Fault intelligent diagnosis method and system for direct current switch machine

Abstract

The invention relates to an intelligent fault diagnosis method and system for a direct current switch machine. The method comprises the steps that a turnout action current curve of the direct current switch machine is collected in real time; carrying out segmentation processing on the current curve through a curve characteristic analysis algorithm; identifying a global highest point of the current curve; positioning a starting point of a conversion stage after the global highest point; detecting a critical position at which the current value is changed from descending to stable change; positioning the starting point of the locking stage after the starting point of the conversion stage; detecting the position where the current value starts to rise after reaching the local lowest point; positioning an action end point after the starting point of the locking stage; detecting a zero position of a current value; inputting the segmented current curve into a pre-trained long short-term memory network classification model, and outputting a fault type classification result; performing operation feature analysis based on a segmentation result; and generating a diagnosis report containing fault positioning and maintenance suggestions and triggering pre-alarm. The problems of low manual analysis efficiency and high omission ratio are solved, and real-time automatic diagnosis of turnout faults is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of turnout fault diagnosis, in particular to a DC switch machine fault intelligent diagnosis method and system. BACKGROUND

[0002] In the prior art, manual means is usually used for identification and analysis of turnout fault curves: after turnout maintenance is completed, the current curve in the normal state is set as the reference curve of the turnout group on the signal centralized monitoring system. The current curve is checked regularly according to the specified period, and is compared with the reference curve. If the action time and current have large deviations from the reference curve, it indicates that the turnout operation state has changed greatly, and timely inspection and treatment should be performed. When the turnout action curve record is found to be poor or the current monitoring is inaccurate, the record is made and processed to ensure that the monitoring equipment is in good operation. The manual analysis means requires that the analyst has certain business experience, and when the turnout fails, the positioning of the equipment fault position takes a long time. For a large station, there are many turnout equipments, and manual analysis needs to consume a large amount of manpower and material resources, and the efficiency is low. SUMMARY

[0003] In order to solve the problems of low efficiency and high missed detection rate of manual analysis, and realize real-time automatic diagnosis of turnout fault, the present application provides a DC switch machine fault intelligent diagnosis method and system.

[0004] The above invention purpose of the present application is realized by the following technical scheme:

[0005] A DC switch machine fault intelligent diagnosis method, comprising the steps of:

[0006] real-time collection of a turnout action current curve of the DC switch machine;

[0007] segmentation processing of the current curve by a curve feature analysis algorithm;

[0008] input of the segmented current curve into a pre-trained long short-term memory network classification model, and output of a fault type classification result;

[0009] operation feature analysis based on the segmentation result;

[0010] generation of a diagnosis report containing fault positioning and maintenance suggestions, and triggering of a pre-alarm.

[0011] In a preferred example of the present application: the step of segmentation processing of the current curve by the curve feature analysis algorithm specifically comprises the steps of:

[0012] identification of a global maximum point of the current curve;

[0013] positioning of a transition stage starting point after the global maximum point: detection of a critical position where the current value changes from falling to stable change;

[0014] locating the start of the locking phase after the start of the transition phase: detecting the position where the current value starts to rise after reaching a local minimum;

[0015] locating the end of the action after the start of the locking phase: detecting the position where the current value returns to zero.

[0016] In a preferred example of the present application, the step of locating the end of the action after the start of the locking phase: detecting the position where the current value returns to zero. is followed by the steps of:

[0017] if the start of the transition phase is not identified, determining that the unlocking phase has failed;

[0018] if the start of the transition phase is identified but the start of the locking phase is not identified, determining that the transition phase has failed.

[0019] In a preferred example of the present application, the training process of the long short-term memory network classification model comprises:

[0020] training using a labeled current curve sample set, the samples containing normal states and various fault types;

[0021] adopting an early stopping mechanism to prevent overfitting and dynamically adjusting the learning rate;

[0022] saving the final model when the accuracy of the validation set reaches the preset standard.

[0023] In a preferred example of the present application, the step of performing running feature analysis based on the segmentation result comprises the steps of:

[0024] detecting whether the total duration of the action exceeds a preset range;

[0025] detecting whether the current values of the transition phase and the locking phase exceed the phase threshold;

[0026] detecting whether the fluctuation frequency of the current in the transition phase exceeds the fluctuation threshold.

[0027] In a preferred example of the present application, the step of detecting whether the fluctuation frequency of the current in the transition phase exceeds the fluctuation threshold comprises:

[0028] identifying a sequence of consecutive wave crests and troughs in the transition phase;

[0029] counting the number of times the difference between adjacent wave crests and troughs exceeds a set amplitude;

[0030] generating a fluctuation warning when the number of fluctuations exceeds the frequency threshold.

[0031] In a preferred example of the present application, the step of generating a diagnostic report containing fault location and maintenance recommendations and triggering a pre-warning comprises:

[0032] When the current is zero fault is identified, an indoor fuse and outdoor circuit maintenance scheme is output;

[0033] When the sawtooth fluctuation fault is identified, a motor commutator maintenance scheme is output.

[0034] The second application purpose is achieved by the following technical solutions.

[0035] A DC switch machine fault intelligent diagnosis system, comprising:

[0036] A data acquisition module is configured to acquire a turnout action current signal in real time.

[0037] An intelligent segmentation module is configured to perform segmented processing on the current curve by using a curve feature analysis algorithm.

[0038] A fault classification module is configured to input the segmented current curve into a pre-trained long short-term memory network classification model and output a fault type classification result.

[0039] A health assessment module is configured to perform operation feature analysis based on the segmentation result.

[0040] A decision output module is configured to generate a diagnosis report containing fault positioning and maintenance suggestions and trigger a pre-alarm.

[0041] The third application purpose is achieved by the following technical solutions.

[0042] A computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the DC switch machine fault intelligent diagnosis method according to any one of claims 1 to 7.

[0043] The fourth application purpose is achieved by the following technical solutions.

[0044] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the DC switch machine fault intelligent diagnosis method according to any one of claims 1 to 7.

[0045] In summary, the present application has at least one of the following beneficial technical effects.

[0046] The model automatically analyzes the real-time turnout curve collected by the centralized monitoring system, extracts the curve with the analysis result of fault, gives a pre-warning prompt, and gives the fault cause, maintenance suggestion, and treatment suggestion. In this way, the work efficiency of the analyst and the signal worker is greatly improved, and the requirements of manpower and material resources are reduced: the fault curve is quickly identified, and the fault position is quickly located. For large stations with many turnout devices, the effect is obvious. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is a flowchart of a DC switch machine fault intelligent diagnosis method of the present application;

[0048] Figure 2 is a schematic diagram of steps S10 to S30 in the embodiment of the present application;

[0049] Figure 3 is a schematic diagram of steps S40 to S50 in the embodiment of the present application. DETAILED DESCRIPTION

[0050] The present application will be further described in detail below in combination with the drawings.

[0051] In an embodiment, as shown in Figure 1 , the present application discloses a DC switch machine fault intelligent diagnosis method, which specifically includes the following steps:

[0052] S10: Real-time collection of turnout action current curve of DC switch machine;

[0053] S20: Subsection processing of current curve by curve feature analysis algorithm;

[0054] S30: Input of the subsectioned current curve into a pre-trained long short-term memory network classification model to output fault type classification result;

[0055] S40: Running feature analysis based on the subsection result;

[0056] S50: Generation of a diagnosis report containing fault positioning and maintenance suggestion and triggering of pre-warning.

[0057] As shown in Figure 2 , in an embodiment, step S20 specifically includes the following steps:

[0058] S21: Identification of global maximum point of current curve;

[0059] S22: Positioning of transition phase starting point after global maximum point: detection of critical position where current value changes from falling to stable change;

[0060] S23: Positioning of locking phase starting point after transition phase starting point: detection of position where current value starts to rise after reaching local minimum point;

[0061] S24: Positioning the end point of the locking phase after the start point of the locking phase: detecting the position where the current value returns to zero;

[0062] S25: If the start point of the transition phase is not identified, determining that the unlocking phase is faulty;

[0063] S26: If the start point of the transition phase is identified but the start point of the locking phase is not identified, determining that the transition phase is faulty.

[0064] In the embodiment, the switch action current curve is segmented, and the second-order difference algorithm is used to accurately divide the unlocking, transition and locking phase boundaries. The difference value of adjacent data points of the curve is calculated, which is the first-order difference. The difference value of adjacent elements of the first-order difference is calculated again, which is the second-order difference. The value of a data point of the curve is greater than the values of the left and right adjacent data points, and the absolute value of the difference is greater than a set threshold (0.05A in this paper), which is a peak value. The value of a data point of the curve is less than the values of the left and right adjacent data points, and the absolute value of the difference is greater than a set threshold (0A in this paper), which is a valley value. The maximum value in the data points of the whole curve is the global peak value.

[0065] For a specific action curve data list D[], the first-order difference list FD[], the second-order difference list SD[], the peak value list P[], the valley value list V[] and the global peak value MP can be accurately calculated. Experiments show that:

[0066] Start condition of the transition zone:

[0067] The first index of the point after the global peak value MP is i, and i satisfies:

[0068] -0.1<FD[i-1]<0, 0<SD[i-1]<0.08

[0069] Start condition of the locking zone:

[0070] The first index of the point after the start index i of the transition zone is j, and j satisfies:

[0071] j==V[k], D[j]>0, FD[j+1]>=0, j+10<=D.Count

[0072] (k is the index of V[] traversed from back to front, and the first index satisfying the above conditions) End condition of the locking zone:

[0073] The first index of the point after the start index j of the locking zone is k, and k satisfies:

[0074] D[k]==0

[0075] The indexes i, j and k of the start point of the transition zone, the start point of the locking zone and the end point of the locking zone are calculated, so that the turnout curve can be accurately divided into three segments (except for the points with 0 before and after the curve). As shown in FIG. 8. If there is no point satisfying the start condition of the transition zone, the action curve only exists in the unlocking zone, and the unlocking zone fails when the action occurs. If there is a point satisfying the start condition of the transition zone, but there is no point satisfying the start condition of the locking zone, the action curve only exists in the unlocking zone and the transition zone, and the transition zone fails when the action occurs. Figure 2

[0076] In an embodiment, the training process of the long short-term memory network classification model comprises:

[0077] The labeled current curve sample set is used for training, and the sample contains normal state and various fault types;

[0078] The early stopping mechanism is used to prevent overfitting, and the learning rate is dynamically adjusted;

[0079] When the accuracy of the verification set reaches the preset standard, the final model is saved.

[0080] In this embodiment, LSTM (Long Short-Term Memory) is a special type of recurrent neural network (RNN) that can solve the problem of gradient disappearance or gradient explosion when processing long-time dependencies. The structure of LSTM controls the flow of information through the "gate" mechanism, enabling it to effectively learn and remember information in long time series. LSTM is widely used in sequence prediction, time series analysis, natural language processing (NLP) and other fields. For the classification problem of turnout action curves, the LSTM model is also very suitable. Through the Tensorflow framework, a turnout action curve classification LSTM model is constructed. After dimension alignment (zero padding) of the turnout curve data, the curve data set (a total of 40,000, simulated by data expansion from actual normal and fault curves, a total of 10 categories, 4,000 in each category, all labeled data) is divided into a training set (70%), a verification set (10%) and a test set (20%). The training set is used for model training, the verification set is used for accuracy verification during model training, and the test set is used to calculate the accuracy of the trained model. During model training, early stopping is set to prevent overfitting; when the loss on the verification set no longer decreases, the learning rate is automatically reduced; a total of 30 rounds of training, batch_size = 32. After training, if the model accuracy is higher than 90%, the model file and optimal parameters are saved, so that subsequent new data can be predicted and breakpoint training based on the optimal parameters of the model.

[0081] As shown in FIG. 8, the start point of the transition zone is the point where the curve changes from the unlocking zone to the locking zone, and the end point of the transition zone is the point where the curve changes from the locking zone to the unlocking zone. Figure 3 ​As shown, in an embodiment, step S40 specifically comprises steps of:

[0082] S41: detecting whether the total action time length exceeds a preset range;

[0083] S42: detecting whether the current values of the conversion phase and the locking phase exceed phase threshold values;

[0084] S43: detecting whether the conversion phase current fluctuation frequency exceeds a fluctuation threshold value.

[0085] Step S43 specifically comprises steps of:

[0086] S431: identifying a sequence of consecutive wave crests and troughs in the conversion phase;

[0087] S432: counting the number of times that the difference between adjacent wave crests and troughs exceeds a set amplitude;

[0088] S433: generating a fluctuation alarm when the fluctuation number exceeds the frequency threshold value.

[0089] In the present embodiment, upper and lower limit values of the action time length, upper and lower limit values of the conversion and locking zones, and upper limit and fluctuation threshold value parameters of the conversion zone fluctuation number are added.

[0090] Action time length over-limit: the actual action curve time length is calculated as (curve point number * 40) / 1000, and then:

[0091] Time length over-limit condition:

[0092] Time length > upper limit | time length < lower limit

[0093] Each phase current over-limit: according to the implementation principle 1, the conversion zone start subscript i, the locking zone start subscript j, and the locking zone end subscript k have been calculated, and on this basis,

[0094] Conversion zone current over-limit condition:

[0095] V[m] > conversion zone upper limit | V[m] < conversion zone lower limit

[0096] (m is the V[] subscript from i+3 to j-3, 3 is the size of the sliding window, and error tolerance processing is performed)

[0097] Locking zone current over-limit condition:

[0098] V[n] > locking zone upper limit | V[n] < locking zone lower limit

[0099] (n is the v[] subscript from j to k)

[0100] Transition phase current curve fluctuation: traverse list D[] from i to j, calculate all extreme points index ex[] list of curve, and increase a sliding window size WS (here is 3), then:

[0101] Curve upward fluctuation condition:

[0102] Times (D[ex[o]]-D[ex[o]-1]>=fluctuation threshold)>=times upper limit

[0103] (ex[o]-ex[o-1]<WS)

[0104] Times (D[ex[o]]-D[ex[o]-ex[o-1]]>=fluctuation threshold)>=times upper limit

[0105] (ex[o]-ex[o-1]>=WS)

[0106] O is ex[] index from 1 to ex[].Count-1 traversal

[0107] Curve downward fluctuation condition:

[0108] Times (D[ex[o]-1]-D[ex[o]]>=fluctuation threshold)>=times upper limit

[0109] (ex[o]-ex[o-1]>WS)

[0110] Times (D[ex[o]-ex[o-1]]-D[ex[o]]>=fluctuation threshold)>=times upper limit

[0111] (ex[o]-ex[o-1]<=WS)

[0112] O is ex[] index from 1 to ex[].Count-1 traversal

[0113] Curve bidirectional fluctuation condition:

[0114] Curve upward fluctuation & curve downward fluctuation

[0115] The accuracy of the DC turnout action curve diagnosis model on the test set reached 99.82%, which was quite accurate for fault curve classification; and the second-order difference algorithm for curve segmentation was also very good for each specific curve (the offset error of the stage boundary line before and after it would not exceed 3 data points); the analysis of the data characteristics of each stage of the action curve (fluctuation, overrun) was also accurate. After deploying the operation and maintenance model to multiple field sites of the centralized monitoring system, the on-site experiment verification showed that the model could accurately segment, identify fault types, prompt pre-alarm, and give fault handling suggestions and maintenance recommendations for various fault curves of turnouts; the work efficiency of signal analysts and signal workers was improved, and consistent praise was received from field users.

[0116] After the operation and maintenance model is deployed, users can set parameters according to actual conditions; at the same time, users can also add actual running turnout action curve data to the knowledge base to realize the breakpoint continuous training of the model, update the model parameters, and make the model more suitable for the actual running conditions of field sites (new fault type labels and corresponding handling suggestions can also be added), thereby continuously improving the diagnosis effect of the model.

[0117] In an embodiment, step S50 specifically includes the following steps:

[0118] S51: when the current zero fault is identified, output an indoor fuse and outdoor circuit maintenance scheme;

[0119] S52: when the sawtooth fluctuation fault is identified, output a motor commutator maintenance scheme.

[0120] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0121] In an embodiment, a DC switch machine fault intelligent diagnosis system is provided, which corresponds to the DC switch machine fault intelligent diagnosis method described above. The system includes:

[0122] A data acquisition module for acquiring a turnout action current signal in real time;

[0123] An intelligent segmentation module for segmenting the current curve through a curve feature analysis algorithm;

[0124] A fault classification module for inputting the segmented current curve into a pre-trained long short-term memory network classification model and outputting a fault type classification result;

[0125] A health assessment module for performing operation feature analysis based on the segmentation result;

[0126] The decision output module is configured to generate a diagnostic report containing fault location and maintenance suggestions and trigger a pre-warning.

[0127] The specific limitation of the DC switch machine fault intelligent diagnosis system can refer to the limitation of the DC switch machine fault intelligent diagnosis method in the above, and will not be repeated here. Each module in the DC switch machine fault intelligent diagnosis system can be realized by software, hardware and their combination. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to call and execute the operation corresponding to each module by the processor.

[0128] In one embodiment, a computer device, which can be a server, is provided, and its internal structure diagram can be shown as follows. The computer device includes a processor, a memory, a network interface and a database connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the DC switch machine fault intelligent diagnosis method.

[0129] In one embodiment, a computer device is provided, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the DC switch machine fault intelligent diagnosis method.

[0130] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by the processor to implement the steps of the DC switch machine fault intelligent diagnosis method.

[0131] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0132] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is exemplified. In actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the above-described functions.

[0133] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for intelligent fault diagnosis of a DC switch machine, characterized in that, Including the following steps: Real-time acquisition of turnout operating current curves of DC switch machines; The current curve is segmented using a curve feature analysis algorithm; The segmented current curves are input into a pre-trained long short-term memory network classification model, which outputs the fault type classification results. Perform operational characteristic analysis based on segmentation results; Generate a diagnostic report containing fault location and maintenance suggestions, and trigger pre-alarms.

2. The intelligent fault diagnosis method for a DC switch machine according to claim 1, characterized in that, The step of segmenting the current curve using a curve feature analysis algorithm specifically includes the following steps: Identify the global peak of the current curve; After the global peak, locate the starting point of the transition phase: detect the critical position where the current value changes from decreasing to stable change; Locate the start of the latching phase after the start of the transition phase: the position where the current value begins to rise after reaching a local minimum point; Locate the end point of the action after the start of the locking phase: the position where the current value returns to zero.

3. The intelligent fault diagnosis method for a DC switch machine according to claim 2, characterized in that, Following the step of locating the end point of the action after the start of the locking phase: detecting the position where the current value returns to zero, the method further includes the following steps: If the start point of the transition phase is not identified, it is determined to be a failure in the unlocking phase; If the start point of the transition phase is identified but the start point of the locking phase is not identified, it is determined to be a transition phase failure.

4. The intelligent fault diagnosis method for a DC switch machine according to claim 1, characterized in that, The training process of the Long Short-Term Memory (LSTM) network classification model includes: Training was performed using a set of labeled current curve samples, which included normal conditions and various fault types. An early stopping mechanism is used to prevent overfitting, and the learning rate is dynamically adjusted. Save the final model when the accuracy of the validation set reaches the preset standard.

5. The intelligent fault diagnosis method for a DC switch machine according to claim 1, characterized in that, The steps for performing runtime feature analysis based on segmentation results specifically include the following steps: Check if the total duration of the detection action exceeds the preset range; Detect whether the current values ​​during the transition and latch-up phases exceed the phase thresholds; Detect whether the frequency of current fluctuations during the conversion phase exceeds the fluctuation threshold.

6. The intelligent fault diagnosis method for a DC switch machine according to claim 5, characterized in that, The step of detecting whether the current fluctuation frequency during the conversion phase exceeds the fluctuation threshold specifically includes: identifying the continuous peak and trough sequence during the conversion phase; Count the number of times the difference between adjacent peaks and troughs exceeds a set amplitude; A fluctuation alarm is generated when the number of fluctuations exceeds the frequency threshold.

7. The intelligent fault diagnosis method for a DC switch machine according to claim 1, characterized in that, The steps of generating a diagnostic report containing fault location and maintenance suggestions and triggering a pre-alarm specifically include: When a fault with zero current is identified, output a troubleshooting plan for the indoor fuse and the outdoor circuit. When a sawtooth fluctuation fault is identified, the output motor commutator maintenance plan is implemented.

8. A fault intelligent diagnosis system for DC switch machines, characterized in that, include: The data acquisition module is used to acquire the turnout operating current signal in real time; The intelligent segmentation module is used to segment the current curve using a curve feature analysis algorithm; The fault classification module is used to input the segmented current curve into a pre-trained long short-term memory network classification model and output the fault type classification result. The health assessment module is used to perform operational characteristic analysis based on segmented results; The decision output module is used to generate diagnostic reports that include fault location and maintenance suggestions, and to trigger pre-alarms.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent fault diagnosis method for a DC switch machine as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent fault diagnosis method for a DC switch machine as described in any one of claims 1 to 7.

Images