阅读说明：本技术 用于监视铁路轨道设备的道岔的方法 (Method for monitoring a switch of a railway track system ) 是由 K.沃姆 于 2020-01-27 设计创作，主要内容包括：除了别的之外,本发明涉及一种用于确定针对铁路轨道设备的道岔(W)的分类模型(KM,KM’)的装置,该分类模型能够根据在道岔回转期间所测量的测量值来确定道岔(W)的故障。根据本发明规定,针对多个道岔回转分别确定参考回转数据组,该参考回转数据组分别涉及在相应的道岔回转期间所测量的至少两个物理测量参量,并且基于该多个参考回转数据组确定分类模型(KM,KM’)。(The invention relates to a device for determining a classification model (KM, KM') for a switch (W) of a railway track system, which allows the determination of a fault of the switch (W) from measured values measured during a switch revolution. According to the invention, reference revolution data sets are determined for a plurality of switch revolutions, each reference revolution data set relating to at least two physical measured variables measured during a respective switch revolution, and a classification model (KM, KM') is determined on the basis of the plurality of reference revolution data sets.)

1. A method for determining a classification model (KM, KM') for a switch (W) of a railway track system, which classification model is able to determine a fault of the switch (W) from measured values measured during a switch revolution, wherein reference revolution data sets are determined for a plurality of switch revolutions, each reference revolution data set relating to at least two physical measured variables measured during a respective switch revolution, and

determining the classification model (KM, KM') based on a plurality of reference revolution data sets,

it is characterized in that the preparation method is characterized in that,

-for each switch turn of the switch (W), creating an at least two-dimensional feature vector (M) associated with a predefined vector space as a reference turn data set, at least two vector components of the feature vector relating to at least two physical measurement quantities measured during the switch turn, and

-defining a spatial section within a vector space using the feature vector (M), wherein the spatial section forms a classification model (KM, KM') and can be checked for forming a fault Signal (SF) by: whether the feature vectors (M) generated for the subsequent turnout turns after the completion of the classification model (KM, KM') lie outside the space section to a greater extent than a predetermined extent.

2. The method of claim 1, wherein the first and second light sources are selected from the group consisting of,

it is characterized in that the preparation method is characterized in that,

the classification model (KM, KM') is determined taking into account or only on the basis of those reference revolution data sets whose associated switch revolutions are to be regarded as fault-free.

3. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

the classification model (KM, KM') is also determined at least on the basis of a reference revolution data set which relates to a predefined number of turnout revolutions after the initial installation of the turnout (W) or to a predefined time interval after the initial installation of the turnout (W).

4. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

the classification model (KM, KM') is also determined at least on the basis of a reference revolution data set which relates to a predefined number of turnout revolutions after the maintenance of the turnout (W) or to a predefined time interval after the maintenance of the turnout (W).

5. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

the classification model (KM, KM') is also determined at least on the basis of a reference revolution data set which relates to a predefined number of turnout revolutions after the maintenance of the turnout (W) or to a predefined time interval after the maintenance of the turnout (W).

6. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

-the first classification model (KM) is determined on the basis of a reference revolution data set which relates to a predefined number of switch revolutions after the initial installation of the switch (W) or to a predefined time interval after the initial installation of the switch (W), and

-modifying the first classification model (KM) to form a second classification model (KM') based on a reference revolution data set relating to a predetermined number of switch revolutions after a first maintenance or first repair of the switch points (W) or relating to a predetermined time interval after a first maintenance or first repair of the switch points (W).

7. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

-after each repair or maintenance, modifying the existing classification model (KM) to form an updated classification model (KM') based on a reference revolution data set relating to a predetermined number of switch revolutions after a respective maintenance or repair of the switch points (W) or to a predetermined time interval after a respective maintenance or repair of the switch points (W).

8. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

-the reference revolution data set specifies the revolution duration of the switch points (W) as at least one of the measured physical measured variables.

9. The method according to any one of the preceding claims,

it is characterized in that the preparation method is characterized in that,

the classification model (KM, KM') is determined under consideration or on the basis of a class of support vector machine methods.

10. A method for determining a fault in a switch (W) of a railway track device,

it is characterized in that the preparation method is characterized in that,

-creating a revolution data set during or after a switch turn of the switch (W), said revolution data set relating to at least two physical measurement quantities measured during the switch turn,

-the revolution data set is compared with a classification model (KM, KM') which has been determined for the same at least two measurement quantities according to the method as claimed in any of the preceding claims, and

-generating a fault Signal (SF) indicative of a fault behavior of a switch (W) in case the revolution data set lies outside a switch state range defined by the classification model (KM, KM') as allowed switch states.

11. An apparatus (200, 300, 400, 500) for determining a classification model (KM, KM ') for a switch (W) of a railway track system, which classification model is capable of determining a fault of the switch (W), wherein the apparatus (200, 300, 400, 500) is designed to determine the classification model (KM, KM') on the basis of a plurality of reference revolution data sets, which relate in each case to at least two physical measurement variables measured during a respective switch revolution, characterized in that the apparatus (200, 300, 400, 500) is designed to,

-for each switch turn of the switch (W), creating an at least two-dimensional feature vector (M) associated with a predefined vector space as a reference turn data set, at least two vector components of the feature vector relating to at least two physical measurement quantities measured during the switch turn, and

-defining a spatial section within a vector space using the feature vectors (M), wherein the spatial section forms the classification model (KM, KM') and the following checks can be made for forming a fault Signal (SF): after completion of the classification model (KM, KM'), whether the feature vectors (M) generated for the subsequent turnout turns lie outside the space section to a greater extent than specified.

12. An apparatus (200, 300, 400, 500) for determining a fault of a switch (W) of a railway track device,

it is characterized in that the preparation method is characterized in that,

the device (200, 300, 400, 500) is designed to create a rotary data record during or after the completion of a switch turn of the switch (W), which relates to at least two physical measured variables measured during the switch turn, to compare the rotary data record with a classification model (KM, KM '), which has been determined on the basis of a plurality of reference rotary data records, and to generate a fault Signal (SF) which indicates a fault behavior of the switch (W) if the rotary data record lies outside a switch state range defined by the classification model (KM, KM') as an allowed switch state.

13. The device (200, 300, 400, 500) according to claim 11 or 12,

it is characterized in that the preparation method is characterized in that,

the apparatus (200, 300, 400, 500) has a computing device (210) and a memory (220) having stored therein a Computer Program Product (CPP) which, when executed by the computing device (210), causes the computing device to perform the method according to any one of claims 1 to 10.

14. A Computer Program Product (CPP),

it is characterized in that the preparation method is characterized in that,

the Computer Program Product (CPP) is adapted to, when executed by a computing device (210), cause the computing device to perform the method according to any of claims 1 to 10.

Technical Field

The invention relates to a method and a device which enable particularly reliable monitoring of points of a railway track system or provide a basis for the same, in particular in the form of a classification model.

Background

A method for monitoring a switch of a railway track equipment is known from korean patent document KR 101823067B 1. In previously known methods, for switches that are considered to be operational or to be fault-free, the current consumption of the switch drive of the switch is detected and the corresponding reference value is stored. If, in the subsequent switch operation, it is determined that the current measured value is not correlated with the reference measured value, a corresponding fault signal is generated, which indicates a fault in the switch.

The document US 20180154913 a1 describes a computer-implemented method for informing a user of the presence of a fault in an electromechanical system in a railway track infrastructure. The method includes receiving electrical usage data (the electrical usage data specifying a value of an electrical usage parameter, the value associated with the electromechanical system) and receiving temperature data showing a current temperature of the electromechanical system. Further, it is determined whether a value of the electrical utilization parameter is indicative of a fault in the electromechanical system based on a predetermined relationship between the electrical utilization parameter and the temperature. If this is the case, a warning is issued to indicate the presence of the fault.

Disclosure of Invention

The object of the invention is, inter alia, to provide a method for determining a classification model which enables particularly reliable monitoring of a switch of a railway track system.

In order to solve this technical problem, according to the invention, a method having the features according to claim 1 is provided. Advantageous embodiments of the method are specified in the dependent claims.

Then, according to the invention, reference revolution data sets are determined for a plurality of switch revolutions, each reference revolution data set relating to at least two physical measured variables measured during a respective switch revolution, and a classification model is determined on the basis of the plurality of reference revolution data sets.

The method according to the invention has the main advantage that, unlike previously known methods, the switch monitoring is not carried out on the basis of a single physical measured variable (in this case the current), but on the basis of at least two or more measured variables, as a result of which an extended classification model is formed and a particularly reliable fault detection is possible.

It is considered advantageous to determine classification models considering or based only on those reference revolution data sets whose associated switch revolutions are considered to be fault-free.

Preferably, for each switch turn of the switch, an at least two-dimensional feature vector associated with a predefined vector space is created as a reference turn data set, at least two vector components of the feature vector relating to at least two physical measurement variables measured during the switch turn.

Preferably, a spatial section within the vector space is defined by the feature vectors, which spatial section forms a classification model and can be checked for the formation of a fault signal: after the classification model is completed, the feature vectors generated for the subsequent turnout turns are outside the spatial segment to a predetermined extent.

Advantageously, the classification model is determined at least also on the basis of reference revolution data sets which relate to a predetermined number of switch revolutions after the initial installation of the switch or to a predetermined time interval after the initial installation of the switch. That is, these reference revolution data sets created after the initial installation define, with a probability of dominance, and form a positive case for the operational turnout.

Alternatively or additionally, it can advantageously be provided that the classification model is determined at least also on the basis of reference revolution data sets which relate to a predetermined number of switch revolutions after the maintenance of the switch or to a predetermined time interval after the maintenance of the switch. That is, these reference revolution data sets created after maintenance define, with a probability of dominance, and form a positive case for operable switches.

Alternatively or additionally, it can advantageously be provided that the classification model is determined at least also on the basis of reference revolution data sets which relate to a predetermined number of switch revolutions after the maintenance of the switch or to a predetermined time interval after the maintenance of the switch. That is, these reference revolution data sets created after maintenance define, with probability of dominance, and form a positive case for the operational turnout.

Advantageously, the first classification model is determined on the basis of reference revolution data sets which relate to a predetermined number of switch revolutions after the initial installation of the switch or to a predetermined time interval after the initial installation of the switch. The first classification model can then be modified in an advantageous manner on the basis of a reference revolution data set to form a second classification model, the reference revolution data set relating to a predefined number of switch revolutions after the first maintenance or first repair of the switch or relating to a predefined time interval after the first maintenance or first repair of the switch.

It is particularly advantageous to modify the existing classification model after each repair or maintenance to form an updated classification model on the basis of a reference revolution data set which relates to a predetermined number of switch revolutions after the respective maintenance or repair of the switch or to a predetermined time interval after the respective maintenance or repair of the switch.

Preferably, the reference revolution data set specifies the revolution duration of the switch as at least one of the measured physical measured variables. The slew duration of a switch is a particularly suitable measurement variable for detecting faults.

Particularly preferably, the classification model is determined taking into account or based on a Class of Support Vector Machine methods (One-Class-Support-Vector-Machine-Verfahren).

In the formation of the second and/or updated classification model, the warning signal can advantageously be generated for reference revolution data sets which lie outside the switch state range defined by the respective preceding classification model as permissible switch states. In the presence of the warning signal, a check of the measurement and/or a check of the switch function can be carried out.

Furthermore, the invention relates to a method for determining a fault in a switch in a railway track system. In this method, according to the invention, a rotation data set is created during or after the completion of a switch turn of the switch, said rotation data set relating to at least two physical measured variables measured during the switch turn; comparing the revolution data set with a classification model, which has been determined for the same at least two measurement quantities according to the method as described above; and generating a fault signal indicative of a fault behavior of the switch if the slew data set is outside a switch state range defined by the classification model as an allowed switch state. The last-mentioned method according to the invention is therefore based on the use of a classification model based on at least two physical measurement variables and can therefore be carried out particularly reliably; in this regard, reference is made to the above embodiments in connection with the method for determining a classification model, which are correspondingly applicable here.

The invention further relates to a device for determining a classification model for a switch of a railway track system, which classification model enables a fault to be determined for the switch. In this case, the device is designed to determine a classification model on the basis of a plurality of reference revolution data sets, which relate in each case to at least two physical measured variables measured during the respective switch revolution. With regard to the advantages of the device according to the invention, reference is made to the above embodiments in connection with the method according to the invention for determining a classification model, since these embodiments apply correspondingly here.

The invention further relates to a device for determining a fault of a switch of a railway track system. In this respect, according to the invention, the device is designed to create a rotation data record during or after the completion of a switch rotation of the switch, said rotation data record relating to at least two physical measured variables measured during the switch rotation; comparing the gyration data set with a classification model, the classification model having been determined based on a plurality of reference gyration data sets; and generating a fault signal indicative of a fault behavior of the switch if the slew data set is outside a switch state range defined by the classification model as an allowed switch state. With regard to the advantages of the last-mentioned device according to the invention, reference is made to the above embodiments in conjunction with the method according to the invention for determining a fault in a switch of a railway track system, which apply correspondingly here.

Advantageously, the apparatus has a computing device and a memory in which is stored a computer program product which, when executed by the computing device, causes the computing device to perform one or all of the methods described above.

Furthermore, the invention relates to a computer program product which, when executed by a computing device, causes the computing device to perform one or all of the above described methods.

Drawings

The invention is illustrated in more detail below with reference to examples; here, by way of example:

figure 1 shows a first embodiment for the method according to the invention according to a flow chart,

figure 2 shows a second embodiment for the method according to the invention according to a flow chart,

figure 3 shows an embodiment for an apparatus for determining a classification model according to the invention according to a block diagram,

figure 4 shows a second embodiment for an apparatus for determining a classification model according to a block diagram,

figure 5 shows an embodiment of a method for monitoring switches of a railway track equipment according to the invention according to a flow chart,

fig. 6 shows a first exemplary embodiment of a device for determining a fault of a switch of a railway track system according to a block diagram, an

Fig. 7 shows a second exemplary embodiment of a device for determining a fault of a switch of a railway track system according to a block diagram.

For purposes of clarity, the same reference numbers will be used in the drawings to refer to the same or like components.

Detailed Description

Fig. 1 shows an exemplary embodiment of a method for determining a classification model KM according to a flowchart, which makes it possible to determine a fault of a switch W of a railway track system from measured values measured during a switch back.

Within the scope of method step 110, it is monitored whether a certain starting signal S is present for starting the method or for starting the classification model KM. If this is the case, a subsequent acquisition process 120 for acquiring the reference revolution data set is started.

Within the scope of the acquisition process 120, a monitoring step 121 for identifying and monitoring the respective next switch turn is first initiated. If the start of a new switch turn is detected in method step 121, then in a subsequent method step 122 at least two physical measurement variables are acquired in each case in a measurement-related manner for the respective switch turn. The physical measured variable may be, for example, the current consumption or the maximum current of the electric drive motor of the respective switch W or the switch-back time of the switch W. Alternatively or additionally, further physical measured variables can also be taken into account, for example the maximum electrical power consumption of the switch W and/or a possible phase offset between current and voltage at the drive motor of the switch W.

In a subsequent method step 123, reference revolution data sets are determined for the respective switch revolutions, which reference revolution data sets relate to at least two physical measured variables. In the following, it is assumed by way of example that a two-dimensional or multidimensional feature vector is created as the reference revolution data set, the vector components of which feature vector relate to the physical measured variables measured during the respective switch revolution.

In fig. 1, the eigenvectors formed in method step 123 are denoted by the reference symbol Mi, where the index i denotes the ith switch turn after the presence of the start signal S. Thus, the feature vector M1 will represent the first feature vector after the presence of the activation signal S, and the feature vector Mn will represent the nth feature vector after the presence of the activation signal S.

For example, if two physical measurement variables are measured, such as current consumption and switch back time, the eigenvector at the i-th switch back after the input of the start signal S will be a two-dimensional vector, which is expressed, for example, as follows:

Mi＝(I,T)

where I denotes the current during the ith turnout turn, and T denotes the turn duration during the ith turnout turn.

In a subsequent method step 124, it is checked whether a sufficient number of switch revolutions have been detected after the input of the start signal S or a predetermined minimum number of revolutions has been reached. For example, in method step 124, it may be checked whether a number n of 10 switch revolutions has already been detected. If this is the case, the measured feature vectors M1, ·, M10 are output in method step 124. If the number n of turnout revolutions has not yet been reached, the monitoring of the turnout revolutions continues in method step 121 until a predetermined number of turnout revolutions has been reached.

Instead of a predefined number of switch revolutions, it is also possible in method step 124 to check whether a predefined time interval T has elapsed after the input of the start signal S. If this is the case, the method step 130 is continued, otherwise the respective recording of the next feature vector is continued in the method step 121.

After the acquisition process 120 is completed, a classification model KM is generated in a subsequent method step 130 on the basis of the generated feature vectors M1. It is considered particularly advantageous to determine the classification model KM under consideration or on the basis of a Class of Support vector Machine methods (One-Class-Support-vector-Machine-Verfahren). In this regard, reference is made herein to the known literature wherein the generation of classification models based on a class of support vector machine methods is described in detail, for example:

"Support Vector Method for Novelty Detection", BernhardRobert Williamson, Alex Smola, John Shawe-Taylor, John Platt, Advances in Neural Information Processing Systems 12, 6.2000, 582. sub.588, Massachusetts institute of technology, and

-Estimating the Support of a High-Dimensional Distribution (Estimating the Support of a High-Dimensional Distribution), "BernhardJohn C.Platt, John C.Shawe-Taylor, Alex J.Smola, Robert C.Williamson, neuro-computing archive (Neural computing archive), Vol.13, 7.2001, p.1443-.

In summary, the classification model KM in the method according to fig. 1 is created on the basis of feature vectors or reference revolution data sets which relate to a predetermined number of switch revolutions after the presence of the start signal S or to switch revolutions which are carried out within a predetermined time interval after the occurrence of the start signal S.

If the start signal S is generated after the switch W has been reinstalled or after maintenance or repair of the switch W, it can be assumed with a prevailing probability that the feature vector M or the corresponding reference revolution data set characterizes an operable or faultless switch W and that a classification model can thus be formed which is "trained" for the recognition of faultless switch revolutions. Thus, training is only carried out in the method according to fig. 1 on the basis of positive examples relating to faultless turnout turns; counter-examples for faulty switches are not necessary for teaching or training the classification model KM.

In the embodiment according to fig. 1, the classification model KM is generated based on a class of support vector machine methods; alternatively, it is of course possible to use further methods, by means of which the classification model KM can be created only from the positive examples, i.e. only from the reference revolution data set considered "fault-free". In this connection, mention may be made, for example, of the methods described in the following literature citations:

"Review of Novelty Detection" (A Review of Novelty Detection), "Marco A.F.Pimentel, David A.Clifton, Lei Clifton, Lionel Tarassenko, Signal Processing (Signal Processing), Vol.99, 6.2014, pp.215-,

"Recent Trends Survey of One category (A Survey of Recent Trends in One Class Class Classification)," Shehroz S.Khan, Michael G.Madden, Artificial Intelligence and Cognitive sciences (Artificial Intelligence and Cognitive Science), pp 188-

"Review of novel Detection Methods", Dubravko Miljkovic, 33 rd MIPRO International conference, 5 months 2010, IEEE.

Fig. 2 shows a method for determining a classification model KM', which is created by updating or modifying an existing classification model KM on the basis of an already existing classification model KM:

after the presence of the start signal S and the subsequent acquisition of the reference revolution data set in the acquisition process 120 (see for this, for example, the embodiment in connection with fig. 1), the already existing classification model KM is modified in a modification method 131 on the basis of the newly generated feature vectors M1, …, Mn. This modification is particularly easy to implement by integrating the newly generated feature vectors M1, …, Mn into the existing classification model KM, whereby a modified or new classification model KM' is generated.

The feature vectors considered for forming the existing classification model KM may also be considered together with the newly generated feature vectors M1, …, Mn to form a modified or new classification model KM'.

In other respects, the embodiment described above in connection with fig. 1 applies correspondingly to the method according to fig. 2.

Fig. 3 shows an embodiment for an apparatus 200 for determining a classification model KM. The device 200 includes a computing device 210 and a memory 220.

In the memory 220 a computer program product CPP is stored, which contains a control program module SPM, a software module SM120 and a software module SM130 for generating the classification model KM. The software modules SM120 and SM130 are controlled by a control program module SPM.

Once the control program module SPM receives the corresponding start signal S, the software module SM120 executes the acquisition process 120 set forth above in connection with fig. 1 and 2, i.e. the method steps 121 to 124 for generating the reference revolution data set or feature vector M.

Software module SM130 (in a manner controlled by control program module SPM) uses the reference revolution data set or the corresponding feature vectors M of software module SM120 to form a classification model KM according to method step 130, as already explained above in connection with fig. 1 and 2.

Fig. 4 shows an embodiment for an apparatus 300 which is suitable not only for generating a classification model KM, but also for modifying an already existing classification model KM and generating a modified classification model KM'. For this purpose, the apparatus 300 has an additional software module SM131, which is capable of forming an updated or modified classification model KM' based on the classification model KM which has been generated previously and on the newly created feature vectors M, as already explained above in connection with the embodiment according to fig. 2 or the corresponding modification method 131.

Fig. 5 shows an embodiment of a method for determining a fault of a switch W of a railway track system according to a flow chart. Within the scope of method step 140, each switch turn of switch W is monitored and a corresponding turn data set is generated, preferably in the form of a feature vector M. In an evaluation step 150, it is checked whether the respective revolution data set characterizes a faultless switch revolution according to a predefined classification model KM. If it is determined that the revolution data set lies outside the switch state range defined by the classification model KM as an allowed switch state, a fault signal SF is generated.

The classification model KM may have been generated, for example, within the scope of the method according to fig. 1 or within the scope of the method according to fig. 2.

Fig. 6 shows an embodiment of an apparatus 400 for determining a fault of a switch W of a railway track device. Device 400 includes computing device 210 and memory 220. In the memory 220 a computer program product CPP is stored, which has a control program module SPM, a software module SM140, a software module SM150 and a classification model KM.

If the control program module SPM determines that a new switch revolution has occurred, the software module SM140 generates a revolution data set or eigenvector M which characterizes the corresponding switch revolution on the basis of at least two physical measurement variables.

Subsequently, it is checked by the software module SM150 whether the acquired revolution data set or the feature vector M lies outside the switch state range defined by the classification model KM as an additional switch state. If this is the case, a fault signal SF is generated.

The software module SM140 preferably performs the method step 140 as it has been set forth in connection with fig. 5. The software module SM150 preferably performs the analysis step 150 as it has been set forth in connection with fig. 5.

Fig. 7 shows a further exemplary embodiment of a device 500 for determining a fault of a switch W of a railway track system. In the arrangement according to fig. 7, in addition to the software modules SM140 and SM150, there are also software modules SM120, SM130 and SM131, which are adapted to generate the classification model KM and to modify or update an existing classification model KM to form an updated classification model KM'. With regard to the software modules SM120, SM130 and SM131, reference is made to the embodiments above in connection with fig. 3 and 4, which are correspondingly applicable here.

In the exemplary embodiment according to fig. 7, the device 500 can thus not only identify a fault from the revolution data set or the newly measured feature vectors and generate a fault signal SF if necessary, but can also generate a classification model KM or a modified classification model KM'.

The control program module SPM is preferably designed to trigger the formation of the classification model KM by means of the software modules SM120 and SM130, respectively, in the presence of the start signal S as long as the classification model KM has not been formed before. Preferably, it is necessary to regenerate the classification model after the first commissioning of the switch W.

If a previously generated classification model KM already exists, the control program module SPM, preferably the software module SM131, is activated upon application of the start signal S in order to update the existing classification model KM to form an updated classification model KM'. The updating of the respectively present classification model preferably takes place after each maintenance or repair, respectively.

Preferably, the formation of the first classification model and the formation of the updated classification model are each carried out on the basis of a predefined number of switch revolutions after the input of the start signal S or within a predefined time interval after the input of the start signal S. The start signal S is preferably generated after the switch W has been reinstalled and after maintenance and/or repair of the switch W and is input into the control program module SPM.

Although the invention has been illustrated and described in detail by means of preferred embodiments, the invention is not limited to the disclosed examples and other variants can be derived therefrom by the person skilled in the art without departing from the scope of protection of the invention.

List of reference numerals

110 method step

120 acquisition process

121 monitoring step

122 method step

123 method step

124 method step

130 method step

131 modification method

140 method step

150 analysis step

200 device

210 computing device

220 memory

300 device

400 device

500 device

CPP computer program product

KM classification model

KM' classification model

M1 feature vector

M feature vectors

Mi feature vector

Mn eigenvector

S Start signal

SF fault signal

SM120 software module

SM130 software module

SM131 software module

SM140 software module

SM150 software module

SPM control program module

W turnout