on the acquired fault power curves corresponding to the first and second extensions and the preset fault template curve set, the integrated rank correlation metric algorithm is used to determine the predicted fault types corresponding to the first and second extensions, including: Obtain the fault power curves corresponding to the first and second extensions; For the first and second extensions, each of the single template test groups is traversed, and the distance value between each fault template curve and the corresponding fault power curve in the single template test group is calculated one by one. The smallest distance value is selected from all the distance values, and the fault type label corresponding to the smallest distance value is used as the fault type to be tested. For the first extension and the second extension, the total number of votes for all the fault types to be tested is counted, and the fault type with the highest total number of votes is obtained. It is then determined whether the fault type with the highest total number of votes is unique. If so, the fault type with the highest total number of votes is taken as the predicted fault type. Otherwise, the corresponding total metric value is calculated based on each fault type with the highest total number of votes. The minimum total metric value is selected from all the total metric values, and the fault type with the highest total number of votes corresponding to the minimum total metric value is taken as the predicted fault type.

[0007] Optionally, the step of calculating the corresponding total metric value based on each of the tested fault types with the highest total votes, filtering out the minimum total metric value from all total metric values, and using the tested fault type with the highest total votes corresponding to the minimum total metric value as the predicted fault type includes: For each of the tested fault types with the highest total number of votes, for all distance values ​​in each single template test group, the distance value corresponding to the tested fault type with the highest total number of votes is extracted and accumulated to obtain the total metric value; The minimum total metric value is selected from all total metric values, and the test fault type with the highest total votes corresponding to the minimum total metric value is used as the predicted fault type.

[0008] Optionally, the combined conflict flag is 0 or 1; The conflict detection value is calculated based on the actual monitoring duration corresponding to the first and second extensions, combined with the conflict detection rules that meet the conditions, including: Obtain the actual monitoring duration corresponding to the first and second extensions; For the first and second extensions, based on the corresponding actual monitoring duration and the corresponding theoretical monitoring time range, a preset discrimination function is used to determine whether the corresponding actual monitoring duration is within the corresponding theoretical monitoring time range, and the corresponding judgment value is output. The judgment value is equal to 0 or 1, where 1 represents true and 0 represents false. The combined conflict flag is multiplied by the judgment values ​​corresponding to the first and second extensions to calculate the conflict detection value.

[0009] Optionally, each rule in the fault cause screening rule base includes a preset fault type for the first extension, a preset fault type for the second extension, a judgment result based on a first conditional judgment statement, and a judgment result based on a second conditional judgment statement; wherein, the first conditional judgment statement is that the set of fault cause factors of the preset fault type of the first extension directly or indirectly causes the second extension to have the corresponding preset fault type, and the judgment result is the set of fault cause factors of the first extension when the first conditional judgment statement is true; the second conditional judgment statement is that the set of fault cause factors of the preset fault type of the second extension directly or indirectly causes the first extension to have the corresponding preset fault type, and the judgment result is the set of fault cause factors of the second extension when the second conditional judgment statement is true. The method of querying a preset fault cause filtering rule base based on the predicted fault types corresponding to the first and second extensions yields a set of fault cause factors corresponding to the first and second extensions, including: Based on the predicted fault types corresponding to the first and second extensions, each rule in the fault cause filtering rule base is queried, and a matching rule is found to obtain the fault cause factor set corresponding to the first and second extensions.

[0010] Optionally, each rule in the maintenance support rule base includes a preset fault type for the first extension, a preset fault type for the second extension, a preset fault cause range, preset monitoring data, and a judgment result based on a third condition judgment statement. The third condition judgment statement is a set of the preset fault type for the first extension, the preset fault type for the second extension, the preset fault cause range, and the preset monitoring data, and the judgment result is a fault handling suggestion. Based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data, the system queries a preset maintenance support rule base and outputs fault handling suggestions, including: Based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data, query each rule in the maintenance support rule base, find the matching rule, and output fault handling suggestions.

[0011] Secondly, embodiments of the present invention provide a multi-module coupled intelligent fault diagnosis device for dual-machine traction turnouts, comprising: The single-unit fault classification module is used to determine the predicted fault type of the first and second extensions based on the acquired fault power curves and the preset fault template curve set, respectively, using an integrated rank correlation metric algorithm. The first query module is used to obtain a combination of predicted fault types based on the predicted fault types corresponding to the first extension and the second extension, and to query a preset conflict detection rule library based on the predicted fault type combination to filter out conflict detection rules that meet the conditions. Each conflict detection rule includes a combination conflict flag corresponding to the combination of fault types of the two machines and the theoretical time range for monitoring corresponding to the first extension and the second extension. The combined conflict calculation module is used to calculate the conflict detection value based on the actual monitoring duration corresponding to the first and second extensions and the conflict detection rules that meet the conditions. The judgment module is used to determine whether the conflict detection value is 1. If it is, the second query module queries the preset fault cause filtering rule library. The second query module is used to query the preset fault cause screening rule base based on the predicted fault types corresponding to the first extension and the second extension, and obtain the fault cause factor set corresponding to the first extension and the second extension. The encoding module is used to perform non-empty logic judgment and cascade encoding based on the fault cause factor sets corresponding to the first and second extensions to obtain the encoding result; The third query module is used to query a preset fault location coding table based on the coding result to obtain the corresponding fault cause range; The output module is used to query a preset maintenance support rule base and output fault handling suggestions based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data.

[0012] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent fault diagnosis method for dual-machine traction turnouts with multi-module coupling as described in the first aspect.

[0013] Fourthly, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent fault diagnosis method for dual-machine traction turnouts with multi-module coupling as described in the first aspect.

[0014] Fifthly, embodiments of the present invention provide a computer program product containing instructions that, when the computer program product is run on a computer device, cause the computer device to execute the intelligent fault diagnosis method for dual-machine traction turnouts with multi-module coupling as described in the first aspect.

[0015] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following: This invention provides a multi-module coupled intelligent fault diagnosis method for dual-machine traction turnouts. First, by acquiring the fault power curves of the sub-machines and a preset set of fault template curves, an integrated rank correlation metric algorithm is used to determine the predicted fault type of each sub-machine. Then, based on the actual monitoring duration of each sub-machine, combined with the combined conflict markers and the theoretical monitoring duration range of each sub-machine, conflict detection values ​​are calculated. Finally, cause localization is performed to determine the fault cause range and output fault handling suggestions. This avoids the logic of "first determining the root cause of the fault, then identifying the fault type" in the prior art. By changing the logical order, this invention avoids the "randomness" of root cause localization caused by the similarity of fault curves, making the localization of the fault root cause more accurate and reliable, and effectively improving the accuracy of fault diagnosis.

[0016] This invention not only determines the predicted fault type, but also queries the fault cause screening rule base based on the predicted fault type combination to obtain the fault cause factor set corresponding to each sub-unit. Through operations such as encoding, the fault cause range is determined, and finally, the maintenance support rule base is queried to output fault handling suggestions. This improves the diagnostic depth from the data level to the knowledge level, breaks through the bottleneck of existing turnout fault diagnosis technology, and provides more targeted and guiding suggestions for fault handling. It helps to solve turnout fault problems more efficiently and improve maintenance efficiency and accuracy.

[0017] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The flowchart is as proposed in the literature [1] "Research on Fault Diagnosis of Multi-Machine Traction Turnout Based on Fretchet Distance and TWSVM"; Figure 2 This is a flowchart of the intelligent fault diagnosis method for dual-machine traction turnout with multi-module coupling provided in the embodiments of the present invention; Figure 3 This is another flowchart of the intelligent fault diagnosis method for dual-machine traction turnout with multi-module coupling provided in this embodiment of the invention; Figure 4 This is an example of a dual-machine traction turnout fault diagnosis system provided in an embodiment of the present invention; Figure 5 A comparison diagram of the dual-machine fault location test results of the method provided in the embodiments of the present invention and the method in reference [1] is shown; Figure 6 This is a schematic diagram of a multi-module coupled dual-machine traction turnout fault intelligent diagnosis device provided in an embodiment of the present invention. Detailed Implementation

[0020] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0021] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," "far," "near," "front," and "rear," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0022] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] The inventors discovered that existing intelligent fault diagnosis methods for single switch machine turnouts are quite abundant, with two typical implementation modes: The first mode typically uses the turnout operation curve as a carrier and utilizes similarity measurement to identify the fault type. This mode belongs to the model-driven method, that is, it determines the fault type by calculating the similarity between the fault curve to be identified and the fault template curve, and takes the closest template fault type as the diagnosis result. The advantage of this type of method is that it is simple and efficient, and does not require collecting sufficient fault samples to train the model, so it is easy to apply in engineering. The main disadvantage is that each type uses a single curve template for measurement, which has a certain degree of one-sidedness, resulting in the method's accuracy and generalization ability needing improvement. The second mode achieves turnout fault classification through data-driven machine learning. It uses machine learning algorithms to learn and train a classifier from the features of known fault samples, and then uses it to identify the type of unknown fault samples. The advantages of this type of method are good accuracy and generalization performance, but the disadvantages are also obvious: First, there are few fault samples that can be used for training in the field and they are imbalanced, which can easily lead to overfitting of the algorithm. It is often necessary to increase the training cost to improve the stability of classification. Second, data-driven methods represented by deep learning have black box characteristics. The decisions made are usually not easy to explain, verify and improve, which makes it difficult to gain the trust of users in the low-fault-tolerance field of rail transportation.

[0024] Compared to the increasingly advanced intelligent diagnosis methods for single-machine turnout faults, there are very few research results on intelligent fault diagnosis for turnouts with dual or more traction machines (dual-machine traction and multi-machine traction are consistent in circuit principle, both belonging to combined circuits, only the number of machines is different).

[0025] Wu Yongcheng proposed a fault diagnosis method for multi-motor traction turnouts (see: Wu Yongcheng, Yang Changqiong, He Tao. Research on Fault Diagnosis of Multi-motor Traction Turnouts Based on Fretchet Distance and TWSVM [J]. Journal of Railway Science and Engineering, 2019, 16(11): 2866-2872.). The overall idea of ​​this method is to "first determine the root cause of the fault, and then identify the fault type". Figure 1The specific implementation process of this method is as follows: First, the Friesian distance is used as a similarity metric algorithm to calculate the similarity between the power curve of each switch and the standard power curve, and the switch with the lowest similarity is taken as the fault source; then, a two-way support vector machine model is trained using a known fault sample set to determine the specific fault type of the fault source. This method has the following obvious drawbacks: First, the logic of "first determining the root cause of the fault, then identifying the fault type" is unreasonable, leading to unstable root cause localization and affecting accuracy. This is because some fault curves with high incidence rates (such as 1DQJ, where the first starting relay self-closing circuit is disconnected, indicating a non-connected circuit, etc.) are very close to the standard curve, so they may be excluded as "fault root causes" in the first step, resulting in a certain degree of "randomness" in root cause localization. Second, the diagnostic depth of the method remains at the data level and cannot reach the knowledge level. That is, it can only ultimately identify the fault type, but cannot determine the scope of the fault cause based on the fault type and provide fault handling suggestions. This is also a bottleneck of existing switch fault diagnosis technology.

[0026] To address the aforementioned issues, the inventors developed a multi-module coupled intelligent diagnostic method and device for dual-machine traction turnout faults. This method integrates single-machine fault type identification, dual-machine fault root cause location, fault cause range confirmation, and fault handling strategy development, thereby improving the accuracy and diagnostic depth of dual-machine fault analysis.

[0027] Example 1 This embodiment proposes a multi-module coupled intelligent fault diagnosis method for dual-motor traction turnouts. (See reference...) Figure 2 The method takes fault monitoring data from each extension unit as input, and processes it sequentially through three modules: a single-unit fault classification module, a combined conflict detection module, and a fault location analysis module, outputting fault handling suggestions. The execution logic of this method consists of three stages. The first stage involves the single-unit fault classification module performing preliminary fault identification, obtaining the fault power curves of two extension units (extension J1 and extension J2), and using an integrated rank correlation metric algorithm to obtain the predicted fault types (i.e., faults of the two extension units) for each extension unit. Figure 2 (e.g., J1 is "blocked", J2 is "power supply abnormality").

[0028] The second stage involves a combined conflict detection module performing dual-machine logical conflict detection. The purpose is to check whether the predicted fault types of the two sub-units are logically reasonable. The predicted fault types of the two sub-units are combined into a "fault pair" (e.g., <three-phase power loss, unlocking jammed circuit>). A matching rule is searched in the conflict detection rule base. Each rule defines whether a certain fault combination is reasonable (if reasonable, the combined conflict is marked as 1; otherwise, it is 0), and the theoretical range of the monitoring duration for the two sub-units when this fault combination occurs. The actual monitoring duration of the two sub-units (hereinafter referred to as actual time) is obtained and compared with the theoretical range in the matched rule. If the actual time of both sub-units is within their respective theoretical ranges and the combined conflict is marked as 1, the logic is considered self-consistent, the conflict detection value is 1 (no conflict), and the process continues. If the actual time of any sub-unit exceeds the theoretical range, the conflict detection value is not 1, the process terminates, and manual processing is initiated. This prevents erroneous conclusions from affecting fault diagnosis, thus avoiding misdiagnosis. Although the designed combined conflict detection module cannot directly improve the accuracy of fault analysis, it has a certain probability of discovering the fault type identification results of a single machine with problems based on the logical constraints of the dual-machine traction action. This allows for timely interruption of fault diagnosis to avoid giving incorrect conclusions and amplifying the impact of the fault, which is of practical significance on site.

[0029] The third stage involves the fault location and analysis module providing cause localization and maintenance guidance. After conflict detection, based on the confirmed combination of dual-machine fault types, conditional fault cause screening is performed from the cause screening rule base to narrow down the cause range and identify the fault cause sets for each sub-machine. Logical judgment (non-empty judgment) and cascaded coding are then performed on the fault cause sets of each sub-machine, converting them into a code result that can be quickly queried by the machine. The fault location coding table is then consulted to further narrow down the possible causes, resulting in a more precise final fault cause range (e.g., "internal mechanical jamming of the switch machine" or "protective relay not activating"). Combining the fault type of each sub-machine, the final fault cause range, and specific centralized signal monitoring data (e.g., current and voltage values), the maintenance support rule base is queried, and maintenance support (i.e., fault handling) suggestions are output (e.g., "check if the connector at point X is loose," "replace cylinder Y").

[0030] When performing the method of this embodiment, refer to Figure 3 Specifically, this may include the following steps: Step S101: Based on the obtained fault power curves corresponding to the first extension and the second extension respectively and the preset fault template curve set, the predicted fault type corresponding to the first extension and the second extension is determined by the integrated rank correlation metric algorithm. Step S102: Based on the predicted fault types corresponding to the first extension and the second extension, obtain the predicted fault type combination, and query the preset conflict detection rule library based on the predicted fault type combination to filter out the conflict detection rules that meet the conditions. Each conflict detection rule includes the combined conflict flag corresponding to the dual-machine fault type combination and the theoretical time range for monitoring corresponding to the first extension and the second extension. Step S103: Based on the actual monitoring durations of the first and second extensions, and combined with the conflict detection rules that meet the conditions, calculate the conflict detection value. Step S104: Determine if the conflict detection value is 1. If it is, proceed to step S105. Step S105: Based on the predicted fault types corresponding to the first extension and the second extension, query the preset fault cause filtering rule base to obtain the fault cause factor set corresponding to the first extension and the second extension. Step S106: Perform non-empty logic judgment and cascade coding based on the fault cause factor sets corresponding to the first and second extensions to obtain the coding result; Step S107: Based on the coding results, query the preset fault location coding table to obtain the corresponding fault cause range; Step S108: Based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data, query the preset maintenance support rule base and output fault handling suggestions.

[0031] Step S101 corresponds to the first stage described above, steps S102-S104 correspond to the second stage described above, and steps S105-S108 correspond to the third stage described above. To provide a clearer explanation of the above-described intelligent fault diagnosis method for dual-machine traction turnouts with multi-module coupling, a detailed explanation of each step is provided below.

[0032] In step S101 above, the fault template curve set includes multiple single-template test groups, and each single-template test group includes multiple fault template curves and a fault type label corresponding to each fault template curve. It is worth noting that normal curves are also regarded as a special type of fault template curve to facilitate modeling. When defining the fault type label corresponding to a normal curve, it is marked as normal to distinguish it from other fault template curves.

[0033] The specific process of constructing a set of fault template curves can be as follows: First, assign a unique fault type label to each known preset fault type; then, for each preset fault type, collect historical fault template curves, emphasizing diversity during collection (covering turnouts in different turnout areas, different operating periods, and different environmental conditions) to ensure the generalization ability of the fault template curves. Store all collected fault template curves in a pre-constructed template curve sample library, where each preset fault type has a corresponding template curve sample library; finally, using the Bootstrap sampling method, randomly select one fault template curve from each template curve sample library with replacement to form a single template test group. That is, each single template test group includes multiple fault template curves and the fault type label corresponding to each fault template curve. Sampling is performed continuously N times in the same way to obtain N single template test groups. All single template test groups form a set of fault template curves.

[0034] Based on the obtained fault power curves corresponding to the first and second extensions and the preset fault template curve set, the specific process of determining the predicted fault types corresponding to the first and second extensions using the integrated rank correlation metric algorithm may include the following steps: Step S1011: Obtain the fault power curves corresponding to the first extension and the second extension; In step S1011 above, after a fault occurs, the fault power curves corresponding to the first sub-unit J1 and the second sub-unit J2 can be obtained from the signal microcomputer monitoring to determine the predicted fault type of each sub-unit.

[0035] Step S1012: For the first extension and the second extension, traverse each single template test group and calculate the distance value between each fault template curve and the corresponding fault power curve in the single template test group. Filter out the smallest distance value from all the distance values ​​and use the fault type label corresponding to the smallest distance value as the fault type to be tested. In step S1012 above, the Spearman rank correlation distance can be used as the distance value. The type of fault to be tested can be determined using a single-template test set. The Spearman rank correlation distance is perfectly suited for time-series data such as turnout power curves, which focus more on trends than absolute values, and is more practical than traditional methods such as Euclidean distance. Based on the fault power curve P of the first substation, the process of calculating the distance value is illustrated using the first single-template test set as an example: In the first single-template test group, the template curve for the k-th type of fault is: The fault power curve P of the first extension unit and the template curve of the kth type of fault are as follows: Distance between Calculated using the following formula (1): In the above formula (1), , They are respectively , The Middle There are n elements, with a total of n elements. One element; Used to calculate rank. The value ranges from [0,2], and the smaller the value, the higher the correlation.

[0036] Spearman's rank correlation distance is used as the distance value because it better characterizes the nonlinear relationship and morphological features between the sequences (i.e., the fault power curve and the fault template curve) compared to other metrics. According to equation (1), the set of distance values ​​between the fault power curve P of the first extension unit and all fault template curves in the first single-template test group can be calculated. .

[0037] Then, using the minimum search function Find the smallest distance value in the set of distance values. and the corresponding fault type labels ,Right now: The fault type label corresponding to the smallest distance value is taken as the fault type to be tested.

[0038] Step S1013: For the first extension and the second extension, count the total number of votes for all the fault types to be tested, find the fault type with the highest total number of votes, and determine whether the fault type with the highest total number of votes is unique; if yes, proceed to step S1014; if no, proceed to step S1015. Step S1014: The fault type with the highest total number of votes is selected as the predicted fault type; Step S1015: Calculate the corresponding total metric value based on the fault type with the highest total votes. Select the minimum total metric value from all total metric values, and use the fault type with the highest total votes corresponding to the minimum total metric value as the predicted fault type. The specific process may include the following steps: Step S10151: For each fault type under test with the highest total number of votes, extract the distance value corresponding to the fault type under test with the highest total number of votes for all distance values ​​of each single template test group, and sum them up to obtain the total metric value. In step S10151 above, when two or more fault types under test receive the same highest total number of votes (e.g., "locking and jamming" and "friction coupling failure" both receive 500 votes), for "locking and jamming," for all distance values ​​in each single template test group, the distance values ​​between the fault template curve and the fault power curve P for all fault types labeled "locking and jamming" are found and accumulated to obtain a total metric value Sum_A. Similarly, a total metric value Sum_B is obtained for "friction coupling failure."

[0039] Step S10152: Select the minimum total metric value from all total metric values. The test fault type with the highest total votes corresponding to the minimum total metric value is used as the predicted fault type.

[0040] In step S10152 above, the magnitudes of all total metrics (such as Sum_A and Sum_B) are compared to obtain the test fault type with the highest total votes corresponding to the smallest total metric value. It has the highest average similarity and the smallest distance with the fault power curve P, and is therefore selected as the final predicted fault type.

[0041] In this embodiment, drawing on the bagging-based ensemble learning model in machine learning, an ensemble rank correlation metric algorithm is proposed to determine the predicted fault type. This algorithm is essentially a similarity metric model-driven method, but unlike existing similarity metric algorithms (which use only a single template curve for each fault type), this algorithm constructs a fault template curve set and sets N single-template test groups to calculate distance values. It combines the total number of votes and the total metric value to determine the predicted fault type. This avoids the limitations of existing methods that only use a single template curve for similarity measurement, improving recognition accuracy and generalization ability, thus enhancing robustness, confidence, and making the predicted fault type identification results more stable and accurate. Furthermore, using a similarity metric model-driven approach avoids the problems of limited samples and high training costs associated with data-driven methods. Using 120 turnout fault curves for testing, the ensemble rank correlation metric algorithm proposed in this embodiment achieves a recognition accuracy of 98.33%, higher than mainstream similarity metric methods already used for turnout fault identification, such as Fraser distance (93.33%), Hausdorff distance (95.83%), and dynamic time curvature (95%).

[0042] In step S102 above, combined conflict detection requires the combined implementation of single and dual-machine control principles and action sequence logic, which is impossible with existing data-driven methods. To solve this problem, this embodiment adopts a knowledge-driven approach, pre-designing a rule-based expert system. That is, the single and dual-machine control principles and action sequence logic are summarized into conflict detection rules in advance, a conflict detection rule base is constructed, and then the conflict detection rule base is queried based on factual data (predicted combinations of fault types) to select conflict detection rules that meet the conditions, thus completing the conflict detection task.

[0043] The process of building a conflict detection rule base may specifically include the following steps: First, input all possible combinations of dual-machine failure types. For each pair of dual-machine fault type combinations, manual analysis is performed based on the action logic constraints of dual-machine traction to determine whether such a pair of dual-machine fault type combinations exists, thus obtaining a combination conflict flag. , A value of 1 indicates the existence of this combination of dual-machine fault types. A value of 0 indicates that this combination of dual-machine fault types does not exist. For example, if the fault type of the first extension J1 is "1DQJ not energized," where 1DQJ represents the 1st start relay, then according to the sequential start mechanism, the 1DQJ of the second extension J2 also cannot be energized. Therefore, J2 should not exhibit any fault type other than "1DQJ not energized." If the dual-machine fault type combination is (1DQJ not energized, 1DQJ not energized), then... The value is 1, otherwise, The value is 0.

[0044] Then, by combining theory with practice, the theoretical time range for monitoring the first extension J1 under the corresponding fault type was derived. The theoretical time range for monitoring of the second extension unit J2 should be maintained. In conjunction with the above content, a conflict detection rule is established for domain knowledge, using the following formula (3) with the sentence structure "IF condition THEN conclusion": Each conflict detection rule includes a combination of dual-machine fault types and corresponding combination conflict flags, the theoretical time range for monitoring the first sub-machine, and the theoretical time range for monitoring the second sub-machine.

[0045] Finally, all the written conflict detection rules are integrated into a database or file to form a conflict detection rule base. .

[0046] Predicted fault types based on the first extension unit Predicted fault types of the second extension To obtain the predicted fault type combination And based on the combination of predicted fault types By querying the aforementioned conflict detection rule base and filtering for conflict detection rules that meet the conditions, the predicted fault types mentioned above are combined. As a retrieval condition, it is matched with the "dual-machine fault type combination" field of each conflict detection rule. The conflict detection rule that is equal to the field of "dual-machine fault type combination" is found. This matching process is the key-value pair query process. The final conflict detection rule is the conflict detection rule that meets the conditions.

[0047] In this embodiment, domain knowledge that is difficult to quantify (single and dual-machine control principles and action sequence logic) is transformed into computable conflict detection rules, thereby enabling logical verification of data-driven results and predicting fault types based on the first sub-machine. Predicted fault types of the second extension The system queries the conflict detection rule base and matches each conflict detection rule. This matching process is an efficient key-value pair query process with fast query speed.

[0048] In step S103 above, the specific process of calculating the collision detection value may include the following steps: Step S1031: Obtain the actual monitoring duration corresponding to the first and second extensions; Step S1032: For the first extension and the second extension, based on the corresponding actual monitoring duration and the corresponding theoretical monitoring duration range, a preset discrimination function is used to determine whether the corresponding actual monitoring duration is within the corresponding theoretical monitoring duration range, and the corresponding judgment value is output. The judgment value is equal to 0 or 1, where 1 represents true and 0 represents false. Step S1033: According to the following formula (4), multiply the combined conflict flag with the judgment values ​​corresponding to the first extension and the second extension to calculate the conflict detection value. : In equation (4): This indicates the preset discrimination function.

[0049] In step S104 above, it is determined whether the collision detection value is 1. If the collision detection value is 1, then... If the value is 1, it indicates that the collision detection has passed. Proceed to step S105. If the collision detection value is 1... A value of 0 indicates that the conflict detection failed and the process needs to be interrupted and handled manually.

[0050] In step S105 above, to achieve correct fault root cause location and fault cause confirmation, it is necessary to integrate multi-source knowledge such as single and dual-machine control principles, dual-machine traction control principles, and fault causes. This embodiment adopts a knowledge-driven approach, designing a rule-based expert system. Specifically, the single and dual-machine control principles, dual-machine traction control principles, and fault causes are pre-summarized into rules to obtain a cause-selection rule base. Then, based on factual data (predicted fault types corresponding to the first and second sub-machines), the cause-selection rule base can be queried to find matching rules, obtaining the fault cause factor set corresponding to the first and second sub-machines, thus completing the fault cause confirmation.

[0051] The process of constructing the cause screening rule base may specifically include the following steps: First, input all possible combinations of dual-machine fault types, and for each pair of dual-machine fault type combinations, denote the preset fault type of the first sub-machine J1 as event A, and the preset fault type of the second sub-machine J2 as event B. Then, manually determine the set of fault causes of the first sub-machine under event A based on the dual-machine traction control principle. (A collection of all possible causes for event A) and a collection of possible causes for the failure of extension J2 under event B. (A collection of all possible causes for event B). Then, given the event trigger symbol " Establish the following two propositions: Proposition 1 This means: There are inherent faults in the first extension unit J1 itself. Event A is triggered, and event A can directly or indirectly trigger event B.

[0052] Proposition 2 This means: There are inherent faults in the second extension unit J2 itself. Event B is triggered, and event B can directly or indirectly trigger event A.

[0053] The above propositions can be judged by on-site experts based on the dual-machine traction control principle, and a causal screening rule base can be established based on this domain knowledge. Each rule in the cause filtering rule base includes the preset fault type of the first extension (i.e., event A), the preset fault type of the second extension (i.e., event B), the judgment result based on the first condition judgment statement, and the judgment result based on the second condition judgment statement. The first condition judgment statement and the second condition judgment statement are shown in the following formula (5): In equation (5) above, the first judgment statement is the first conditional judgment statement, the condition being proposition 1 above, that is, the set of fault-causing factors of the preset fault type of the first extension directly or indirectly causes the second extension to have the corresponding preset fault type, and the judgment result is the set of fault-causing factors of the first extension when the first conditional judgment statement is true; the second judgment statement is the second conditional judgment statement, the condition being proposition 2 above, that is, the set of fault-causing factors of the preset fault type of the second extension directly or indirectly causes the first extension to have the corresponding preset fault type, and the judgment result is the set of fault-causing factors of the second extension when the second conditional judgment statement is true. Obviously, the rule being true is equivalent to filtering the fault causes, which is used to clarify the final range of fault causes.

[0054] Based on the aforementioned fault cause screening rule base, using the predicted fault types corresponding to the first and second extensions, each rule in the fault cause screening rule base is queried to find the matching rule, which is the predicted fault type of the first extension. Predicted fault types of the second extension As search criteria, the system matches the fields of the preset fault type for the first extension and the preset fault type for the second extension in each rule of the fault cause filtering rule base, respectively, to find fields that are equal to... and According to the rule, obtain the fault cause factor set of the first extension and the fault cause factor set of the second extension corresponding to the rule.

[0055] In this embodiment, based on the fault cause screening rule base, the root cause of the extension unit failure is located, that is, the fault cause set corresponding to the extension unit is found. Moreover, the range of the fault cause set is much smaller than the corresponding fault cause set, which further narrows the investigation scope and enables accurate fault location. This lays the foundation for outputting accurate and reasonable fault handling suggestions, thereby greatly improving maintenance efficiency.

[0056] In step S106 above, the specific process of obtaining the encoding result may include the following steps: According to the preset judgment function, the following formula (6) is used to sequentially perform non-empty logic judgment on the fault cause set of the first extension, the fault cause set of the second extension, and the intersection of the fault cause set of the first extension and the second extension, and the judgment result is concatenated and encoded to obtain the encoding result: In the above formula (6), This represents the preset judgment function, which is a true / false judgment function. The judgment result for an empty set is 0, and the judgment result for a non-empty set is 1. This indicates cascaded encoding.

[0057] By concatenating the logical judgment results (whether the set is empty) in a fixed order, a structured and unique encoded result is transformed. This encoded result contains rich information on the causal relationship of the fault, which is the basis for subsequent output of accurate fault handling suggestions.

[0058] In step S107 above, a priority decision is made in advance based on the theoretically existing encoding results. And clarify the scope of the cause of the fault. The data shown in Table 1 is obtained as follows: Table 1 Fault Location Coding and Priority Decision The basis for making the above priority decision is that the probability of both machines having independent fault points is much lower than that of a single point of failure or a common line failure. Therefore, in accordance with the principle of emergency fault handling, priority is given to faults caused by a single fault source that cause asynchronous failure of the two machines or a common line failure, and then the possibility of each machine having a fault is considered as a secondary decision.

[0059] By referring to Table 1 based on the coding results obtained in step S106 above, the priority decision and the scope of the fault cause can be determined.

[0060] In step S108 above, to provide reasonable fault handling suggestions, it is necessary to integrate multi-source knowledge, including single and dual-machine control principles and action sequence logic, work instructions, and other monitoring parameters. Therefore, in this embodiment, a knowledge-driven approach is adopted to design a rule-based expert system. This system summarizes the single and dual-machine control principles and action sequence logic, work instructions, and other monitoring parameters into rules to obtain a maintenance support rule base. Then, based on factual data (the scope of fault causes, the predicted fault types corresponding to the first and second sub-units, and the acquired monitoring data), the system can query the maintenance support rule base to find matching rules and output accurate and reasonable fault handling suggestions.

[0061] The process of building the maintenance support rule base may specifically include the following steps: First, for each combination of dual-machine fault types and the corresponding preset fault cause range H, combined with the three-phase current... , indicating voltage Based on the preset monitoring data and the on-site operation instructions as expert knowledge, establish the third conditional judgment statement shown in the following formula (7): In the above formula (7), the conditions are the preset fault type of the first extension J1, the fault type of the second extension J2, the preset fault cause range, and the set of preset monitoring data, and the judgment result is the fault handling suggestion E.

[0062] Each rule includes the preset fault type for the first extension, the preset fault type for the second extension, the preset fault cause range, preset monitoring data, and the judgment result based on the third conditional judgment statement. A support rule base is built and maintained based on all the rules. .

[0063] Based on the fault cause range, the predicted fault types corresponding to the first and second extensions, and the acquired monitoring data, query each rule in the aforementioned maintenance support rule base, find matching rules, that is, find the rules that match the preset fault type of the first extension, the preset fault type of the second extension, the preset fault cause range, and the preset monitoring data one by one with the predicted fault type of the first extension, the predicted fault type of the second extension, the fault cause range, and the monitoring data, and output the fault handling suggestions corresponding to the rule.

[0064] In this embodiment, an integrated rank correlation metric algorithm (Bootstrap sampling + Spearman rank correlation distance + integrated voting) is employed. This algorithm possesses strong anti-interference and fault tolerance capabilities, enabling accurate prediction of fault types and achieving single-machine fault type identification with high confidence. It addresses the issue of low accuracy in model-driven methods while avoiding the limitations of data-driven methods, such as limited samples and high training costs. Furthermore, a conflict detection rule base is introduced to detect conflicts based on the control principles and action sequence logic of single and dual machines. This fundamentally filters out logically impossible false diagnoses, significantly reducing false alarm and false negative rates. Compared to considering the predicted fault types of two sub-machines in isolation, the conflict detection rule base couples the predicted fault types of the two sub-machines for analysis, achieving information fusion and high reliability. This embodiment innovatively designs a cause-selection rule base. Based on propositions 1 and 2, it infers and selects the set of fault-causing factors most likely to lead to the predicted fault type. Based on the encoding results, it accurately locates the fault cause range, achieving precise fault location and root cause analysis, going beyond simple fault symptom labeling to transcend mere symptom diagnosis. Based on the maintenance support rule base, the final output is specific and actionable fault handling suggestions, which do not stop at the level of identifying fault types. This improves the depth of diagnosis from the data level to the knowledge level, breaking through the bottleneck of existing turnout fault diagnosis technology. It provides more targeted and guiding suggestions for fault handling, helps to solve turnout fault problems more efficiently, improves maintenance efficiency and accuracy, and enhances the accuracy and depth of dual-machine fault analysis.

[0065] Compared to the existing technology's logic of "first determining the root cause of the fault, then identifying the fault type," the method in this embodiment changes the logical order, adopting a "first single-machine fault type identification, then dual-machine fault root cause localization" approach to diagnose dual-machine traction turnout faults. It sequentially performs single-machine fault type identification, combined conflict detection, and fault location analysis, which better aligns with the working logic of manual handling of dual-machine traction turnout faults on-site. This avoids the "randomness" introduced by prior fault root cause localization in complex situations where both machines exhibit abnormal behavior, making fault root cause localization more accurate and reliable. Therefore, compared to existing technologies, it significantly improves the stability and accuracy of dual-machine traction turnout fault location diagnosis.

[0066] The core of the method in this embodiment is three rule bases (conflict detection rule base, fault cause screening rule base, and maintenance support rule base). These three rule bases are essentially a systematic organization of single and dual-machine control principles, expert experience, action sequence logic, and work instructions, forming a digital and structured rule base. This can mimic manual fault root cause localization in dual-machine systems. The method in this embodiment can effectively utilize this knowledge and solve the problem of weak diagnostic capabilities of pure data-driven models when there are few fault samples.

[0067] For example, a fault diagnosis system for dual-track traction turnouts was developed based on the S700K dual-track turnout. This example is an offline system, facilitating testing, optimization, and application demonstration. See [link / reference]. Figure 4 The system interface consists of three display windows: monitoring information, fault location analysis, and fault handling suggestions, which are shown in the accompanying case application. Using a dual-machine traction switch fault in a subway as a test case, the developed system sample is used for review and application, and the interface is analyzed as follows: Figure 4 As shown. The monitoring information may include the turnout operating direction and the monitoring duration and fault power curves corresponding to the first substation J1 and the second substation J2, respectively. Taking the fault power curves corresponding to the first substation J1 and the second substation J2, the predicted fault type of the first substation J1 is determined using an integrated rank correlation metric algorithm. The predicted fault type for extension J2 is "1DQJ self-closing circuit not formed". The phrase "1DQJ was cut off after forming an autism" is used. As a condition, the conflict detection rule base is matched step by step, and the conclusion is: and combined with the actual monitoring duration and The collision detection value is calculated to be 1 according to the above formula (4), thus passing the collision detection. After screening, the causes of the fault are determined. , By matching the fault cause filtering rule base and performing non-empty logic judgment and cascade coding according to the above formula (7), the coding result obtained is 100. Based on the fault location coding table, the fault cause range is narrowed down to the failure of BHJ of J1 to be attracted. The conclusion is explained according to the circuit: the failure of BHJ of J1 to be attracted will cause QDJ (cut-off relay) to fail to close automatically, and after QDJ is released and falls, it will directly cut off the self-closing circuit of 1DQJ of J2. Therefore, the failure of BHJ of J1 to be attracted indirectly triggers the cut-off of the self-closing circuit of 1DQJ of J2. The final fault handling suggestion given by the system is to first check BHJ of J1 and DBQ that supplies it, and then consider other possible problems. After manual search, the fault point in this case is that the BHJ coil of J1 is broken. Therefore, the system gives reasonable fault analysis and handling suggestions.

[0068] To further verify the reliability and accuracy of the method in this embodiment, dual-machine fault location tests were conducted using data from 43 collected dual-machine fault cases. The final fault location results were categorized into three scenarios: location failure (number 1), location success (number 2), and location error (number 3). See [link to documentation]. Figure 5 The results of dual-machine fault location tests were statistically analyzed using the method of this embodiment and the method in reference [1] (i.e., the above-mentioned research on multi-machine traction turnout fault diagnosis based on Fretchet distance and TWSVM). It can be found that the method of this embodiment had 41 cases with correct location, 1 case with incorrect location, and 1 case with location failure, with a total accuracy rate of 95.35%. In contrast, the method in reference [1] had only 27 cases with correct location, with an accuracy rate of 62.79%. The method of this embodiment has a higher accuracy rate, higher reliability, and higher accuracy.

[0069] Example 2 Based on the same inventive concept, see [reference] Figure 6 This application also proposes a multi-module coupled intelligent fault diagnosis device for dual-machine traction turnouts, comprising: The single-machine fault classification module 101 is used to determine the predicted fault type of the first and second extensions based on the acquired fault power curves corresponding to the first and second extensions and the preset fault template curve set, respectively, using an integrated rank correlation metric algorithm. The first query module 102 is used to obtain a combination of predicted fault types based on the predicted fault types corresponding to the first extension and the second extension, and to query a preset conflict detection rule library based on the combination of predicted fault types to filter out conflict detection rules that meet the conditions. Each conflict detection rule includes a combination conflict flag corresponding to the combination of fault types of the two machines and the theoretical time range for monitoring corresponding to the first extension and the second extension. The combined conflict calculation module 103 is used to calculate the conflict detection value based on the actual monitoring duration of the first extension and the second extension, combined with the conflict detection rules that meet the conditions. The judgment module 104 is used to determine whether the conflict detection value is 1. If it is, the second query module queries the preset fault cause filtering rule library. The second query module 105 is used to query a preset fault cause filtering rule base based on the predicted fault types corresponding to the first extension and the second extension, and obtain the fault cause factor set corresponding to the first extension and the second extension. Encoding module 106 is used to perform non-empty logic judgment and cascade encoding based on the fault cause factor sets corresponding to the first extension and the second extension to obtain the encoding result; The third query module 107 is used to query a preset fault location coding table based on the coding results to obtain the corresponding fault cause range; The output module 108 is used to query the preset maintenance support rule base and output fault handling suggestions based on the fault cause range, the predicted fault type corresponding to the first extension and the second extension, and the acquired monitoring data.

[0070] The combined conflict detection module of Embodiment 1 may include a first query module 102, a combined conflict calculation module 103 and a judgment module 104, and the fault location analysis module may include a second query module 105, an encoding module 106, a third query module 107 and an output module 108.

[0071] The intelligent diagnostic device for faults of dual-motor traction turnouts with multi-module coupling provided in this embodiment of the invention has a similar implementation principle and technical effect to that of Embodiment 1, and will not be repeated here.

[0072] Example 3 Based on the same inventive concept, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent fault diagnosis method for dual-machine traction turnouts with multi-module coupling as described in Embodiment 1.

[0073] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments; or it may exist independently and not assembled into the device / apparatus. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to Embodiment 1 of the present invention.

[0074] According to embodiments of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0075] Example 4 Based on the same inventive concept, this application also proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent fault diagnosis method for dual-machine traction turnout with multi-module coupling as described in Embodiment 1.

[0076] Example 5 Based on the same inventive concept, this application proposes a computer program product containing instructions. When the computer program product is run on a computer device, the computer device executes the intelligent fault diagnosis method for dual-machine traction turnout with multi-module coupling in Embodiment 1.

[0077] The principles by which the above-mentioned devices, clients, media, and related equipment in this embodiment of the invention solve the problem are similar to those of the aforementioned methods. Therefore, their implementation can refer to the implementation of the aforementioned methods, and repeated details will not be repeated.

[0078] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0079] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0080] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0081] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0082] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A method for intelligent diagnosis of double machine traction turnout fault in multi-module coupling, characterized in that, The method comprises the following steps: Based on the obtained fault power curve corresponding to the first and second extension sets respectively and the preset fault template curve set, the integrated rank correlation measure algorithm is used to determine the predicted fault type corresponding to the first and second extension sets; Based on the predicted fault type corresponding to the first and second extension sets, the predicted fault type combination is obtained, and the preset conflict detection rule library is queried based on the predicted fault type combination to screen the conflict detection rule meeting the condition, wherein each conflict detection rule includes the combination conflict label corresponding to the double-machine fault type combination and the theoretical time range that the monitoring of the first and second extension sets should last; Based on the obtained actual monitoring duration corresponding to the first and second extension sets, the conflict detection value is calculated based on the conflict detection rule meeting the condition; If the conflict detection value is 1, the fault cause screening rule library is queried based on the predicted fault type corresponding to the first and second extension sets to obtain the fault cause set corresponding to the first and second extension sets; Non-empty logical judgment and cascade coding are performed on the fault cause set corresponding to the first and second extension sets to obtain the coding result; The preset fault positioning coding table is queried based on the coding result to obtain the corresponding fault cause range; The preset maintenance support rule library is queried based on the fault cause range, the predicted fault type corresponding to the first and second extension sets and the obtained monitoring data to output the fault processing suggestion.

2. The method according to claim 1, characterized in that, The fault template curve set comprises a plurality of single template test groups, wherein each single template test group comprises a plurality of fault template curves and a fault type label corresponding to each fault template curve; Based on the obtained fault power curve corresponding to the first and second extension sets respectively and the preset fault template curve set, the integrated rank correlation measure algorithm is used to determine the predicted fault type corresponding to the first and second extension sets, which comprises the following steps: Obtain the fault power curve corresponding to the first and second extension sets; For the first and second extension sets, each single template test group is traversed respectively, and the distance value between each fault template curve in the single template test group and the corresponding fault power curve is calculated one by one to obtain the smallest distance value from all distance values, and the fault type label corresponding to the smallest distance value is taken as the to-be-tested fault type; For the first and second extension sets, the total number of all to-be-tested fault types is counted to obtain the to-be-tested fault type with the highest total number, and it is judged whether the to-be-tested fault type with the highest total number is unique; if yes, the to-be-tested fault type with the highest total number is taken as the predicted fault type; otherwise, the total metric value corresponding to each to-be-tested fault type with the highest total number is calculated, and the minimum total metric value is selected from all total metric values, and the to-be-tested fault type with the highest total number corresponding to the minimum total metric value is taken as the predicted fault type.

3. The method according to claim 2, characterized in that, The total metric value corresponding to each to-be-tested fault type with the highest total number is calculated, and the minimum total metric value is selected from all total metric values, and the to-be-tested fault type with the highest total number corresponding to the minimum total metric value is taken as the predicted fault type, which comprises the following steps: For each of the top total vote types of the to-be-tested fault, for all distance values of each of the single template test groups, the distance value corresponding to the top total vote type of the to-be-tested fault is extracted and accumulated to obtain a total metric value; From all the total metric values, the minimum total metric value is screened out, and the top total vote type of the to-be-tested fault corresponding to the minimum total metric value is taken as the predicted fault type.

4. The method according to claim 1, characterized in that, The combination conflict flag is 0 or 1; The conflict detection value is calculated based on the acquired actual monitoring duration corresponding to the first extension and the second extension and the conflict detection rule meeting the condition, including: Acquiring the actual monitoring duration corresponding to the first extension and the second extension; For the first extension and the second extension, based on the corresponding actual monitoring duration, in combination with the corresponding theoretical time range in which monitoring should be continued, whether the corresponding actual monitoring duration is within the corresponding theoretical time range in which monitoring should be continued is judged by a preset judgment function, and a corresponding judgment value is output, the judgment value is equal to 0 or 1, wherein 1 represents true and 0 represents false; The combination conflict flag is multiplied by the judgment value corresponding to the first extension and the second extension to calculate the conflict detection value.

5. The method according to claim 1, characterized in that, Each rule of the fault cause screening rule library includes a preset fault type of the first extension, a preset fault type of the second extension, a judgment result based on a first condition judgment statement, and a judgment result based on a second condition judgment statement; wherein the first condition judgment statement is that the fault cause set of the preset fault type of the first extension directly or indirectly causes the corresponding preset fault type of the second extension, and the judgment result is the fault cause set of the first extension for which the first condition judgment statement is true; the second condition judgment statement is that the fault cause set of the preset fault type of the second extension directly or indirectly causes the corresponding preset fault type of the first extension, and the judgment result is the fault cause set of the second extension for which the second condition judgment statement is true; The fault cause set corresponding to the first extension and the second extension is obtained by querying the preset fault cause screening rule library based on the predicted fault types of the first extension and the second extension, including: Based on the predicted fault types of the first extension and the second extension, each rule in the fault cause screening rule library is queried to find a matching rule, and the fault cause set corresponding to the first extension and the second extension is obtained.

6. The method according to claim 1, wherein, Each rule of the maintenance support rule library includes a preset fault type of the first extension, a preset fault type of the second extension, a preset fault cause range, a preset monitoring data, and a judgment result based on a third condition judgment statement, wherein the third condition judgment statement is a set of the preset fault type of the first extension, the preset fault type of the second extension, the preset fault cause range, and the preset monitoring data, and the judgment result is a fault handling suggestion; The fault handling suggestion is output by querying the preset maintenance support rule library based on the fault cause range, the predicted fault types of the first extension and the second extension, and the acquired monitoring data, including: Based on the fault cause range, the predicted fault type corresponding to the first and second extensions, and the obtained monitoring data, each rule of the maintenance support rule library is queried to find a matched rule, and a fault processing suggestion is output.

7. A device for intelligent diagnosis of double machine traction turnout fault of multi-module coupling, characterized in that, The method comprises the following steps: A single-machine fault classification module is configured to determine the predicted fault type corresponding to the first and second extensions based on the obtained fault power curve corresponding to each of the first and second extensions and a preset fault template curve set, and an integrated rank correlation measure algorithm. A first query module is configured to obtain a predicted fault type combination based on the predicted fault type corresponding to the first and second extensions, and query a preset conflict detection rule library based on the predicted fault type combination to filter a conflict detection rule meeting a condition, wherein each conflict detection rule comprises a combined conflict marker corresponding to a double-machine fault type combination and a theoretical time range during which monitoring of the first and second extensions should be continued. A combined conflict calculation module is configured to calculate a conflict detection value based on the obtained actual monitoring duration of the first and second extensions and the conflict detection rule meeting the condition. A judgment module is configured to determine whether the conflict detection value is 1, and if so, a second query module is configured to query a preset fault cause screening rule library. The second query module is configured to query the preset fault cause screening rule library based on the predicted fault type corresponding to the first and second extensions to obtain a fault cause set corresponding to the first and second extensions. An encoding module is configured to perform non-empty logical judgment and cascade encoding based on the fault cause set corresponding to the first and second extensions to obtain an encoding result. A third query module is configured to query a preset fault positioning encoding table based on the encoding result to obtain a corresponding fault cause range. An output module is configured to query a preset maintenance support rule library based on the fault cause range, the predicted fault type corresponding to the first and second extensions, and the obtained monitoring data to output a fault processing suggestion.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the multi-module coupled double-machine traction turnout fault intelligent diagnosis method of any one of claims 1-6.

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the multi-module coupled double-machine traction turnout fault intelligent diagnosis method of any one of claims 1-6.

10. A computer program product comprising instructions which, when the computer program product is run on computer equipment, cause the computer equipment to perform the multi-module coupled double-machine traction turnout fault intelligent diagnosis method of any one of claims 1-6.

Images