Pantograph status evaluation method based on multi-point triaxial acceleration measurement

Abstract

The present invention discloses a pantograph state evaluation method based on multi-point three-axis acceleration measurement, comprising the following steps: S1, collecting three-axis acceleration data of the pantograph bow angle; S2, discriminating and pre-processing the three-axis acceleration data to filter out abnormal data; S3, performing comparative analysis on the pre-processed three-axis acceleration data to screen and / or complete valid data, including historical comparison and cross comparison; S4, constructing a pantograph state evaluation system based on the valid data to evaluate the pantograph state. The present invention proposes a multi-point and multi-axis pantograph-net acceleration acquisition method, establishes a discriminating and analyzing system to verify the correctness of the data, and finally comprehensively evaluates the pantograph state based on multi-dimensional data.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit pantograph-net hard point detection, and in particular to a pantograph state evaluation method based on multi-point three-axis acceleration measurement. Background Art

[0002] With the rapid development of rail transit, the importance of pantograph-catenary pressure hard points has become increasingly prominent.

[0003] The installation of pressure sensors often requires changing the structure of the pantograph, which introduces certain safety hazards, while hard-point sensors only need to be installed in a hanging manner, which has almost no impact on the pantograph body. At present, hard-point sensors often only use their vertical acceleration, and the acceleration values ​​in the yaw direction and along the line are not collected and applied. On the other hand, the vertical acceleration value is often directly output after being collected. At present, there is no reasonable verification method for the correctness of the value itself, and there is no secondary mining and comprehensive application of the data. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a pantograph state evaluation method based on multi-point three-axis acceleration measurement to solve the problem of inaccurate pantograph state evaluation results.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a pantograph state evaluation method based on multi-point triaxial acceleration measurement, comprising the following steps:

[0006] S1. Collect the three-axis acceleration data of the pantograph angle;

[0007] S2, pre-process the three-axis acceleration data to filter out abnormal data;

[0008] S3, by comparing and analyzing the pre-processed three-axis acceleration data, to screen and / or complete the valid data, including historical comparison and cross comparison;

[0009] S4. Based on valid data, a pantograph status evaluation system is constructed to evaluate the pantograph status.

[0010] Furthermore: the three-axis acceleration data includes three-axis acceleration in the vertical direction, along-line direction and yaw direction.

[0011] Furthermore: in the step S2, the three-axis acceleration data is pre-processed, including abnormal jump point determination and filtering, and / or abnormal zero value point determination and filtering.

[0012] Further: the abnormal jump point determination and filtering includes:

[0013] S211, determining whether the triaxial acceleration data belongs to an abnormal jump point or a normal fault detection point;

[0014] S212. Using cluster analysis, locate and filter out abnormal jump points in the three-axis acceleration data.

[0015] Further: the abnormal zero value point identification and filtering includes:

[0016] S221, judging the abnormal zero value, including abnormal jump to 0 value and abnormal straight line;

[0017] S222. Filter out invalid three-axis acceleration data using gradient analysis.

[0018] Furthermore: the historical comparison in step S3 includes: using the repeatability of the data to construct a discrimination mechanism to achieve repeatability analysis and / or reliability analysis.

[0019] Further: the use of data repeatability to construct a discrimination mechanism is specifically:

[0020] S311. Using repeated measurement data variance analysis, locate the section where the pantograph acceleration changes abnormally;

[0021] S312, calculating and counting the correlation coefficient between the entire data and the historical test data to determine whether the current test data is reliable;

[0022] S313: When the correlation coefficient is greater than a preset confidence threshold R, the data is considered reliable.

[0023] Furthermore, the cross comparison in step S3 includes: data calibration and data completion of abnormal points through three-axis acceleration data fluctuation matching recognition, waveform inconsistency recognition and missing value reinforcement.

[0024] Further: the three-axis acceleration data fluctuation matching and identification includes:

[0025] Determine that the yaw acceleration direction of each detection point of the pantograph angle is consistent;

[0026] It is determined that the acceleration direction along the line of each detection point of the pantograph angle is consistent.

[0027] Further: the waveform inconsistency identification includes:

[0028] Determine whether the coaxial acceleration data of each detection point have nearly the same characteristics in the time domain;

[0029] The three-axis acceleration data is transformed into the frequency domain to determine whether the waveform vibration period of the same slider and the same direction is consistent, thereby determining whether the acceleration change period is consistent;

[0030] The Sieboldt-Huang transform is used to extract the time-frequency domain features of the triaxial acceleration data to determine whether there is only a time difference in the instantaneous energy of the coaxial data of the front and rear sliders, and whether the instantaneous energy of the previous and next moments is consistent.

[0031] Further: the missing value reinforcement includes:

[0032] Find the location of the abnormal point. When only a single axis and a single detection point data is abnormal, you can continue to use the data of the corresponding detection point on the same axis to perform an inverse transformation to complete the abnormal detection point data.

[0033] Further: the evaluation of the pantograph state in step S4 includes a steady state evaluation and a non-steady state evaluation.

[0034] Further: the steady state evaluation includes:

[0035] Trained through bp neural network;

[0036] Weighting different axial acceleration values ​​according to the training results;

[0037] Calculate the weighted square error of each axis acceleration;

[0038] When the sum of the variances is within a preset threshold range, it is considered that the pantograph is in a stable state;

[0039] When the sum of the variances is greater than a preset threshold, it is considered that the pantograph is in a non-stationary state.

[0040] Furthermore: the non-steady state includes left-right offset load, front-back offset load, vertical offset load, lateral offset load and along-line offset load.

[0041] Further: evaluating the left and right eccentric loads includes: determining whether the pull-out value distribution is uneven by analyzing and calculating the impact of the left and right slide plates;

[0042] The evaluation of the front and rear offset loads includes: determining whether the guide height slope is too large by analyzing and calculating the impact of the front and rear slide plates;

[0043] The vertical eccentric load is evaluated, including: analyzing and calculating the impact in the vertical direction, and determining whether the contact line is hard bent;

[0044] evaluating the yaw eccentric load, including determining whether the pull-out value distribution is too large by analyzing and calculating the impact in the yaw direction;

[0045] The evaluation of the eccentric load along the line includes: analyzing and calculating the impact along the line direction, and determining whether the contact line tension is insufficient.

[0046] Further: the evaluation of the left and right load is specifically as follows: if N is the preset threshold value. If N ≥ 20%, it means that the left and right sides of the slide are impacted too unevenly, and a status prompt is required. The pull-out value is unevenly distributed.

[0047] The evaluation of the front and rear load is specifically as follows: N is the preset threshold value. If N ≥ 20%, it means that the front and rear skateboards are impacted too unevenly, and a status prompt is required. The guide slope is too large.

[0048] The vertical eccentric load is evaluated as follows: M is the preset threshold, M≥2, that is, the vertical hard point variance is greater than the preset threshold M of the mean of the three-axis hard point variance, which means that the impact on the plumb bob is too large, and a status prompt is required, and the contact line is hard bent;

[0049] The evaluation of the yaw load is specifically as follows: M is the preset threshold, M≥2, that is, if the hard point variance in the yaw direction is greater than the preset threshold M of the mean of the hard point variances of the three axes, it means that the impact in the yaw direction is too large, and a status prompt is required, and the pull-out value distribution is too large;

[0050] The specific evaluation of the eccentric load along the line is as follows: M is the preset threshold, M≥2, that is, the hard point variance along the line is greater than the preset threshold M of the mean of the three-axis hard point variance, which means that the impact along the line is too large, and a status prompt is required, and the contact line tension is insufficient;

[0051] Among them, S ij is the variance of the j-th axis of the ith detection point.

[0052] The beneficial effects of the present invention are as follows: the present invention proposes a multi-point multi-axis pantograph acceleration acquisition method, establishes a discrimination analysis system to verify the correctness of the data, and finally comprehensively evaluates the pantograph status based on multi-dimensional data. The present invention adopts a multi-point multi-axis pantograph hard point acquisition method, adopts an effective acceleration data filtering and discrimination mechanism, effectively filters out abnormal acceleration data, and effectively evaluates the pantograph status through multi-axis multi-point acceleration data. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of the present invention;

[0054] Figure 2 It is a schematic diagram of the position of the acceleration measurement sensor in the present invention. DETAILED DESCRIPTION

[0055] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0056] like Figure 1 As shown, a pantograph state evaluation method based on multi-point triaxial acceleration measurement includes the following steps:

[0057] S1, collect three-axis acceleration data of four bow angles through acceleration measurement sensors;

[0058] like Figure 2 As shown, under the four bow corners of the pantograph, a set of acceleration measurement sensors are installed at four points respectively. Each sensor can detect three-axis acceleration in the vertical direction, along the line direction, and yaw direction.

[0059] It should be particularly emphasized that the hard point sensor is hung below the bow angle, does not invade the core current collection area of ​​the pantograph, and does not change the main structure of the pantograph, which greatly ensures the safety of the pantograph.

[0060] The high and low voltage isolation module, photoelectric conversion module, signal transmission module, data storage and processing module used in the acquisition system are not the focus of the present invention and will not be described in detail here.

[0061] S2. Pre-process the triaxial acceleration data, including filtering out abnormal jump points and abnormal zero-value points;

[0062] Since most of the domestic contact networks are powered by high voltage of AC27.5KV or DC1500V, the dynamic current drawing of the pantograph will bring about very complex electromagnetic effects. The stable acquisition of sensor signals in this environment is a major problem in the current field of contact network detection. In the actual acquisition process, abnormal jump points, value zeroing and other abnormalities sometimes occur. In order to accurately evaluate the current movement state of the pantograph, the present invention mentions a preprocessing method for the collected data.

[0063] Abnormal jump point identification: The system has a reasonable mechanism to determine whether it is an abnormal jump point or a normal fault detection point, and uses cluster analysis to locate and filter out abnormal jump points in the data.

[0064] Zero value discrimination: The system has the ability to discriminate abnormal zero values, including abnormal jumps to 0 values ​​and abnormal straight line pulling of data. It uses gradient analysis to filter out invalid collected data.

[0065] S3, by comparing and analyzing the pre-processed three-axis acceleration data, to screen and / or complete the valid data, including historical comparison and cross comparison;

[0066] The overhead line is a purely physical structure, so the detection data of each detection vehicle should have a certain degree of repeatability. The detection data in some cases is extremely distorted and should be effectively filtered out. However, the speed, environment, and time of each vehicle detection are different. Therefore, how to use the repeatability of data to establish a reasonable discrimination mechanism is one of the difficulties in data analysis. The present invention mentions the following historical comparative analysis method.

[0067] Repeatability analysis: The system has the ability to identify the repeatability of historical curves, mainly using variance analysis of repeated measurement data to locate the section where the pantograph acceleration changes abnormally;

[0068] Reliability analysis: The system has a credibility test for the curve, which is mainly used to evaluate whether the current test data is reliable. By statistically calculating the correlation coefficient r(x, y) between the entire data and the historical test data, when the correlation coefficient is greater than the set credibility threshold R, the data is considered reliable.

[0069] The formula for calculating the correlation coefficient is:

[0070]

[0071] In the above formula, Cov(x,y) is the covariance of two data, Var[x] and Var[y] are the variances of data x and y respectively.

[0072] As shown in the acquisition system, 4 bow angle points and a total of 12-axis acceleration curves are collected. Under the actual operating environment, these waveforms should have a certain correlation, which can be used to perform discriminant analysis of the detection data and filter and calibrate a small number of abnormal points in the data of certain dimensions. The present invention mainly mentions the following method.

[0073] Fluctuation matching and identification of collected data: Due to installation process, electromagnetic interference, sensor status and other reasons, some axial acceleration curves of some detection points may be detected incorrectly. Therefore, certain discrimination analysis is required. Combined with the pantograph operation mechanism, the yaw direction is the x-axis direction, the line direction is the y-axis direction, and the vertical direction is the z-axis direction. The yaw acceleration direction of the four detection points is kept consistent, that is, is the acceleration vector of the x-axis of the ith monitoring point, is the acceleration vector of the jth monitoring point along the y axis; let the acceleration direction along the line of detection points 1, 2 and detection points 3, 4 be consistent, that is, They are the acceleration vectors of detection points 1, 2, 3, and 4 on the y-axis respectively;

[0074] The waveform inconsistency identification is specifically as follows: the characteristics (mean peak value standard deviation) of the coaxial acceleration data of each detection point in the time domain should differ by less than 1% of the total value, that is,

[0075] ave[a 1j ]≈ave[a 2j ]≈ave[a 3j ]≈ave[a 4j ]

[0076] max[a 1j ]≈max[a 2j ]≈max[a 3j ]≈max[a 4j ]

[0077] σ[a 1j ]≈σ[a 2j ]≈σ[a 3j ]≈σ[a 4j ]

[0078] When the data is transformed into the frequency domain (Fourier), the waveform vibration period of the same slider in the same direction must be consistent, and the acceleration change period of detection points 1 and 4 must be consistent with the acceleration change period of detection points 2 and 3, that is,

[0079]

[0080] By using the HHT transform to extract the time-frequency domain features of the data, the instantaneous energy of the coaxial detection data of the front and rear sliders will only have a time difference, and the instantaneous energy of the front and rear moments will remain consistent, that is,

[0081] E 1j (t1) = E 4j (t1+Δt)

[0082] E 2j (t1) = E 3j (t1+Δt)

[0083] The missing value reinforcement is specifically as follows: using multi-point multi-axis hard point data discrimination analysis to find the location of the abnormal point. When only a single axis and a single detection point data is abnormal, the data of the corresponding detection point on the same axis can be used for inverse transformation to complete the abnormal detection point data, that is,

[0084] a ij '=f(a nj ),n≠i

[0085] S4. Based on valid data, a pantograph status evaluation system is constructed to evaluate the pantograph status.

[0086] After the above data have passed the discriminant analysis test and confirmed that all the test data meet the basic physical conditions, the hard point data can be used to construct a pantograph motion state evaluation system, which is mainly divided into the pantograph stable state and the pantograph non-stable state.

[0087] 1. Steady state

[0088] The stable state of the pantograph is the most ideal dynamic operation state of the pantograph network, which means that the pantograph operates stably and fluctuates evenly. The influence of different axial accelerations on the current collection of the pantograph network is different. The present invention uses a genetic BP neural network, different axial accelerations as independent variables, and the instantaneous current value obtained by the vehicle as the dependent variable. The different axial acceleration values ​​are weighted by the BP neural network training results:

[0089] I = f(a j ,θ)

[0090] In the above formula, I is the instantaneous current, a j is a certain axial acceleration, θ is the weight;

[0091] After adding weights to the acceleration of each axis, calculate the weighted square sum of the acceleration of each axis and set the threshold σ max It is considered that the pantograph corresponding to the axial acceleration within the threshold range is in a stable current receiving state.

[0092] In addition, when the overall state is stable, the local variance in some sections may be too large during segmentation. In this case, it can be considered that the quality of the bow-net in this section is poor, which may lead to unstable current receiving in the bow-net, etc. The defects of the local sections can also be identified based on this.

[0093] 2. Non-stationary state

[0094] When the hard point variance analysis is greater than a certain threshold, the pantograph is considered to be in a non-stable state. From the business perspective, this system gives the direction of the abnormal problem to guide the repair of the pantograph-network problem.

[0095] Remember S ij Represents the variance of the j-th axis of the ith detection point, and the following state evaluation system can be constructed.

[0096] The left and right eccentric loads are specifically: This means that the left and right sides of the skateboard are impacted too unevenly, and a status prompt is required. Possible causes include uneven distribution of pull-out values.

[0097] The front and rear bias load is specifically: This means that the front and rear skateboards are impacted unevenly and a status prompt is required. Possible causes include excessive slope.

[0098] The vertical eccentric load is specifically: That is, if the vertical hard point variance is greater than twice the mean of the three-axis hard point variance, it means that the impact on the plumb bob is too large and a status prompt is required. The possible cause is the hard bend of the contact line.

[0099] The lateral load is specifically: That is, if the hard point variance in the yaw direction is greater than twice the mean of the hard point variances of the three axes, it means that the impact in the yaw direction is too large and a status prompt is required. The possible cause is that the pull-out value distribution is too large;

[0100] The specific load along the line is: That is, if the hard point variance along the line is greater than twice the mean of the hard point variance of the three axes, it means that the impact along the line is too large and a status prompt is required. The possible cause is insufficient contact line tension.

Claims

1. A pantograph state evaluation method based on multi-point triaxial acceleration measurement, characterized in that: The following steps are involved: S1, collect the three-axis acceleration data of the four bow angles of the pantograph; S2, pre-process the three-axis acceleration data to filter out abnormal data; S3, by comparing and analyzing the pre-processed three-axis acceleration data, to screen and / or complete the valid data, including historical comparison and cross comparison; S4. Based on the effective data, a pantograph status evaluation system is constructed to evaluate the pantograph status; The historical comparison in step S3 includes: using the repeatability of the data to construct a discrimination mechanism to achieve repeatability analysis and / or reliability analysis; the use of the repeatability of the data to construct a discrimination mechanism is specifically: S311. Using repeated measurement data variance analysis, locate the section where the pantograph acceleration changes abnormally; S312, calculating and counting the correlation coefficient between the entire data and the historical test data to determine whether the current test data is reliable; S313, when the correlation coefficient is greater than a preset confidence threshold R, the data is considered reliable; The cross comparison in step S3 includes: data calibration and data completion of abnormal points through three-axis acceleration data fluctuation matching recognition, waveform inconsistency recognition and missing value reinforcement; In step S4, the evaluation of the pantograph state includes a steady state evaluation and a non-steady state evaluation; The steady state evaluation includes: Trained through bp neural network; Weighting different axial acceleration values ​​according to the training results; Calculate the weighted square error of each axis acceleration; When the sum of the variances is within a preset threshold range, it is considered that the pantograph is in a stable state; When the sum of the variances is greater than a preset threshold, it is considered that the pantograph is in a non-stationary state; The non-stationary state includes left-right eccentric load, front-back eccentric load, vertical eccentric load, lateral eccentric load and along-line eccentric load; The evaluation of the left and right eccentric load is specifically as follows: , N is the preset threshold, N ≥ 20%, which means that the left and right sides of the skateboard are impacted too unevenly, and a status prompt is required, and the pull-out value is unevenly distributed; The evaluation of the front and rear load is specifically as follows: , N is the preset threshold, N ≥ 20%, which means that the front and rear skateboards are impacted too unevenly, and a status prompt is required, and the guide slope is too large; The vertical eccentric load is evaluated as follows: , M is the preset threshold, M≥2, that is, the vertical hard point variance is greater than the preset threshold M of the mean of the three-axis hard point variance, which means that the impact on the plumb bob direction is too large, and a status prompt is required, and the contact line is hard bent; The evaluation of the yaw load is specifically as follows: , M is the preset threshold, M≥2, that is, the hard point variance in the yaw direction is greater than the preset threshold M of the mean of the three-axis hard point variance, which means that the impact in the yaw direction is too large, and a status prompt is required, and the pull-out value distribution is too large; The specific evaluation of the eccentric load along the line is as follows: , M is the preset threshold, M≥2, that is, the hard point variance along the line is greater than the preset threshold M of the mean of the three-axis hard point variance, which means that the impact along the line is too large, and a status prompt is required, and the contact line tension is insufficient; Among them, S ij is the variance of the j-th axis of the ith detection point.

2. The pantograph state evaluation method based on multi-point triaxial acceleration measurement according to claim 1 is characterized in that: The three-axis acceleration data includes three-axis accelerations in the vertical direction, along-line direction, and yaw direction.

3. The pantograph state evaluation method based on multi-point triaxial acceleration measurement according to claim 1 is characterized in that: In the step S2, the three-axis acceleration data is pre-processed, including abnormal jump point determination and filtering, and / or abnormal zero value point determination and filtering.

4. The pantograph state evaluation method based on multi-point triaxial acceleration measurement according to claim 3 is characterized in that: The abnormal jump point identification and filtering includes: S211, determining whether the triaxial acceleration data belongs to an abnormal jump point or a normal fault detection point; S212. Using cluster analysis, locate and filter out abnormal jump points in the three-axis acceleration data.

5. The pantograph state evaluation method based on multi-point triaxial acceleration measurement according to claim 3 is characterized in that: The abnormal zero value point identification and filtering includes: S221, judging the abnormal zero value, including abnormal jump to 0 value and abnormal straight line; S222. Filter out invalid three-axis acceleration data using gradient analysis.

6. The pantograph state evaluation method based on multi-point acceleration measurement according to claim 1, characterized in that: The three-axis acceleration data fluctuation matching and identification includes: Determine that the yaw acceleration direction of each detection point of the pantograph angle is consistent; It is determined that the acceleration direction along the line of each detection point of the pantograph angle is consistent.

7. The pantograph state evaluation method based on multi-point acceleration measurement according to claim 1, characterized in that: The waveform inconsistency identification includes: Determine whether the coaxial acceleration data of each detection point have nearly the same characteristics in the time domain; The three-axis acceleration data is transformed into the frequency domain to determine whether the waveform vibration period of the same slider and the same direction is consistent, thereby determining whether the acceleration change period is consistent; The Sieboldt-Huang transform is used to extract the time-frequency domain features of the triaxial acceleration data to determine whether there is only a time difference in the instantaneous energy of the coaxial data of the front and rear sliders, and whether the instantaneous energy of the previous and next moments is consistent.

8. The pantograph state evaluation method based on multi-point acceleration measurement according to claim 1, characterized in that: The missing value enhancement includes: Find the location of the abnormal point. When only a single axis and a single detection point data is abnormal, you can continue to use the data of the corresponding detection point on the same axis to perform an inverse transformation to complete the abnormal detection point data.

Images