CN115805971A

CN115805971A - Method for monitoring polygonal fault of train wheel on line

Info

Publication number: CN115805971A
Application number: CN202211677372.9A
Authority: CN
Inventors: 曹日起; 侯永强; 刘闯; 刘传杨
Original assignee: Dalian Baishengyuan Technology Co ltd
Current assignee: Dalian Baishengyuan Technology Co ltd
Priority date: 2022-12-26
Filing date: 2022-12-26
Publication date: 2023-03-17

Abstract

The invention belongs to the technical field of rail train fault detection, and discloses a method for monitoring a train wheel polygon fault on line. According to the wheel polygon fault monitoring method, a mode that the average value of polygon indexes in a certain time is used as a polygon evaluation index and a mode that the polygon order index appearing most times in a certain time is selected as the polygon order evaluation index are introduced, so that the stability and reliability of the evaluation index are ensured, and misdiagnosis and missed diagnosis are effectively avoided; the vehicle speed is introduced as a condition for fault judgment and alarm state output, and a method for time dimension judgment and fault state value counting judgment is introduced, so that the influence of the track state on judgment of the polygonal fault state of the wheel is effectively avoided, and the interference of a large amount of fault alarm information prompts on drivers and passengers is reduced. The wheel polygon fault detection method has accurate and stable monitoring result and small requirement on system performance, and meets the requirement of a low-cost embedded device real-time online monitoring system.

Description

Method for monitoring polygonal fault of train wheel on line

Technical Field

The invention belongs to the technical field of rail train fault detection, and relates to a method for monitoring a polygonal fault of a train wheel on line.

Background

By the end of 2021 years, the national railway business mileage is 15 kilometers, the high-speed railway operation mileage reaches 4 kilometers, and the high-speed railway stably stays in the first place in the world. By 7 months in 2022, 277 urban rail transit lines are opened in 51 cities of 31 provinces (autonomous regions and direct prefectures) and Xinjiang production and construction military consortia, and the operating mileage is 0.91 kilometer.

In the train operation process, a plurality of technical indexes related to economy, comfort and safety need to pay attention, and one of the indexes is a wheel polygon fault index. The wheel polygon is also called as wheel corrugation or periodic non-rounding of the wheel, is commonly existed in the operation of subway vehicles, ordinary railways and high-speed trains, and can aggravate the vibration of the vehicles and generate noise to influence the riding comfort; and the vibration can also affect the service life of the wheel rail system, and affect the stability, the economy and the safety of train operation.

The detection modes of the polygonal state of the wheel in use at present are mainly divided into three categories: ground detection, trackside detection and real-time online monitoring. At present, the detection of the polygonal state of wheels in most train maintenance work sections in China mainly takes ground detection as a main part, the ground detection is carried out by using specific equipment, and the detection method has the advantages that the detection accuracy is high, but the detection is carried out after the wheel set is disassembled, generally only carried out during four-stage maintenance or five-stage maintenance, and the detection frequency is too low. The trackside detection generally uses an image detection technology, a high-definition camera array is installed beside a track, images of wheel sets and treads of vehicles are completely collected, fault information is automatically analyzed and identified, the trackside detection device has the advantages that the detection accuracy is high, the trackside detection device is fixed in position, a train can be detected only when passing by the trackside detection device, and the detection result is influenced by the passing speed of the train. The real-time on-line monitoring generally refers to detecting by using a vibration acceleration sensor mounted on an axle box of a commercial vehicle, obtaining a polygonal evaluation index through a detection algorithm, and performing real-time detection.

The existing real-time online monitoring method adopts short-time indexes to perform alarm judgment, the identification on the polygonal fault degree of the wheel is good, however, in the actual running process of a train, due to the fact that running track states are different, when the train runs in a track section with a poor state, polygonal evaluation index values are influenced by impact signals to be higher, misjudgment is easily caused, and when polygonal faults exist in the wheel, the short-time indexes are adopted to perform alarm judgment, the polygonal faults can be continuously reported, and a large amount of alarm information prompts are caused.

Disclosure of Invention

The invention provides a specific solution to the difficult problem. The method is realized by adopting a method for analyzing and processing the vibration acceleration signals acquired by the sensor, not only can the polygon defect of the wheel be monitored on line in real time, but also the situations of error report and the like which are easy to generate by adopting short-time index alarm judgment are effectively avoided.

In order to achieve the purpose, the technical scheme adopted by the invention for solving the problems is as follows:

acquiring real-time speed and wheel diameter information of a train; calculating the real-time frequency conversion of the wheel according to the real-time speed and the diameter of the wheel, and then calculating a polygonal order frequency value sequence of the wheel;

acquiring a collected train wheel vertical vibration acceleration signal; after the vibration acceleration signal is subjected to low-pass filtering, intercepting a middle section, and simultaneously performing frequency domain transformation;

according to the vibration acceleration signal obtained after the middle section is intercepted, firstly calculating a time domain effective value, and then calculating a vibration effective value index;

according to frequency domain data obtained by frequency domain transformation, firstly extracting a certain number of peak values and corresponding frequencies in the frequency domain data to form a peak value sequence, and extracting frequency domain amplitude values corresponding to the wheel polygon order frequency value sequence from the peak value sequence by adopting a difference threshold algorithm to form a wheel polygon amplitude-frequency characteristic sequence;

according to the amplitude-frequency characteristic sequence of the wheel polygon, taking a group of data with the maximum amplitude as a characteristic quantity, and calculating a polygon amplitude index and a polygon order index;

calculating a polygon index according to the vibration effective value index and the polygon amplitude index;

calculating a polygon evaluation index and a polygon order evaluation index according to the polygon index and the polygon order index;

when the vehicle speed is greater than the preset value, comparing the polygon evaluation index with an alarm threshold value, and judging whether a fault state counting condition is met or not by combining the polygon order index, and counting; and when the vehicle speed is less than the preset value, judging whether the sum of the times of the fault state exceeds the alarm judgment times, outputting an alarm state, and setting the relevant alarm parameter to be 0.

A method for monitoring polygonal faults of train wheels on line comprises the following specific implementation steps:

step 1: acquiring the real-time speed TSpd and the wheel diameter TDia of the train;

step 2: calculating a wheel real-time rotation frequency St = TSpd 1000/(3.6 TDia pi) according to the real-time vehicle speed TSpd and the wheel diameter TDia, and then calculating a wheel polygon order frequency value sequence On (St);

and step 3: acquiring a collected train wheel vertical vibration acceleration signal;

and 4, step 4: after the vibration acceleration signal obtained in the step (3) is subjected to low-pass filtering, intercepting a middle section, and simultaneously carrying out FFTW frequency domain transformation;

and 5: according to the vibration acceleration signal obtained after intercepting the middle section in the step 4, firstly calculating a time domain effective value RMS, and then calculating a vibration effective value index RmsdB =20log (RMS);

step 6: according to the frequency domain data obtained by the FFTW frequency domain transformation in the step 4, firstly extracting a certain number of peak values and corresponding frequencies in the frequency domain data to form a P (x, y) sequence, then sequencing the P (x, y) sequence from large to small according to frequency domain amplitude values to form a peak value sequence Py (x, y), extracting the frequency domain amplitude values corresponding to the wheel polygon order frequency value sequence from the peak value sequence by adopting a difference threshold algorithm, and forming a wheel polygon amplitude-frequency characteristic sequence OPn (x, y);

and 7: according to the amplitude-frequency characteristic sequence OPn (x, y) of the wheel polygon obtained in the step 6, taking a group of data OPmax (x, y) with the maximum frequency domain amplitude as a characterization quantity, and calculating a polygon amplitude index AdB =20log (OPmax (y)) and a polygon order index Polyn = OPmax (x)/St;

and 8: obtaining a vibration effective value index RmsdB according to the step 5, obtaining a polygon amplitude index AdB value and a weighting coefficient a in the step 7, and calculating a polygon index KndB = a AdB + (1-a) RmsdB;

and step 9: averaging the polygon index Kndb from 1 second to N seconds as the N second polygon evaluation index

KdB within the initial N-1 second takes a value of 0;

step 10: selecting a polygon order index Polyn with the largest occurrence frequency within 1-N seconds as a final polygon order evaluation index Poly;

step 11: outputting a polygon evaluation index KDB and a polygon order evaluation index Poly;

step 12: when the vehicle speed is greater than a preset value, comparing the polygon evaluation index KDB with an alarm threshold value, judging whether a fault state counting condition is met or not by combining the polygon order evaluation index Poly, and counting the fault state times of prejudgment state times ALTN1 or early warning state times ALTN2 or alarm state times ALTN 3; when the vehicle speed is less than the preset value, judging whether the sum of the failure state times exceeds the alarm judgment times, namely whether ALTN1+ ALTN2+ ALTN3 is more than or equal to ALTN, if so, the alarm state ALTRs is a state value ALST corresponding to the maximum value in ALTN1, ALTN2 and ALTN3, otherwise, the alarm state ALTRs =0;

step 13: outputting an alarm state ALttr; and sets the ALTN1, ALTN2, ALTN3, ALTr values to 0.

Preferably, in step 2, the wheel polygon order frequency value sequence On (St) is the product of the wheel rotation frequency St and the polygon order On sequence, and the order On of the wheel with the polygon fault is generally between 15 and 40;

preferably, in step 4, intercepting the middle section is intercepting the vibration acceleration signal between 10% and 90% of the vibration acceleration signal.

Preferably, in step 6, the method for extracting the peak sequence Py (x, y) includes: a least square fitting algorithm is adopted to extract a certain number of peak values and corresponding frequencies in frequency domain data to form a P (x, y) sequence, then the P (x, y) sequence is sequenced from large to small according to frequency domain amplitude values to form a peak value sequence Py (x, y), the sequence is a group of coordinate values, an abscissa x represents the frequency, and an ordinate y represents the corresponding frequency domain amplitude value. Further, in the P (x, y) sequence, 40-60 peaks and corresponding frequencies are extracted.

As a preferable mode, in step 6, the method for extracting the frequency domain amplitude corresponding to the wheel polygon order frequency value sequence from the peak value sequence by using a difference threshold algorithm to form a wheel polygon amplitude-frequency feature sequence OPn (x, y) includes: setting a coupling coefficient, coupling the frequency values in the wheel polygon order frequency value sequence On (St) and the peak value sequence Py (x, y) one by one, when | On (St) -Py (x) | is less than or equal to the coupling coefficient, indicating that the frequency value is fault characteristic frequency, extracting the frequency value and outputting the frequency value and corresponding frequency domain amplitude to form a wheel polygon amplitude-frequency characteristic sequence OPn (x, y), wherein the sequence OPn (x, y) is a group of coordinate values, an abscissa x represents the wheel polygon order frequency value, and an ordinate y represents the corresponding frequency domain amplitude.

Preferably, in step 8, the weighting coefficient a ranges from 0.2 to 0.8, and generally ranges from a =0.5.

Preferably, in step 9, due to the short-time stress fluctuation of the wheels and the dynamic change characteristics of the vibration signals during the actual running process of the train, the calculated polygon index may generate numerical fluctuation to cause misjudgment, and in order to calculate the polygon evaluation index more stably and reliably, the invention introduces a mode of taking the polygon index average value in a certain time as the polygon evaluation index. Preferably, the polygon index within 60s is averaged to be used as the polygon evaluation index.

As a preferable mode, in step 10, due to the short-time stress fluctuation of the wheels and the dynamic change characteristic of the vibration signal during the actual operation of the train, the calculated polygon order index may generate numerical value fluctuation to cause missed judgment, and in order to calculate the polygon order evaluation index more stably and reliably, the invention introduces a mode of selecting the polygon order index which appears most times within a certain time as the polygon order evaluation index. Preferably, the polygon order index appearing most times within 60s is selected as the polygon order evaluation index.

Preferably, in step 12, the method for determining whether to alarm includes: and when the vehicle speed is greater than the preset value, comparing the polygon evaluation index KDB with an alarm threshold value. When KDB is less than the alarm threshold value, the operation is normal, and the state value ALtst =0; every time when the lowest value of the continuous n-second alarm threshold value is less than or equal to Kdb and less than the middle value of the alarm threshold value and the order Poly is the same, the state value ALST =1, and the number of prejudgment state times ALTN1 is added with 1; when the intermediate value of the continuous n-second alarm threshold value is less than or equal to Kdb and less than the maximum value of the alarm threshold value and the order Poly is the same, the state value ALST =2, and the early warning state number of times value ALTN2 is added with 1; whenever KDB is greater than or equal to the maximum alarm threshold for n consecutive seconds and the order Poly is the same, the status value ALST =3 and the number of alarm states ALTN3 plus 1. When the vehicle speed is less than the preset value, judging whether the sum of the failure state times exceeds the alarm judgment times, namely whether ALTN1+ ALTN2+ ALTN3 is more than or equal to ALTN, if so, the alarm state ALTRs is a state value ALST corresponding to the maximum value in ALTN1, ALTN2 and ALTN3, and carrying out alarm prompt; otherwise, the alarm state ALttr =0, and no alarm prompt is performed. If the short-time index is adopted for alarm judgment, in the actual running process of the train, due to the difference of the running track states, when the train runs in a track section with a poor state, the evaluation index is higher, and false alarm is caused; and when the wheel has polygon fault, the polygon fault can be continuously reported, so that a great deal of alarm information is prompted. Therefore, the invention introduces the vehicle speed as the condition for fault judgment and alarm state output, and introduces the method for time dimension judgment and fault state value counting judgment. When the vehicle speed is greater than the preset value, judging whether the polygon evaluation index continuously exceeds the alarm threshold value within a certain time and whether the order evaluation indexes are the same, and taking the polygon evaluation index as a condition for counting the fault state; when the vehicle speed is less than the preset value, if the sum of the failure state times exceeds a threshold value, a wheel polygon failure is diagnosed, and an alarm prompt is carried out; further, the preset values of ALTN1, ALTN2 and ALTN3 are 0; the ALST =1 is prejudged, the ALST =2 is early-warning, and the ALST =3 is alarming; further, the value range of n is 1-60 seconds; furthermore, the value range of the ALTN is 50-300.

The invention has the beneficial effects that: according to the wheel polygon fault monitoring method, a mode that the average value of polygon indexes in a certain time is used as a polygon evaluation index and a mode that the polygon order index appearing most times in a certain time is selected as the polygon order evaluation index are introduced, so that the stability and reliability of the evaluation index are ensured, and misdiagnosis and missed diagnosis are effectively avoided; the vehicle speed is introduced as a condition for fault judgment and alarm state output, and a method for time dimension judgment and fault state value counting judgment is introduced, so that the influence of the track state on the judgment of the polygonal fault state of the wheel is effectively avoided, and the interference of a large amount of fault alarm information prompts on drivers and passengers is reduced. The wheel polygon fault detection method has accurate and stable monitoring result and small requirement on system performance, and meets the requirement of a low-cost embedded device real-time online monitoring system.

Drawings

FIG. 1 is a schematic diagram of a system hardware structure of an embodiment of the method of the present invention.

Fig. 2 is a flow chart of the monitoring method of the present invention.

FIG. 3 is a plot of the measured point time domain waveform.

FIG. 4 is a plot of the frequency domain waveforms of the survey points.

FIG. 5 is a trend chart of the polygon evaluation index in the conventional method.

FIG. 6 is a diagram of the trend of the polygon evaluation index in the present method.

FIG. 7 is a diagram illustrating a trend of a polygon order evaluation index in the conventional method.

FIG. 8 is a graph of the trend of the polygon order evaluation index in the present method.

Fig. 9 shows the polygon evaluation index and the polygon order evaluation index trend of a certain wheel all day.

FIG. 10 is a trend of polygon fault status values throughout the day for a certain wheel.

FIG. 11 is a polygon fault alarm state value trend throughout the day for a certain wheel.

Detailed Description

The following detailed description of the invention refers to the accompanying drawings.

Example 1

As shown in the system hardware structure diagram of fig. 1, the vibration sensor is arranged above the axle box bearing of the train, and a vibration acceleration sensor is arranged on each of 8 wheels of each carriage. The data acquisition module acquires a vertical vibration acceleration signal of the wheel, which is acquired by the sensor, and sends the vertical vibration acceleration signal to the diagnostic analysis algorithm processing module; the communication module acquires real-time train speed and wheel diameter information of the train sent by the vehicle-mounted central control unit through the Ethernet, and sends the information to the diagnostic analysis algorithm processing module for calculating the real-time frequency conversion of the wheels; the wheel polygon evaluation index, the wheel polygon order evaluation index and the alarm information calculated by the diagnosis analysis algorithm processing module are sent to the vehicle-mounted central control unit from the communication module through the Ethernet in real time.

As shown in the flow chart of the monitoring method in fig. 2, the specific steps are as follows:

step 2: calculating a wheel real-time rotation frequency St = TSpd 1000/(3.6 TDia pi) according to the real-time vehicle speed TSpd and the wheel diameter TDia, and then calculating a wheel polygon order frequency value sequence On (St); the wheel polygon order frequency value sequence On (St) is the product of the wheel rotation frequency St and the polygon order On sequence, and the order On of the wheel polygon fault is generally between 15 and 40;

and 3, step 3: acquiring a collected train wheel vertical vibration acceleration signal;

and 4, step 4: the method comprises the following steps of performing 2500Hz low-pass filtering on a vibration acceleration signal, selecting a 6-order Butterworth filter as wheel polygon judgment data by the filter, wherein the transfer function form of the filter is as follows:

fourier transform is carried out on the filtered signals to obtain frequency domain data, a Fourier transform formula selects an FFTW formula to calculate, and the calculation formula is as follows:

wherein the content of the first and second substances,

according to the property of the correlation function and by adopting recursion division, all coefficients, namely frequency domain amplitude values, can be rapidly calculated.

And 5: calculating a vibration time domain effective value RMS of the filtered signal, intercepting part of the signal to calculate the time domain effective value due to calculation errors caused by edge jitter of the filtered signal, generally taking a value between 10% and 90% of a data segment when intercepting the signal, and then calculating a vibration effective value index RmsdB =20log (RMS);

step 6: a certain number of peak values and corresponding frequencies in frequency domain data are extracted to form a P (x, y) sequence, the extraction algorithm is a least square fitting algorithm, generally 40-60 peak values and corresponding frequencies are extracted and extracted, then the P (x, y) sequence is sequenced from large to small according to frequency domain amplitude values to form a peak value sequence Py (x, y), the sequence is a set of coordinate values, the x axis of an abscissa represents the frequency, and the y axis of an ordinate represents the corresponding frequency domain amplitude values. In the step, the peak value sequences of a plurality of points are extracted to ensure that the extracted peak value sequences are possible fault parts, and then the least square fitting algorithm is adopted to extract data, so that the method is stable and reliable, short in time consumption and high in speed;

and (3) performing fault characteristic coupling by adopting a difference threshold algorithm, coupling the frequency value sequence On (St) of the wheel polygon order frequency value sequence with the frequency value in the peak value sequence Px (x, y), and coupling coefficient is 2.1. The specific calculation mode is that the wheel polygon order frequency value sequence On (St) is coupled with the frequency values in the peak value sequence Py (x, y) one by one, when | On (St) -Py (x) | is less than or equal to 2.1, the frequency value is indicated to be fault characteristic frequency, the frequency value is extracted and output with corresponding frequency domain amplitude value to form a wheel polygon amplitude-frequency characteristic sequence OPn (x, y), the sequence OPn (x, y) is a group of coordinate values, the abscissa x represents the wheel polygon order frequency value, and the ordinate y represents the corresponding frequency domain amplitude value. In the step, a difference threshold algorithm is adopted for fault characteristic coupling, so that floating point calculation or frequency errors caused by rotation speed fluctuation can be abandoned, and failure in characteristic extraction is avoided;

and 7: taking a group of data OPmax (x, y) with the maximum amplitude in a frequency domain in a wheel polygon amplitude-frequency characteristic sequence OPn (x, y) as a characterization quantity, wherein the characterization quantity is a group of coordinates, the OPmax (x) represents a wheel polygon order frequency value, the OPmax (y) represents a corresponding frequency domain amplitude, and a polygon amplitude index AdB =20log (OPmax (y)) and a polygon order index Polyn = OPmax (x)/St are calculated;

and step 8: calculating a polygon index KndB = a AdB + (1-a) RmsdB + Kfac, wherein the value range of the weighting coefficient a is 0.2-0.8, the preferred value is a =0.5, the value range of the Kfac is 0-50, and the polygon index is used for adapting to the statistical type of alarm values of various vehicle types; a weight algorithm is introduced into the polygonal indexes, and the vibration effective value is used as a component of an index parameter, so that the monitoring result can be ensured to be more accurate;

KdB within the initial N-1 second takes a value of 0; the value range of N is 20-100 seconds, and the preferable value is N =60; the method introduces the polygon index average value in a certain time as the polygon evaluation index, can effectively avoid the misjudgment caused by the polygon index value fluctuation caused by the short-time stress fluctuation of wheels, the dynamic change characteristic of vibration signals and the like in the actual running process of the train, and can ensure that the polygon evaluation index is more stable and reliable;

step 10: selecting the order with the most polygon orders Polyn within 1-N seconds as a final polygon order evaluation index Poly; the value range of N is 20-100 seconds, and the preferable value is N =60; the method introduces a mode of selecting the polygon order index which appears most times within a certain time as the polygon order evaluation index, can effectively avoid the missing judgment caused by the fluctuation of the polygon order index due to the short-time stress fluctuation of wheels, the dynamic change characteristic of vibration signals and the like in the actual running process of the train, and can ensure that the polygon order evaluation index is more stable and reliable;

step 12: the preset values of ALTN1, ALTN2 and ALTN3 are 0; when the vehicle speed TSpd is more than or equal to 100km/h, the polygon evaluation index KDB is compared with an alarm threshold value. When KdB < 10 (dB), the operation is normal, and the state value ALtst =0; every time the polygon evaluation index Kdb is continuously n seconds, 10 (dB) is more than or equal to Kdb and less than 18 (dB), and when the grades Poly are the same, the state value ALtst =1, and the number of prejudgment states ALtn1 is added with 1; every time the polygon evaluation index Kdb is continuously n seconds, 18 (dB) is more than or equal to Kdb and less than 23 (dB), and when the grades Poly are the same, the state value ALST =2, and the number of times ALTN2 of the early warning state is added with 1; every time the polygon evaluation index Kdb is continuously n seconds, kdb is not less than 23 (dB), and the order Poly is the same, the state value ALST =3, and the alarm state number of times ALTN3 is added with 1; when the vehicle speed TSpd is less than or equal to 100km/h, finishing counting of ALTN1, ALTN2 and ALTN 3; judging whether the sum of the failure state times exceeds the alarm judgment times, namely whether ALTN1+ ALTN2+ ALTN3 is larger than or equal to ALTN, and the alarm state ALTRs are state values ALST corresponding to the maximum values in ALTN1, ALTN2 and ALTN 3; otherwise, the alarm state ALTtr =0; the method is characterized in that the vehicle speed is introduced as a condition for fault judgment and alarm state output, and a method for time dimension judgment and fault state value counting judgment is introduced, namely when the vehicle speed is greater than a preset value, whether polygon evaluation indexes continuously exceed a threshold value within a certain time and whether order evaluation indexes are the same are judged to be used as a condition for fault state counting; when the vehicle speed is less than the preset value, if the sum of the failure state times exceeds a threshold value, a wheel polygon failure is diagnosed, and an alarm prompt is carried out; the condition that the alarm judgment is carried out by adopting short-time indexes, when the train runs in a track section with a poor state, the evaluation index is higher to cause false alarm is avoided, and the condition that when the polygon fault exists in the wheel, the polygon fault is continuously reported to cause a large amount of alarm information prompts is avoided; the value range of n is 1-60, and the preferable value is n =20; the value of ALTN ranges from 50 to 300, and the preferred value is ALTN =100.

Example 2

The method is applied to an intelligent dynamic high-speed rail safety monitoring system in China to carry out polygonal fault diagnosis, and specifically comprises the following steps:

step 1: acquiring a real-time vehicle speed TSpd =344.5Km/h and a wheel diameter TDia =920mm of a train;

step 2: calculating the current wheel rotation frequency St = TSpd 1000/(3.6 TDia pi) =33.12Hz according to the real-time vehicle speed and the wheel diameter, and calculating a sequence of corresponding wheel polygon order frequency values On (St) =496.8Hz, 529.92Hz.. 662.4Hz.. 1324.8 Hz;

and 4, step 4: as shown in fig. 3 to 4, 2500Hz low-pass filtering is performed on the vibration acceleration signal, and a 6 th order butterworth filter is selected as the filter;

fourier transformation is carried out on the filtered signals to obtain frequency domain data, and a Fourier transformation formula selects an FFTW formula to calculate;

and 5: calculating a time domain effective value RMS of a filtered signal, intercepting a part of the signal to calculate a vibration time domain effective value because of calculation errors caused by edge jitter of the filtered signal, generally taking a value between 10% and 90% of a data segment when intercepting the signal, and then calculating a vibration effective value index RmsdB =20log (RMS) =20log (8.9805 g) =19.066dB;

step 6: a certain number of peak values and corresponding frequencies in the frequency domain data are extracted to form a P (x, y) sequence, the extraction algorithm is a least square fitting algorithm, generally 40-60 peak values and corresponding frequencies are extracted and extracted, and then the P (x, y) sequence is sequenced from large to small according to the amplitude of the frequency domain to form a peak value sequence Py (x, y). As in table 1, the top 15 peak sequence is illustrated;

TABLE 1 Peak sequence ordering

Serial number	Frequency (Hz)	Amplitude (g)
			1	661.932	10.3148
2	694.946	1.8826
			3	1323.855	0.9097
4	628.838	0.6765
			5	52.013	0.5837
6	641.694	0.5707
			7	658.646	0.5704
8	645.331	0.5659
			9	646.87	0.5536
10	665.046	0.4974
			11	640.042	0.4966
12	669.579	0.4741
			13	992.817	0.4549
14	675.111	0.4404
			15	618.054	0.4013

And (3) performing fault characteristic coupling by adopting a difference threshold algorithm, coupling the frequency value sequence On (St) of the wheel polygon order frequency value sequence with the frequency value in the peak value sequence Px (x, y), and coupling coefficient is 2.1. The specific calculation mode is that the wheel polygon order frequency value sequence On (St) is coupled with the frequency values in the peak value sequence Py (x, y) one by one, when | On (St) -Py (x) | is less than or equal to 2.1, the frequency value is indicated as fault characteristic frequency, the frequency value is extracted and output with the corresponding frequency domain amplitude value, and the wheel polygon amplitude-frequency characteristic sequence OPn (x, y) is formed. As table 2, the partial extraction results are illustrated;

TABLE 2 order index extraction results

Serial number	Frequency (Hz)	Amplitude (g)	Polygonal order sequence frequency (Hz)
				1	661.932	10.3148	662.4
2	694.946	1.8826	695.52
				3	1323.855	0.9097	1324.8
4	628.838	0.6765	629.28
				5	992.817	0.4549	993.6

And 7: taking a group of data OPmax (x, y) with the largest frequency domain amplitude in the wheel polygon amplitude-frequency characteristic sequence OPn (x, y) as a characteristic quantity, calculating a polygon amplitude index AdB =20log (OPmax (y)) =20log (10.3148 g) =20.269dB and a polygon order index Polyn = OPmax (x)/St =661.932/33.12=19.986, and rounding to 20 orders;

and 8: calculating polygon index Kndb = a + AdB + (1-a) × RmsdB =0.5 × 20.269dB +0.5 × 19.066dB =19.668dB;

high-speed rail dynamic safety monitoring systems typically set a uniform alarm threshold for regulatory administration. The evaluation values obtained based on different algorithms are different, but the preset alarm threshold value cannot be correspondingly adjusted according to different algorithms, and for this reason, an algorithm provider can set a corresponding correction coefficient according to the alarm threshold value preset by the system so as to match the use requirements of the system. In this embodiment, the correction coefficient is-4 dB, and the polygon index after correction is:

KndB＝19.668dB-4dB＝15.668dB；

and step 9: as shown in fig. 5, in the case of no averaging, the polygon index value fluctuates greatly due to the short-time stress fluctuation of the wheel, the dynamic change characteristic of the vibration signal, and the like, and a part of the time period exceeds the early warning threshold of 18dB, which causes misjudgment. As shown in fig. 6, after the averaging processing is performed for 60 seconds, the polygon evaluation index is more stable and reliable;

step 10: as shown in fig. 7, before the processing is not performed, the polygon order index data fluctuates due to the short-time stress fluctuation of the wheel, the dynamic change characteristic of the vibration signal, and the like, and the judgment condition that "the polygon evaluation index KdB continuously exceeds the threshold value for n seconds and the order Poly is the same" is disturbed, thereby causing a failure and a missing judgment. As shown in fig. 8, after the polygon order index appearing most frequently in 60 seconds is selected as the polygon order evaluation index, the polygon order evaluation index is more stable and reliable;

step 11: outputting a polygon evaluation index KDB =14.233dB and a polygon order evaluation index Poly =20;

step 12: as shown in FIG. 9 and FIG. 10, when the vehicle speed TSpd is greater than or equal to 100km/h, the polygon evaluation index Kdb is compared with the alarm threshold value: when KdB < 10 (dB), the operation is normal, and the state value ALtst =0; every time the polygon evaluation index Kdb is continuously 20 seconds (10 dB) ≦ Kdb < 18 dB), and the order Poly is the same, the state value ALtst =1, and the number of prejudgment states ALTN1 plus 1;

as shown in FIG. 11, when the vehicle speed TSpd is less than or equal to 100km/h, the ALTN1, ALTN2, ALTN3 are counted; when the sum of the fault state times ALTN1+ ALTN2+ ALTN3 is larger than or equal to the alarm judgment time ALTN =100, the alarm state ALTRs are state values ALST =1 corresponding to the maximum value ALTN1 in ALTN1, ALTN2 and ALTN 3;

step 13: outputting an alarm state ALTtr =1; and sets the ALTN1, ALTN2, ALTN3, ALTr values to 0.

Outputting 6 times of pre-judgment alarm in total all day;

the intelligent high-speed rail dynamic safety monitoring system adopts the algorithm to operate for 2 months, the alarm is prompted for 14 wheel positions for 136 times, the roundness of the wheel at the alarm position is manually measured, the defect of a polygon which is the same as the monitoring result is really present, and the accuracy rate is 100%. The method has accurate monitoring result and extremely high monitoring stability.

Claims

1. A method for on-line monitoring polygon fault of train wheel is characterized in that,

acquiring a collected train wheel vertical vibration acceleration signal; after low-pass filtering is carried out on the vibration acceleration signals, the middle section is intercepted, and meanwhile, frequency domain transformation is carried out;

according to frequency domain data obtained by frequency domain transformation, firstly extracting peak values and corresponding frequencies in the frequency domain data to form a peak value sequence, and extracting frequency domain amplitude values corresponding to the wheel polygon order frequency value sequence from the peak value sequence by adopting a difference threshold algorithm to form a wheel polygon amplitude-frequency characteristic sequence;

when the vehicle speed is greater than the preset value, comparing the polygon evaluation index with an alarm threshold value, and judging whether a fault state counting condition is met or not by combining the polygon order evaluation index, and counting; and when the vehicle speed is less than the preset value, judging whether the sum of the times of the fault state exceeds the alarm judgment times, outputting an alarm state, and setting the relevant alarm parameter to be 0.

2. The method for monitoring the polygonal fault of the train wheel on line according to claim 1, which is characterized by comprising the following steps:

and 4, step 4: performing low-pass filtering on the vertical vibration acceleration signals of the train wheels obtained in the step (3), intercepting a middle section, and performing FFTW frequency domain transformation at the same time;

and 8: obtaining a vibration effective value index RmsdB according to the step 5, obtaining a polygon amplitude index AdB value and a weighting coefficient a according to the step 7, and calculating a polygon index KndB = a AdB + (1-a) RmsdB;

KdB within the initial N-1 second takes a value of 0;

3. The method of claim 2, wherein the sequence of wheel polygon order frequency values On (St) is a product of the wheel rotation frequency St and a sequence of polygon order On, and the order On of the wheel polygon fault is between 15 and 40.

4. The method of claim 2, wherein the intercepting the mid-section is intercepting the vibration acceleration signal between 10% and 90% of the vibration acceleration signal.

5. The method for on-line monitoring of the polygon fault of the train wheel according to claim 2, wherein the method for extracting the peak sequence Py (x, y) comprises: and extracting 40-60 maximum peak values and corresponding frequencies in the frequency domain data by adopting a least square fitting algorithm to form a P (x, y) sequence, and then sequencing the P (x, y) sequence from large to small according to the frequency domain amplitude to form a peak value sequence Py (x, y), wherein the sequence is a group of coordinate values, the abscissa x represents the frequency, and the ordinate y represents the corresponding frequency domain amplitude.

6. The method for on-line monitoring of train wheel polygon faults as claimed in claim 2, wherein the method for extracting frequency domain amplitude values corresponding to the wheel polygon order frequency value sequence from the peak value sequence by using the difference threshold algorithm to form the wheel polygon amplitude frequency characteristic sequence OPn (x, y) comprises: setting a coupling coefficient, coupling the frequency values in the wheel polygon order frequency value sequence On (St) and the peak value sequence Py (x, y) one by one, when | On (St) -Py (x) | is less than or equal to the coupling coefficient, indicating that the frequency value is fault characteristic frequency, extracting the frequency value and outputting the frequency value and corresponding frequency domain amplitude to form a wheel polygon amplitude-frequency characteristic sequence OPn (x, y), wherein the sequence OPn (x, y) is a group of coordinate values, an abscissa x represents the wheel polygon order frequency value, and an ordinate y represents the corresponding frequency domain amplitude.

7. The method for on-line monitoring of the polygonal fault of the train wheel according to claim 2, further characterized in that the value range of the weighting coefficient a is 0.2-0.8.

8. The method for monitoring the polygonal fault of the train wheel on line according to claim 2 is characterized in that the method for judging whether to alarm is as follows: when the vehicle speed is greater than the preset value, comparing the polygon evaluation index KDB with an alarm threshold value; when KDB is less than the alarm threshold value, the operation is normal, and the state value ALtst =0; every time when the lowest value of the continuous n-second alarm threshold value is less than or equal to Kdb and less than the middle value of the alarm threshold value and the order Poly is the same, the state value ALST =1, and the number of prejudgment state times ALTN1 is added with 1; every time when the intermediate value of the continuous n-second alarm threshold value is less than or equal to KdB and less than the maximum value of the alarm threshold value, and when the order Poly is the same, the state value ALST =2, and the number of times of early warning ALTN2 is added with 1; every time continuous n seconds KDB is larger than or equal to the highest alarm threshold value and the order Poly is the same, the state value ALST =3, and the number of times ALTN3 of alarm states is added with 1; when the vehicle speed is less than the preset value, judging whether the failure state times exceed the alarm judgment times, namely whether ALTN1+ ALTN2+ ALTN3 is more than or equal to ALTN, if so, taking the state value ALST corresponding to the maximum value in the alarm state ALTN1, ALTN2 and ALTN3, and carrying out alarm prompt; otherwise, the alarm state ALttr =0, and no alarm prompt is performed; further, ALTst =1 is prejudged, ALTst =2 is early-warning, and ALTst =3 is alarming.

9. The method for on-line monitoring of the polygonal fault of the train wheel according to claim 8, wherein n has a value ranging from 1 to 60 in n consecutive seconds.

10. The method for on-line monitoring of the polygonal fault of the train wheel according to claim 8, wherein the value of the number of times of alarm determination ALTN ranges from 50 to 300.