CN109186813B - Temperature sensor self-checking device and method - Google Patents

Abstract

The invention provides a temperature sensor self-checking device and a method, wherein a temperature data sequence of a train under a normal condition is subjected to differential processing to obtain a segmented standard deviation sequence, and an abnormal detection threshold value is obtained by performing statistical analysis on the standard deviation sequence; carrying out differential processing on the real-time input temperature data sequence to obtain a segmented standard deviation sequence; judging whether the segmented temperature difference value sequence is abnormal or not based on the threshold value and the standard difference sequence; if the standard deviation sequence of a certain segment of the segmentation is larger than or equal to the threshold value, judging that the temperature difference sequence of the segment is abnormal, and entering the next step, otherwise, judging that the sensor is normal; judging the distribution consistency of the temperature difference sequence with the abnormality in a certain section, the normal reference sequence and the temperature difference sequence in the previous adjacent time section; if the consistency exists, the sensor is judged to be normal, and if not, the sensor is abnormal. The invention can solve the technical problems that the prior art can not carry out quick and effective self-checking on the temperature sensor and can not ensure the safe and efficient operation of the train.

Description

A temperature sensor self-checking device and method

technical field

The invention relates to the technical field of fault diagnosis, in particular to a self-checking device and method applied to a train temperature sensor.

Background technique

The temperature sensor is one of the important components of the train transmission, control, running department and other systems. It is responsible for the monitoring and sensing functions of the key components related to the temperature of the whole train. It is the core device to ensure the safety and normal operation of the equipment. One of the key indicators to monitor. Generally, in most cases, those skilled in the art are concerned with the operation of the object monitored by the sensor, and seldom consider the problem of the internal connection between the monitor and the temperature sensor (system) itself. In fact, as a device, the temperature sensor must have the possibility of failure. When the temperature value measured by the sensor is abnormal, it is usually impossible to determine whether the abnormality is actually a problem with the measured object or whether the sensor or communication system is abnormal. If the real measurement object is abnormal, emergency train safety measures, such as power reduction or shutdown for maintenance, need to be taken. However, if the sensor (system) is abnormal, taking emergency measures will greatly increase the operation and maintenance cost, and may even lead to safety problems in some cases. Therefore, a problem must be solved in practical application, that is, how to ensure the normal operation of the temperature sensor itself, or in other words, how to determine whether the abnormal temperature value that occurs is caused by the abnormality of the sensor itself.

Based on the above problems, the technical key that needs to be solved at present is how to find out whether the temperature sensor itself is normal, including: how to use the information and data related to the temperature sensor to effectively identify the source of abnormality, achieve effective early warning, and How to realize anomaly discrimination by adding additional sensors or relying on existing data.

At present, there are many researches and applications on checking the operating state of the sensor itself, among which there are many detection methods for the abnormality of the temperature sensor. Some judge the working state of the sensor by detecting the voltage and current of the sensor, and some combine professional knowledge to set the temperature threshold. To judge the working state of the sensor, there are also ways to judge the working state of the sensor by adding a sensor in a similar position and comparing the changes of two or more temperature values. However, these existing temperature sensor detection methods all have inaccurate judgment results, misjudgments, missed judgments, etc., or require professionals to use professional knowledge to make judgments, and their application occasions and fields are limited, or they require The installation of additional devices leads to various technical defects such as increased system complexity.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present invention is to provide a temperature sensor self-checking device and method to solve the technical problem that the existing train system cannot perform fast and effective self-checking on the temperature sensor, and thus cannot guarantee the safe and efficient operation of the train.

In order to achieve the above purpose, the present invention specifically provides a technical implementation scheme of a temperature sensor self-checking device. The temperature sensor self-checking device includes:

The abnormal detection threshold calculation module is used to perform differential processing on the temperature data sequence T measured by the sensor under the normal operation of the train to obtain the segmented standard deviation sequence Θ of the temperature difference sequence δT, and perform statistical analysis on the standard deviation sequence Θ. Get the anomaly detection threshold K;

The key feature value extraction module is used to perform differential processing on the real-time input temperature data sequence t measured by the sensor to obtain the segmented standard deviation sequence η of the temperature difference sequence δt;

The first abnormality detection module is used to judge whether there is abnormality in the segmented temperature difference sequence δt according to the abnormality detection threshold K output by the abnormality detection threshold calculation module and the standard deviation sequence n output by the key feature value extraction module; if If the segmented standard deviation sequence η of a certain segment of temperature difference sequence δt is greater than or equal to the abnormality detection threshold K, it is judged that this segment of temperature difference sequence δt is abnormal, and the abnormal temperature difference sequence δt is output, otherwise the sensor is judged normal;

a consistency checking module, configured to perform a distribution consistency check on the abnormal temperature difference sequence δt output by the first anomaly detection module, the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period;

The second abnormality detection module is used for judging whether the distribution consistency test occurrence probability P value output by the consistency test module is less than the set standard, if it is less than the set standard, output a sensor abnormality warning signal, otherwise the sensor is normal.

的原则构建列车该部位对应的异常检测阈值K。Further, the abnormality detection threshold calculation module obtains the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculates the temperature difference sequence δT per unit time ΔT. The temperature difference sequence δT in unit time is segmented according to the same duration T1, the standard deviation θ _i of each temperature difference sequence δT is calculated, and the standard deviation sequence Θ is formed. Analyze the distribution of the standard deviation series Θ, and calculate the mean μ and standard deviation σ of the standard deviation series Θ, according to the probability of occurrence

The principle of constructing the anomaly detection threshold K corresponding to this part of the train.

Among them, θ is the standard deviation of the temperature difference series δT.

x_i为样本值，n为样本数。Among them, ω _i is the weighting coefficient, here

x _i is the sample value, and n is the number of samples.

Further, the key feature value extraction module obtains the temperature data sequence t measured by the real-time input sensor, calculates the temperature difference value sequence δt per unit time Δt, and divides the temperature difference value sequence δt per unit time according to the same duration T2, Calculate the standard deviation of each temperature difference series δt, and form the standard deviation series η.

Further, after the consistency checking module finds that a certain section of temperature difference sequence δt appears abnormal data by comparing the abnormality detection threshold K and the standard deviation sequence η during the operation of the train, it records the information of this section of temperature difference sequence _xt . , and obtain the temperature difference sequence x _t of this segment and the temperature difference sequence y _t of the previous adjacent time period, and simultaneously obtain the temperature difference sequence z1 _t ,…,zn measured by other similar position sensors of the train in the same time period _t , and the temperature difference series x _t suspected to be abnormal and the temperature difference series y _t of the previous adjacent time period, and the temperature difference series measured by other similar position sensors {z1 _t ,...,zn _t } The KS distribution test was performed one by one.

Further, the consistency checking module is judging the temperature difference sequence x t to be checked and the temperature difference sequence y _t of the previous adjacent time period, and the temperature difference sequence {z1 _t , measured by other similar position sensors _. When the distribution consistency of ..., zn _t }, the significance of the temperature difference series x _t is determined by checking the maximum difference value D of the empirical distribution function between the series. When the actual calculated maximum difference value D is greater than a certain set standard value, or when the distribution probability P value corresponding to the maximum difference value D is less than a certain set standard value, there is no consistency between the two temperature difference series. sex.

Among them, the sample size of the temperature difference sequence x _t is n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t ,…,zn _t is n2, F ₁ (x) and F ₂ (x ) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample.

代表D_j绝对距离的最大值。检验统计量Z近似于正态分布，其表达式为：Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j . The test statistic Z is approximately normally distributed and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

为取B(F(x))绝对距离的最大值，x为样本。

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample.

The empirical distribution function B(t) is:

Among them, x is an independent variable, and i is a natural number.

The present invention also specifically provides a technical implementation scheme of a temperature sensor self-checking method, the temperature sensor self-checking method, comprising the following steps:

S10) carry out differential processing to the temperature data sequence T measured by the sensor under normal operation of the train to obtain the segmented standard deviation sequence Θ of the temperature difference sequence δT, and obtain the abnormal detection threshold K by performing statistical analysis on the standard deviation sequence Θ;

S20) perform the same differential processing as in step S10) on the real-time input sensor temperature data sequence t to obtain the segmented standard deviation sequence η of the temperature difference sequence δt;

S30) Judge whether there is an abnormality in the segmented temperature difference sequence δt based on the abnormality detection threshold K obtained in step S10) and the standard deviation sequence η obtained in step S20); if the segmented standard deviation sequence η of a certain segment of temperature difference sequence δt is greater than or equal to the abnormality detection threshold K, then it is judged that the temperature difference sequence δt in this segment is abnormal, and the process goes to step S40), otherwise it is judged that the sensor is normal;

S40) Judging that there is an abnormal temperature difference sequence δt in step S30), the distribution consistency between the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period; if there is consistency, it is judged that the sensor is normal, if not If there is consistency, it is judged that the sensor is abnormal.

Further, the step S10) further includes:

S11) Select the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculate the temperature difference sequence δT according to the unit time ΔT;

S12) the temperature difference sequence δT in unit time is segmented by the same duration T1, calculate the standard deviation θ _i of each section of temperature difference sequence δT, and form the standard deviation sequence θ;

的原则构建列车该部位对应的异常检测阈值K；S13) Analyze the distribution of the standard deviation sequence Θ, and calculate the mean μ and standard deviation σ of the standard deviation sequence Θ, according to the probability of occurrence

The principle of constructing the anomaly detection threshold K corresponding to this part of the train;

Among them, θ is the standard deviation of the temperature difference series δT.

x_i为样本值，n为样本数。Among them, ω _i is the weighting coefficient, here

x _i is the sample value, and n is the number of samples.

Further, the step S20) further includes:

S21) real-time input temperature data sequence t measured by the sensor;

S22) Calculate temperature difference sequence δt according to unit time Δt;

S23) Segment the temperature difference sequence δt in a unit time according to the same duration T2, calculate the standard deviation of each segment of the temperature difference sequence δt, and form a standard deviation sequence η.

Further, the step S40) further includes:

S41) By comparing the abnormality detection threshold K and the standard deviation sequence η, it is found that a certain section of the temperature difference sequence δt appears abnormal data during the train operation, record the information of the temperature difference sequence x _t , and obtain the temperature difference of the section The value sequence x _t and the temperature difference sequence y _t of the previous adjacent time period;

S42) Obtain the temperature difference sequence z1 _t , . . . , zn _t measured by other similar position sensors of the train in the same time period;

S43) Take the temperature difference series x t suspected to be abnormal and the temperature difference series y _t of the previous adjacent time period respectively, and the temperature difference series {z1 _t ,...,zn _{t }} measured by other similar position sensors one by one KS distribution test is carried out;

S44) When the occurrence probability P value of all inspections is less than the set standard, output a sensor abnormality warning signal, otherwise the sensor is normal.

Further, the step S43) further comprises:

Let the sample size of the temperature difference sequence x _t be n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t ,…,zn _t be n2, F ₁ (x) and F ₂ (x) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample.

代表D_j绝对距离的最大值。检验统计量Z近似于正态分布，其表达式为：Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j . The test statistic Z is approximately normally distributed and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

为取B(F(x))绝对距离的最大值，x为样本。

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample.

The empirical distribution function B(t) is:

Among them, x is an independent variable, and i is a natural number;

When judging the distribution consistency between the temperature difference sequence x t to be tested and the temperature difference sequence y _t in the previous adjacent time period, and the temperature difference sequence {z1 _t ,...,zn _{t }} measured by other similar position sensors When , the significance of the temperature difference series x _t is determined by testing the maximum difference value D of the empirical distribution function between the series. When the actual calculated maximum difference value D is greater than a certain set standard value, or when the distribution probability P value corresponding to the maximum difference value D is less than a certain set standard value, there is no consistency between the two temperature difference series. sex.

By implementing the above technical solutions of the temperature sensor self-checking device and method provided by the present invention, the following beneficial effects are obtained:

(1) The present invention performs self-checking and early warning based on the change value of the temperature value measured by the sensor (system) itself. Compared with the technical scheme based on other variables such as current, voltage or comparing the measurement results of multiple devices in the prior art, it can be more Effective and more direct discovery of possible anomalies, monitoring and early warning results will be more real and accurate;

(2) The present invention not only uses the threshold index for self-inspection and early warning, but also further detects and finds abnormalities from the perspective of comparative distribution changes. Compared with the prior art, which only uses one or two indicator indicators to analyze sensor failures, self-inspection The rules and results of early warning are more accurate and effective;

(3) The present invention is based on a large number of normal and abnormal data in the actual operation process to carry out analysis and application. Compared with the problems such as the smaller amount of data based on the prior art, the model results are more reliable, and the factors considered are more sufficient and reasonable. , the testability and practicability are also stronger;

(4) The present invention performs real-time self-detection and early warning based on a large amount of temperature data measured during train operation, and automatically adjusts the threshold value and distribution test segmentation method based on the continuously updated data, which has remarkable efficiency and intelligence.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that are required to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and for those of ordinary skill in the art, other embodiments can also be obtained according to these drawings without creative efforts.

Fig. 1 is a system structure block diagram of a specific embodiment of a temperature sensor self-checking device of the present invention;

2 is a schematic diagram of the working flow principle of a specific embodiment of the temperature sensor self-checking method of the present invention;

Fig. 3 is the program flow chart of a specific embodiment of the temperature sensor self-checking method of the present invention;

4 is a schematic diagram of K-S inspection in a specific embodiment of the temperature sensor self-inspection method of the present invention;

In the figure: 1- anomaly detection threshold calculation module, 2- key feature value extraction module, 3- first anomaly detection module, 4- consistency checking module, 5- second anomaly detection module.

Detailed ways

For the sake of citation and clarity, the technical terms, abbreviations or abbreviations used hereinafter are recorded as follows:

Nonparametric test: refers to the method of using sample data to infer the population distribution pattern when the population variance is unknown or little known. Since nonparametric test methods do not involve parameters about the population distribution during inference, they are called "nonparametric" tests.

K-S test: The Kolmogorov-Smirnov test is based on a cumulative distribution function and is used to test whether two empirical distributions are different or whether one empirical distribution is different from another ideal distribution. It is different from other methods such as t test in that the K-S test does not need to know the distribution of the data, and can be regarded as a nonparametric test method. When the sample size is relatively small, the K-S test, as a nonparametric test, is a commonly used method to analyze whether there is a difference between two groups of data.

To make the purposes, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

As shown in Figures 1 to 4, specific embodiments of the temperature sensor self-checking device and method of the present invention are given, and the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

Example 1

Through the analysis and research of a large amount of data, it is found that the temperature change of each relevant part on the train is usually a relatively slow process. That is to say, the possibility of large fluctuations in temperature changes in a short period of time is small, especially in the case of a large temperature drop in a very short period of time, it is almost difficult to occur. Therefore, based on this consideration, if it is found that the temperature change value fluctuates sharply in a short period of time, and the distribution of the change value in a similar time period is significantly different, it means that the corresponding temperature sensor (system) is likely to appear exception. Therefore, the specific embodiment of the present invention measures the fluctuation level by the standard deviation of the temperature difference value per unit time, and combines the K-S distribution test method to realize the test and comparison of the temperature difference distribution in similar time periods, so as to comprehensively judge whether the temperature sensor is abnormal. The temperature sensor self-checking device described in this embodiment includes two major functions. The first part is the determination (modeling) of the abnormal temperature difference detection threshold, and the second part is combined with the threshold value and K-S test to realize the self-check of the temperature sensor abnormality.

In terms of determining the threshold value of the constant temperature difference detection, firstly, the difference processing is performed on the time series data of each temperature type under normal conditions of the train system, and then the volatility (standard deviation) is calculated in principle segments according to the processed temperature difference data to form the corresponding standard deviation. Then, the corresponding distribution is estimated based on the formed standard deviation sequence, the mean and standard deviation of the fluctuation sequence are calculated, and finally the anomaly detection threshold μ+3σ (that is, the anomaly detection threshold K) is obtained by combining the statistical distribution principle.

In terms of realizing the abnormal self-check of temperature sensor by combining the abnormal detection threshold and K-S test, for the real-time input temperature detection data, firstly calculate the difference and the corresponding fluctuation sequence according to the same rules as in the determination stage of the abnormal temperature detection threshold, and then based on the abnormality obtained above The detection threshold determines the abnormal possibility of the segmental fluctuation sequence. If it is greater than or equal to the abnormal detection threshold, it is determined that the temperature value of this segment is suspected to be abnormal, and then the distribution difference between this segment of the temperature difference sequence and the basic temperature difference sequence and the previous sequence is further judged. If the difference is obvious, it means that the sensor (system) is abnormal; otherwise, it is considered that there is no abnormality.

The specific implementation scheme of the temperature sensor self-checking device based on the above working principle will be described in detail below.

As shown in FIG. 1, an embodiment of a temperature sensor self-checking device specifically includes:

The abnormality detection threshold calculation module 1 is used to perform differential processing on the temperature data sequence T (such as: oil temperature, axle temperature, water temperature, etc.) measured by the sensor under the normal operation of the train to obtain the score of the temperature difference sequence δT. segment standard deviation sequence Θ, and obtain the anomaly detection threshold K by performing statistical analysis on the standard deviation sequence Θ;

The key feature value extraction module 2 is used to perform differential processing on the real-time input temperature data sequence t measured by the sensor to obtain a segmented standard deviation sequence η of the temperature difference sequence δt;

The first abnormality detection module 3 is used to judge whether there is an abnormality in the segmented temperature difference sequence δt according to the abnormality detection threshold K output by the abnormality detection threshold calculation module 1 and the standard deviation sequence n output by the key feature value extraction module 2; If the segmented standard deviation sequence η of the segment temperature difference sequence δt is greater than or equal to the abnormality detection threshold K, it is judged that the temperature difference sequence δt in this segment is abnormal, and the abnormal temperature difference sequence δt is output, otherwise it is judged that the sensor is normal ;

The consistency checking module 4 is used to perform a distribution consistency check on the abnormal temperature difference sequence δt output by the first anomaly detection module 3, the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period;

The second abnormality detection module 5 is used for judging whether the distribution consistency test occurrence probability P value output by the consistency test module 4 is less than the set standard, if it is less than the set standard, it outputs a sensor abnormality warning signal, otherwise the sensor is normal.

的原则构建列车该部位对应的异常检测阈值K。The abnormality detection threshold calculation module 1 obtains the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculates the temperature difference sequence δT per unit time ΔT. The temperature difference sequence δT in unit time is segmented according to the same duration T1, the standard deviation θ _i of each temperature difference sequence δT is calculated, and the standard deviation sequence Θ is formed. Analyze the distribution of the standard deviation series Θ, and calculate the mean μ and standard deviation σ of the standard deviation series Θ, according to the probability of occurrence (ie, the probability of occurrence of the distribution consistency test)

The principle of constructing the anomaly detection threshold K corresponding to this part of the train.

Among them, θ is the standard deviation of the temperature difference series δT.

x_i为样本值，n为样本数。Among them, ω _i is the weighting coefficient, here

x _i is the sample value, and n is the number of samples.

The key feature value extraction module 2 obtains the temperature data sequence t measured by the real-time input sensor, calculates the temperature difference sequence δt per unit time Δt, and divides the temperature difference sequence δt per unit time according to the same duration T2 to calculate the temperature of each segment standard deviation of the difference series δt, and form the standard deviation series η.

The consistency check module 4, by comparing the abnormality detection threshold K and the standard deviation sequence η, finds that there is suspected abnormal data in a certain temperature difference sequence δt during the operation of the train, records the information of the temperature difference sequence x _t , and obtains the data. The temperature difference sequence x _t of the segment and the temperature difference sequence y _t of the previous adjacent time segment are obtained, and the temperature difference sequence z1 _t ,...,zn _t measured by other similar position sensors of the train in the same time segment is obtained at the same time. The suspected abnormal temperature difference sequence x _t and the temperature difference sequence y _t of the previous adjacent time period, and the temperature difference sequence {z1 _t ,...,zn _t } measured by other similar position sensors, respectively perform KS one by one. Distribution test.

The consistency checking module 4 is judging the temperature difference series x t to be checked and the temperature difference series y _t of the previous adjacent time period, and the temperature difference series _{{z1 t ,...,zn t measured by} other similar position sensors When the distribution consistency of }, the significance of the temperature difference series x _t is determined by checking the maximum difference value D of the empirical distribution function between the series. When the actual calculated maximum gap value D is greater than a certain set standard value, or the KS distribution probability P value corresponding to the maximum gap value D (that is, the probability P value of distribution consistency test occurrence) is less than a set standard value, Then there is no consistency between the two temperature difference series.

Among them, the sample size of the temperature difference sequence x _t is n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t ,…,zn _t is n2, F ₁ (x) and F ₂ (x ) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample.

代表D_j绝对距离的最大值。检验统计量Z近似于正态分布，其表达式为：Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j . The test statistic Z is approximately normally distributed and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

为取B(F(x))绝对距离的最大值，x为样本。

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample.

The empirical distribution function B(t) is:

Among them, x is an independent variable, and i is a natural number.

Example 2

In this embodiment, since the data used is mainly the time series temperature data measured by the sensor, during the real-time abnormal self-check, there are relative differences due to the different environments, lines and states of different trains, which cannot be directly passed through. Therefore, it is necessary to classify and reconstruct the temperature sequence data, establish a difference sequence and calculate the volatility in segments before performing anomaly detection. As shown in Figure 2 and Figure 3, an embodiment of a temperature sensor self-checking method specifically includes the following steps:

S10) carry out differential processing to the temperature data sequence T measured by the sensor under normal operation of the train to obtain the segmented standard deviation sequence Θ of the temperature difference sequence δT, and obtain the abnormal detection threshold K by performing statistical analysis on the standard deviation sequence Θ;

S20) perform the same differential processing as in step S10) on the real-time input sensor temperature data sequence t to obtain the segmented standard deviation sequence η of the temperature difference sequence δt;

S30) Judge whether there is an abnormality in the segmented temperature difference sequence δt based on the abnormality detection threshold K obtained in step S10) and the standard deviation sequence η obtained in step S20); if the segmented standard deviation sequence η of a certain segment of temperature difference sequence δt is greater than or equal to the abnormality detection threshold K, then it is judged that the temperature difference sequence δt in this segment is abnormal, and the process goes to step S40), otherwise it is judged that the sensor is normal;

S40) Judging that there is an abnormal temperature difference sequence δt in step S30), the distribution consistency between the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period; if there is consistency, it is judged that the sensor is normal, if not If there is consistency, it is judged that the sensor is abnormal.

Step S10) further comprises:

S11) Select the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculate the temperature difference sequence δT according to the unit time ΔT (eg: 1s);

S12) the temperature difference sequence δT in unit time is segmented by the same duration T1, calculate the standard deviation θ _i of each section of temperature difference sequence δT, and form the standard deviation sequence θ;

的原则构建列车该部位对应的异常检测阈值K；S13) Analyze the distribution of the standard deviation sequence Θ, and calculate the mean μ and standard deviation σ of the standard deviation sequence Θ, according to the probability of occurrence

The principle of constructing the anomaly detection threshold K corresponding to this part of the train;

S14) Calculate the abnormality detection threshold K corresponding to different parts of the same train and different parts of the train in the same manner as steps S11) to S13), and form a temperature sensor abnormality self-checking threshold matrix.

Among them, θ is the standard deviation of the temperature difference series δT.

x_i为样本值，n为样本数。Among them, ω _i is the weighting coefficient, here

x _i is the sample value, and n is the number of samples.

The mean-μ here usually refers to the arithmetic mean of the sample, which represents the number of trends in a set of data, and refers to the sum of all data in a set of data divided by the number of data in this set, which is a reflection of the data set. an indicator of a trend.

Standard Deviation—σ is the square root of the squared deviation from the mean and the mean, or the arithmetic square root of the variance. The standard deviation can reflect the degree of dispersion of a data set, or it can also be called the degree of fluctuation. If the mean is the same, the standard deviation may not be the same. Standard deviation can be used as a measure of uncertainty. For example: in practical measurement science, when repeatable measurements are made, the standard deviation of the set of measurement values represents the precision of those measurements. The standard deviation of the measured value plays a decisive role when it comes to determining whether the measured value matches the predicted value: if the measured mean is too far from the predicted value (and compared to the standard deviation value), the measured value and the predicted value are considered contradictory . Because if the measured values all fall outside a certain numerical range, it is reasonable to infer whether the predicted value is correct.

Step S20) further comprises:

S21) real-time input temperature data sequence t measured by the sensor;

S22) Calculate temperature difference sequence δt according to unit time Δt;

S23) Segment the temperature difference sequence δt in a unit time according to the same duration T2, calculate the standard deviation of each segment of the temperature difference sequence δt, and form a standard deviation sequence η.

Since the above steps are used to judge the abnormality of the sensor (system) based on the distribution threshold of the temperature difference subsection sequence fluctuation, there are still some problems that may falsely report the abnormality, for example, the occasional jump caused by a signal problem at a certain point in time does not belong to the sensor. Or the system is abnormal, and it may be judged as a sensor (system) abnormality through the threshold judgment, resulting in a false alarm. For this reason, it is necessary to strengthen the judgment principle based on the threshold judgment and the characteristics of the overall distribution of the temperature difference.

The reason for improving the accuracy of abnormal self-checking through distribution testing is that, under normal circumstances, in a short period of time, the factors that affect temperature changes at the same location of the same train cannot change greatly and fundamentally, so adjacent short-term The temperature difference distribution over time should belong to the same overall distribution, and there should be no significant distribution differences unless there is a problem with the sensor or system that measures the temperature. In addition, when it is impossible to determine which parameter distribution the temperature difference changes of different trains and different parts belong to, the consistency of the distribution through non-parametric test is more in line with the characteristics of the actual data itself. Therefore, the K-S two-sample distribution test method is described in this embodiment. It has become a very suitable choice in the technical solution.

For two samples from two different populations, if you want to test whether the overall distribution behind them is consistent, you can perform a two-sample K-S test. The distribution of is replaced by the empirical distribution of another sample, and the specific steps are as follows.

Step S40) further comprises:

S41) By comparing the abnormality detection threshold K and the standard deviation sequence η, it is found that a certain section of the temperature difference sequence δt appears abnormal data during the train operation, record the information of the temperature difference sequence x _t , and obtain the temperature difference of the section The value sequence x _t and the temperature difference sequence y _t of the previous adjacent time period;

S42) Obtain the temperature difference sequence z1 _t , . . . , zn _t measured by other similar position sensors of the train in the same time period;

S43) Take the temperature difference series x t suspected to be abnormal and the temperature difference series y _t of the previous adjacent time period respectively, and the temperature difference series {z1 _t ,...,zn _{t }} measured by other similar position sensors one by one KS distribution test is carried out;

S44) When the occurrence probability P value of all inspections is less than the set standard, output a sensor abnormality warning signal, otherwise the sensor is normal, and the sensor abnormality warning signal is not output.

Step S43) further comprises:

Let the sample size of the temperature difference sequence x _t be n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t ,…,zn _t be n2, F ₁ (x) and F ₂ (x) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample.

代表D_j绝对距离的最大值。检验统计量Z近似于正态分布，其表达式为：Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j . The test statistic Z is approximately normally distributed and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

为取B(F(x))绝对距离的最大值，x为样本。

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample.

The empirical distribution function (ie the Kolmogonov distribution function) B(t) is:

Among them, x is an independent variable, and i is a natural number;

As shown in Figure 4, when judging the temperature difference sequence x t to be tested and the temperature difference sequence y _t of the previous adjacent time period, and the temperature difference sequence {z1 _{t ,} ..., When the distribution of zn _t } is consistent, the significance of the temperature difference series x _t is determined by checking the maximum difference value D of the empirical distribution function between the series. When the actual calculated maximum gap value D is greater than a certain set standard value, or the KS distribution probability P value corresponding to the maximum gap value D is less than a set standard value, the two sequences from the same distribution population are rejected. Assume that there is a clear inconsistency between the two series of temperature difference values. That is to say, there is a reason to show that there is an obvious difference between the two sequences, which is inconsistent with the premise that the distribution of temperature difference sequences in a short period of time will not change greatly, and further shows that the abnormality is caused by the temperature sensor (system). .

The temperature sensor self-checking device and method described in the specific embodiments of the present invention are based on the big data platform, combined with the temperature data measured by the sensors of the relevant parts of the train (axles, motors, transformers, converters, etc.) fed back on the spot, A set of temperature sensor (system) abnormal self-checking device and method is constructed, which realizes the automation, intelligent self-checking and early warning of the train sensing system. Fast and effective self-inspection, thereby ensuring the safe and efficient operation of the train. In the specific embodiment of the present invention, by calculating the distribution characteristics of the fluctuation sequence of the temperature difference, the detection threshold of the abnormal temperature difference is determined with a certain probability, and the fluctuation change of the abnormal data of the sensor in a short time is well reflected, and the sensor ( The abnormal change of the measured temperature value caused by the abnormality of the system) greatly improves the self-test efficiency of the abnormality of the sensor (system). At the same time, the specific embodiment of the present invention is based on the characteristics and laws of temperature changes of train components, uses the property that the temperature difference distribution does not change much in a short period of time, and combines the non-parametric test method to compare the difference of the temperature difference distribution between different time periods, which is extremely effective in reducing the sensor. (System) The false alarm rate of abnormal self-check and early warning greatly improves the overall accuracy of prediction.

It should be particularly noted that, in the above-mentioned specific embodiments of the present invention, the method of processing the temperature difference sequence through the standard deviation can also be studied and applied through indicators such as dispersion coefficient and range. At the same time, the K-S non-parametric distribution test method used in the specific embodiment of the present invention can also try to use non-parametric methods such as unit root test, symbol test and other parametric test methods to test the distribution difference. The temperature sensor self-checking device and method described in the specific embodiments of the present invention can be implemented by using the reference codes as R and Python codes, or by using a series of languages such as C, MATLAB, and Java.

By implementing the technical solutions of the temperature sensor self-checking device and method described in the specific embodiments of the present invention, the following technical effects can be produced:

(1) The temperature sensor self-checking device and method described in the specific embodiment of the present invention performs self-checking and early warning based on the change value of the temperature value measured by the sensor (system) itself, compared with other variables such as current, voltage, etc. in the prior art. Compared with the technical solution of measuring results of multiple devices, possible anomalies can be found more effectively and directly, and the monitoring and early warning results will be more real and accurate;

(2) The temperature sensor self-checking device and method described in the specific embodiment of the present invention not only uses the threshold index to carry out self-checking and early warning, but also further detects and finds abnormalities from the perspective of comparative distribution changes. Compared with the prior art, only one or two are used. For the analysis of sensor faults with an indicator index, the rules and results of self-inspection and early warning are more accurate and effective;

(3) The temperature sensor self-checking device and method described in the specific embodiments of the present invention carry out analysis and application based on a large number of normal and abnormal data in the actual operation process, compared with the problems such as less data based on the prior art, The model results are more reliable, the factors considered are more sufficient and reasonable, and the testability and practicability are stronger;

(4) The temperature sensor self-checking device and method described in the specific embodiment of the present invention performs real-time self-checking and early warning based on a large amount of temperature data measured during the operation of the train, and automatically adjusts the threshold value and the distribution check segmentation method based on the continuously updated data. , with remarkable efficiency and intelligence level.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other.

The above descriptions are only preferred embodiments of the present invention, and do not limit the present invention in any form. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art, without departing from the spirit and technical solutions of the present invention, can make many possible changes and modifications to the technical solutions of the present invention by using the methods and technical contents disclosed above, or modify them to be equivalent. Variant equivalent embodiments. Therefore, any simple modification, equivalent replacement, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still belong to the protection scope of the technical solution of the present invention.

Claims (6)

1. a temperature sensor self-checking device, is characterized in that, comprises: The abnormal detection threshold calculation module (1) is used to perform differential processing on the temperature data sequence T measured by the sensor under normal operation of the train to obtain the segmented standard deviation sequence Θ of the temperature difference sequence δT, and by comparing the standard deviation sequence Θ Perform statistical analysis to obtain anomaly detection threshold K; The key feature value extraction module (2) is used to perform differential processing on the real-time input temperature data sequence t measured by the sensor to obtain a segmented standard deviation sequence η of the temperature difference sequence δt; The first abnormality detection module (3) is used to judge the segment temperature according to the abnormality detection threshold value K output by the abnormality detection threshold value calculation module (1), and the standard deviation sequence n output by the key feature value extraction module (2) Whether the difference sequence δt is abnormal; if the segmented standard deviation sequence η of a certain temperature difference sequence δt is greater than or equal to the anomaly detection threshold K, it is judged that there is an abnormality in the temperature difference sequence δt, and the abnormality of this segment is output. Temperature difference sequence δ t, otherwise the sensor is judged to be normal; The consistency checking module (4) is used to perform distribution consistency on the abnormal temperature difference sequence δt output by the first anomaly detection module (3), the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period test; The second abnormality detection module (5) is used for judging whether the distribution consistency test occurrence probability P value output by the consistency test module (4) is less than the set standard, if it is less than the set standard, then output the sensor abnormality warning signal, Otherwise the sensor is normal; The abnormality detection threshold calculation module (1) obtains the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculates the temperature difference sequence δT per unit time ΔT; for the temperature difference sequence δT per unit time According to the same time length T1, calculate the standard deviation θ _i of each temperature difference sequence δT, and form the standard deviation sequence Θ; analyze the distribution of the standard deviation sequence Θ, and calculate the mean μ and standard deviation σ of the standard deviation sequence Θ , according to the probability of occurrence

The principle of constructing the anomaly detection threshold K corresponding to this part of the train; Among them, θ is the standard deviation of the temperature difference series δT;

Among them, ω _i is the weighting coefficient, here

x _i is the sample value, n is the number of samples; The key characteristic value extraction module (2) obtains the temperature data sequence t measured by the real-time input sensor, calculates the temperature difference value sequence δt per unit time Δt, and divides the temperature difference value sequence δt per unit time according to the same duration T2, Calculate the standard deviation of each temperature difference series δt, and form the standard deviation series η. 2. The temperature sensor self-checking device according to claim 1, wherein the consistency checking module (4) finds a certain section of temperature difference sequence in the train running process by comparing the abnormality detection threshold K and the standard deviation sequence η After the suspected abnormal data appears in δt, record the information of the temperature difference sequence x _t , and obtain the temperature difference sequence x _t of this segment and the temperature difference sequence y _t of the previous adjacent time period, and obtain the same time The temperature difference series z1 _t , ... , zn _t measured by other similar position sensors of the train, and the temperature difference series x _t suspected to be abnormal and the temperature difference series y _t of the previous adjacent time period, and other The temperature difference sequence _{ z1 _t , . 3. The temperature sensor self-checking device according to claim 1 or 2, wherein the consistency checking module (4) judges the temperature difference between the temperature difference sequence x _t to be checked and the previous adjacent time period When the value sequence y _t and the temperature difference sequence {z1 _t , _. The significance of x _t ; when the actual calculated maximum difference value D is greater than a set standard value, or the distribution probability P value corresponding to the maximum gap value D is less than a set standard value, then the difference between the two temperatures There is no consistency between sequences; Among them, the sample size of the temperature difference sequence x _t is n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t , ..., znt _t is n2, F ₁ (x) and F ₂ (x ) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample; Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j ; the test statistic Z approximates a normal distribution, and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample; The empirical distribution function B(t) is:

Among them, x is an independent variable, and i is a natural number. 4. a temperature sensor self-checking method, is characterized in that, comprises the following steps: S10) carry out differential processing to the temperature data sequence T measured by the sensor under normal operation of the train to obtain the segmented standard deviation sequence Θ of the temperature difference sequence δT, and obtain the abnormal detection threshold K by performing statistical analysis on the standard deviation sequence Θ; S20) perform the same differential processing as in step S10) on the real-time input sensor temperature data sequence t to obtain the segmented standard deviation sequence η of the temperature difference sequence δt; S30) Judge whether there is an abnormality in the segmented temperature difference sequence δt based on the abnormality detection threshold K obtained in step S10) and the standard deviation sequence η obtained in step S20); if the segmented standard deviation sequence η of a certain segment of temperature difference sequence δt is greater than or equal to the abnormality detection threshold K, then it is judged that the temperature difference sequence δt in this segment is abnormal, and the process goes to step S40), otherwise it is judged that the sensor is normal; S40) Judging that there is an abnormal temperature difference sequence δt in step S30), the distribution consistency between the normal reference sequence and the temperature difference sequence δt in the previous adjacent time period; if there is consistency, it is judged that the sensor is normal, if not If there is consistency, it is judged that the sensor is abnormal; Described step S10) further comprises: S11) Select the temperature data sequence T measured by the sensor during the operation of a certain part of the vehicle under normal conditions, and calculate the temperature difference sequence δT according to the unit time ΔT; S12) the temperature difference sequence δT in unit time is segmented by the same duration T1, calculate the standard deviation θ _i of each section of temperature difference sequence δT, and form the standard deviation sequence θ; S13) Analyze the distribution of the standard deviation sequence Θ, and calculate the mean μ and standard deviation σ of the standard deviation sequence Θ, according to the probability of occurrence

The principle of constructing the anomaly detection threshold K corresponding to this part of the train; Among them, θ is the standard deviation of the temperature difference series δT;

Among them, ω _i is the weighting coefficient, here

x _i is the sample value, n is the number of samples; Described step S20) further comprises: S21) real-time input temperature data sequence t measured by the sensor; S22) Calculate temperature difference sequence δt according to unit time Δt; S23) Segment the temperature difference sequence δt in unit time by the same duration T2, calculate the standard deviation of each segment of the temperature difference sequence δt, and form a standard deviation sequence η. 5. The temperature sensor self-checking method according to claim 4, wherein, The step S40) further includes: S41) By comparing the abnormality detection threshold K and the standard deviation sequence η, it is found that a certain section of the temperature difference sequence δt appears abnormal data during the train operation, record the information of the temperature difference sequence x _t , and obtain the temperature difference of the section The value sequence x _t and the temperature difference sequence y _t of the previous adjacent time period; S42) Obtain the temperature difference value sequence z1 _t , . . . , zn _t measured by other similar position sensors of the train in the same time period; S43) The temperature difference sequence x t that is suspected to be abnormal, the temperature difference sequence y _t of the previous adjacent time period, and the temperature difference sequence {z1 _t , . . . , zn _t } measured by other similar position sensors _, one by one KS distribution test is carried out; S44) When the occurrence probability P value of all inspections is less than the set standard, output a sensor abnormality warning signal, otherwise the sensor is normal. 6. The temperature sensor self-checking method according to claim 4 or 5, wherein the step S43) further comprises: Let the sample size of the temperature difference sequence x _t be n1, the sample size of any difference sequence in the temperature difference sequence y _t , z1 _t , ..., zn _t be n2, F ₁ (x) and F ₂ (x) respectively represent the cumulative empirical distribution function of the two samples, j is the segment identification of the temperature difference sequence, and x is the sample; Denote D _j =F ₁ (x _j )-F ₂ (x _j ),

Represents the maximum value of the absolute distance of D _j ; the test statistic Z approximates a normal distribution, and its expression is:

When the null hypothesis is true, Z converges to the K distribution according to the density distribution d, that is, when the sample is taken from the one-dimensional continuous distribution F,

In order to take the maximum value of the absolute distance of B(F(x)), x is the sample; The empirical distribution function B(t) is:

Among them, x is an independent variable, and i is a natural number; When judging the distribution consistency of the temperature difference sequence x _t to be tested and the temperature difference sequence y _t of the previous adjacent time period, and the temperature difference sequence {z1 _t ,..., zn _t } measured by other similar position sensors When , the significance of the temperature difference series x _t is determined by checking the maximum difference value D of the empirical distribution function between the series; when the actual calculated maximum difference value D is greater than a certain set standard value, or the corresponding maximum difference value D When the distribution probability P value of , is less than a certain standard value, there is no consistency between the two temperature difference series.

Images