Signal machine fault monitoring method based on Weibull distribution

Abstract

The invention discloses a signal machine fault monitoring method based on Weibull distribution, and the method comprises the steps: S1, collecting the historical fault data of a plurality of signal machine subsystems, and determining the fault distribution type and distribution parameters of each signal machine subsystem, so as to obtain the number of safe operation days of each signal machine subsystem; s2, adopting a PSO algorithm to optimize a neural network model, and utilizing the neural network model to predict the fault rate of each annunciator subsystem; and S3, constructing a risk model of each annunciator subsystem by taking the number of safe operation days and the failure rate of each annunciator subsystem as adjoint variables so as to predict possible faults of each annunciator subsystem. The method can solve the problem that a traditional signal machine fault monitoring method is generally based on historical data and experience of a signal machine and cannot accurately monitor and prevent the fault of the signal machine.

Description

Technical Field

[0001] The invention relates to the field of rail transit monitoring, and in particular to a signal machine fault monitoring method based on Weibull distribution. Background Art

[0002] Signals are an important part of rail transit systems, but they are prone to random and uncertain failures during long-term operation, which can lead to traffic chaos and safety issues. However, traditional fault monitoring methods are usually based on historical data and experience of signal machines, lacking accurate monitoring and prevention of signal machine failures.

[0003] As a probability distribution model widely used in the field of reliability engineering, Weibull distribution can better describe the fault characteristics of signal equipment at different stages. By collecting and analyzing the historical data and operation data of the signal, a fault monitoring model based on Weibull distribution can be established, which can achieve early warning and accurate diagnosis of signal faults, improve the reliability and availability of the signal, and ensure the safety and smoothness of road traffic.

[0004] Therefore, it is of great practical significance to develop a signal fault monitoring method based on Weibull distribution to improve the monitoring accuracy and prevention effect. Summary of the invention

[0005] The purpose of the present invention is to provide a signal fault monitoring method based on Weibull distribution, which can solve the problem that traditional signal fault monitoring methods are usually based on historical data and experience of signals and lack accurate monitoring and prevention of signal faults.

[0006] To achieve the above object, the present invention provides a signal fault monitoring method based on Weibull distribution, comprising:

[0007] S1, respectively collecting historical fault data of a plurality of signal subsystems, and determining the fault distribution type and distribution parameter of each signal subsystem, so as to obtain the number of safe operation days of each signal subsystem;

[0008] S2, using a PSO algorithm to optimize a neural network model, and using the neural network model to predict the failure rate of each of the signal subsystems;

[0009] S3, taking the safe operation days and failure rate of each signal subsystem as accompanying variables, constructing a risk model for each signal subsystem to predict possible failures of each signal subsystem.

[0010] Optionally, step S1 includes:

[0011] S1.1, respectively collecting historical fault data of each signal subsystem;

[0012] S1.2, analyzing the historical fault data of each of the signal subsystems by survival analysis method, and determining the fault distribution model and distribution parameters corresponding to the historical fault data distribution of each of the signal subsystems;

[0013] S1.3, obtaining the fault distribution model and distribution parameters of the historical fault data distribution of each of the signal subsystems, and obtaining the time point of the fault occurrence of each of the signal subsystems through the interval time between the fault occurrence time points.

[0014] Optionally, the fault distribution model includes: a two-parameter Weibull distribution model, a three-parameter Weibull distribution model;

[0015] Wherein, the two-parameter Weibull distribution model is characterized by a two-parameter fault distribution function, a two-parameter reliability function, a two-parameter probability density function and a two-parameter failure rate function;

[0016] The three-parameter Weibull distribution model is characterized by a three-parameter failure distribution function, a three-parameter reliability function, a three-parameter probability density function and a three-parameter failure rate function.

[0017] Optionally, the step S1.2 comprises:

[0018] Import the historical fault data of each signal subsystem respectively through the survival statistics tool;

[0019] Using the maximum likelihood estimation method to fit the historical fault data of each signal subsystem respectively, so as to determine the fault distribution model to which each signal subsystem belongs;

[0020] By constructing the likelihood function, the parameters in the distribution model are estimated for each signal subsystem that determines the fault distribution model.

[0021] Optionally, step S2 comprises:

[0022] S2.1, using PSO algorithm to set the optimal parameters of BP neural network;

[0023] S2.2, respectively screening the historical fault data of each of the signal subsystems to obtain training data samples and test data samples corresponding to each of the signal subsystems;

[0024] S2.3, using the optimal parameters obtained by PSO algorithm training and the training data samples of each signal subsystem as input to train the BP neural network, and obtaining the BP neural network prediction model corresponding to each signal subsystem;

[0025] S2.4, respectively input the test data samples of each signal subsystem into the corresponding BP neural network prediction model, and respectively obtain the PSO-BP prediction model of each signal subsystem to respectively predict the failure rate of each signal subsystem.

[0026] Optionally, the setting of the optimal parameters includes:

[0027] Initialize the number of particles and maximum speed for the PSO algorithm;

[0028] Fitness evaluation, use the particle position as the weight and threshold of the BP neural network to perform a complete training, and calculate the fitness f of each particle;

[0029] According to the fitness value f of each particle, update the speed and position of the particle;

[0030] The weight and threshold corresponding to the particle with the best fitness are selected as the optimal parameters.

[0031] Optionally, the calculation formula of the fitness function is:

[0032]

[0033] Where N is the number of samples, y i is the expected output of the ith sample, is the actual output value of the i-th sample.

[0034] Optionally, the velocity change of the particle complies with the following velocity change formula:

[0035]

[0036] In the formula, represents the flying speed of the i-th particle in the k+1th generation; w represents the inertia weight; represents the flying speed of the i-th particle in the k-th generation; c1 and c2 represent learning factors; r1 and r2 represent random distribution numbers; represents the best position of the i-th particle to the k-th generation; represents the position of the i-th particle in the k-th generation; Represents the best position of the particle swarm to the kth generation.

[0037] Optionally, the position change of the particle complies with the following position change formula:

[0038]

[0039] In the formula, represents the position of the i-th particle in the k+1th generation; X i k represents the position of the i-th particle in the k-th generation.

[0040] Optionally, step S3 includes:

[0041] S3.1, taking the number of safe operation days and failure rate of each signal subsystem as accompanying variables, risk modeling is performed on each signal subsystem;

[0042] S3.2, verification and optimization, compares the actual failure data with the predicted results after risk modeling.

[0043] Optionally, the risk function formed by the risk modeling is:

[0044]

[0045] Where β is the distribution shape parameter, η is the distribution scale parameter, γ is the distribution location parameter, t is the time point when the fault occurs, and X(t) is the accompanying variable.

[0046] Optionally, the risk function includes upper and lower control limits;

[0047] Wherein, the upper control limit is a fault threshold, and when the upper control limit is exceeded, it indicates that the current signal subsystem is unavailable;

[0048] The lower control limit is a preventive threshold. When it is lower than the lower control limit, it indicates that potential failures begin to occur in the current signal subsystem.

[0049] Optionally, the multiple signal machine subsystems are: signal light source system, signal control and drive system, mechanical structure system, power supply system, monitoring and maintenance auxiliary system.

[0050] The present invention also provides a computer device, which includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the signal machine fault monitoring method based on Weibull distribution.

[0051] The present invention also provides a computer-readable storage medium, in which at least one program code is stored. The program code is loaded and executed by a processor to implement the signal machine fault monitoring method based on Weibull distribution.

[0052] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. A signal machine fault monitoring method based on Weibull distribution provided by the present invention better fits the signal machine fault mode by introducing Weibull distribution to capture the signal machine fault characteristics; and more accurately reflects the actual operation status and fault trend of the signal machine through accurate estimation and dynamic adjustment of Weibull distribution parameters, thereby improving the monitoring accuracy;

[0054] 2. The signal machine fault monitoring method based on Weibull distribution provided by the present invention dynamically monitors the fault of the signal machine by continuously updating data and adjusting the neural network model to adapt to the changes in the operating state of the signal machine, and has better reliability and availability;

[0055] 3. The present invention provides a signal light fault monitoring method based on Weibull distribution, which can integrate various operating data of the signal light, such as operating time, environmental parameters, maintenance records, etc., that is, by cleaning, integrating and extracting features from multi-source data, the quality and availability of the data are improved, thereby improving the accuracy of fault monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flow chart of a signal fault monitoring method of the present invention;

[0057] Figure 2 It is a modeling flow chart of the PSO-BP prediction model of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the attached Figures 1-2 , the technical content, structural features, objectives achieved and effects of the present invention are described in detail through preferred embodiments.

[0059] It should be noted that the drawings are in a very simplified form and use non-precise proportions. They are only used to conveniently and clearly assist in explaining the embodiments of the present invention, and are not used to limit the conditions for the implementation of the present invention. Therefore, they have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical content disclosed by the present invention.

[0060] In the description of the present invention, it should be noted that the terms "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0061] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0062] The present invention provides a signal machine fault monitoring method based on Weibull distribution, which performs fault monitoring on multiple signal machine subsystems contained in the signal machine; specifically, the present invention adopts five signal machine subsystems, which are: signal light source system, signal control and drive system, mechanical structure system, power supply system, monitoring and maintenance auxiliary system. Figure 1 As shown, the signal machine fault monitoring method comprises the following steps:

[0063] S1, collect historical fault data of signal light source system, signal control and drive system, mechanical structure system, power supply system and monitoring and maintenance auxiliary system respectively, and determine the fault distribution type and distribution parameters of each signal subsystem based on the survival analysis method, so as to conduct a preliminary analysis on the reliability and safe operation days of each signal subsystem;

[0064] S2, using a neural network model optimized by a PSO algorithm to further predict the failure rates of the signal light source system, signal control and drive system, mechanical structure system, power supply system, and monitoring and maintenance auxiliary system;

[0065] S3, taking the safe operation days and failure rate of each signal subsystem as accompanying variables, construct the risk model of each signal subsystem to predict the possible failures of each signal subsystem.

[0066] The fault distribution type of the signal light subsystem is determined by a fault distribution model; the fault distribution model includes: a two-parameter Weibull distribution model and a three-parameter Weibull distribution model.

[0067] Furthermore, the two-parameter Weibull distribution model presents the following functional rules: two-parameter fault distribution function F(t), two-parameter reliability function R(t), two-parameter probability density function f(t), and two-parameter failure rate function λ(t):

[0068]

[0069] Where t≥0, β>0, η>0, t is the time point when the fault occurs, β is the distribution shape parameter, and η is the distribution scale parameter.

[0070] The three-parameter Weibull distribution model presents the following functional rules: three-parameter failure distribution function F'(t), three-parameter reliability function R'(t), three-parameter probability density function f'(t), and three-parameter failure rate function λ'(t):

[0071]

[0072] Where t is the time point when the fault occurs, β is the distribution shape parameter, η is the distribution scale parameter, and γ is the distribution location parameter.

[0073] Wherein, the step S1 comprises the following steps:

[0074] S1.1, collecting historical fault data of the signal light source system, signal control and drive system, mechanical structure system, power supply system and monitoring and maintenance auxiliary system;

[0075] S1.2, analyzing the historical fault data of each signal subsystem by survival analysis method, and determining the fault distribution model and distribution parameters to which the historical fault data of these signal subsystems belong respectively; specifically including:

[0076] S1.2.1, using the survival statistics tool, import the historical fault data of the signal light source system, signal control and drive system, mechanical structure system, power supply system, and monitoring and maintenance auxiliary system respectively;

[0077] S1.2.2, using the maximum likelihood estimation method to fit the historical fault data of the signal light source system, signal control and drive system, mechanical structure system, power supply system, monitoring and maintenance auxiliary system, respectively, to determine the fault distribution model to which these signal subsystems belong;

[0078] S1.2.3, through the constructed likelihood function, the parameters in the distribution model of the signal machine subsystem that determines the fault distribution model are estimated.

[0079] In a specific embodiment of the present invention, taking the signal light source system as an example, the fault distribution model of the signal light source system is a three-parameter Weibull distribution model; further, the distribution shape parameter β in the three-parameter Weibull distribution model of the signal light source system is 1.30274, the distribution scale parameter η is 6.65490, and the distribution position parameter γ is 30.8751.

[0080] Specifically, the three-parameter Weibull distribution model to which the signal light source system belongs presents the following function law:

[0081] The signal light source system fault distribution function F”(t), the signal light source system reliability function R”(t), the signal light source system fault time probability density function f”(t), and the signal light source system fault rate function λ”(t) are respectively:

[0082]

[0083] Where t is the time point when the fault occurs.

[0084] Similarly, the historical fault data of the signal control and drive system, mechanical structure system, power supply system, monitoring and maintenance auxiliary system can be analyzed to obtain the fault distribution models of these signal subsystems and determine the distribution parameters in the fault distribution models.

[0085] S1.3, by obtaining the fault distribution model and distribution parameters of the historical fault data distribution of each signal subsystem, the time point of the failure of each signal subsystem is obtained; and the number of days of safe operation of each signal subsystem is obtained through the interval time between the failure time points.

[0086] Among them, Figure 2 As shown, step S2 includes the following steps:

[0087] S2.1, using PSO (particle swarm optimization) algorithm to set the optimal parameters of BP (back propagation) neural network; specifically including:

[0088] S2.1.1, initialize the number of particles and maximum speed for the PSO algorithm:

[0089] Specifically, select the number of particles as 30 and the maximum particle speed as 0.5.

[0090] S2.1.2, fitness evaluation, use the particle position as the weight and threshold of the BP neural network to conduct a complete training cycle, and get the prediction error, and then calculate the fitness f of each particle based on the prediction error;

[0091] Among them, the formula of fitness function is:

[0092] Where RMSE stands for root mean square error, N is the number of samples, and y i is the expected output of the ith sample, is the actual output value of the i-th sample.

[0093] S2.1.3, according to the fitness value f of each particle, update the particle's velocity and position so that the particle moves towards a better solution.

[0094] The velocity change of the particle complies with the following velocity change formula:

[0095]

[0096] In the formula, represents the flying speed of the i-th particle in the k+1th generation; w represents the inertia weight; represents the flying speed of the i-th particle in the k-th generation; c1 and c2 represent learning factors; r1 and r2 represent random distribution numbers; represents the best position of the i-th particle to the k-th generation; represents the position of the i-th particle in the k-th generation; Represents the best position of the particle swarm to the kth generation.

[0097] The position change of the particle conforms to the following position change formula:

[0098]

[0099] In the formula, represents the position of the i-th particle in the k+1th generation; X i k represents the position of the i-th particle in the k-th generation.

[0100] S2.1.4, select the weight and threshold corresponding to the particle with the best fitness as the optimal parameters.

[0101] S2.2, filtering the historical fault data of each signal subsystem respectively to obtain training data samples and test data samples corresponding to each signal subsystem;

[0102] S2.3, using the optimal parameters obtained by PSO (particle swarm optimization) algorithm training and the training data samples of the signal subsystem as input to train the BP neural network, and obtaining the BP neural network prediction model corresponding to each signal subsystem;

[0103] S2.4, respectively input the test data samples of each signal subsystem into the corresponding BP neural network prediction model, and finally obtain the PSO-BP prediction model of each signal subsystem to predict the failure rate of each signal subsystem.

[0104] Wherein, the step S3 comprises the following steps:

[0105] S3.1, with the failure time point t (the interval between two adjacent failure time points is the number of safe operation days) and failure rate of each signal subsystem as the accompanying variables, risk modeling is performed on each signal subsystem to obtain the upper and lower control limits of each signal subsystem.

[0106] Among them, the upper control limit is the fault threshold. During operation, once it is found that the operating status value of the signal subsystem exceeds the upper control limit, it indicates that the signal subsystem is in an unusable state. Corrective maintenance or replacement should be carried out before reuse; the lower control limit is the preventive threshold. When it is lower than the lower control limit, it indicates that the signal subsystem begins to have potential faults.

[0107] S3.2, verification and optimization, compare the actual fault data with the prediction results after risk modeling to verify the effectiveness and accuracy of the fault monitoring method.

[0108] In a specific embodiment of the present invention, a risk model of each of the signal subsystems is constructed, wherein the risk function is:

[0109]

[0110] The precautionary threshold h* of the risk function is:

[0111]

[0112] Among them, X(t) is the accompanying variable;

[0113] Specifically, take the fault occurrence time t as the x coordinate and take As the y coordinate, a risk graph is formed, whose upper limit represents the upper control limit corresponding to the failure threshold, and the lower limit represents the lower control limit corresponding to the preventive threshold.

[0114] In summary, the signal machine fault monitoring method based on Weibull distribution provided by the present invention better fits the signal machine failure mode by introducing Weibull distribution to capture the signal machine failure characteristics; and better predicts the failure rate of the signal machine through the PSO-BP prediction model, thereby more accurately reflecting the actual operating status and failure trend of the signal machine and improving the monitoring accuracy.

[0115] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the invention. After reading the above content, it will be obvious to those skilled in the art for various modifications and substitutions of the present invention. Therefore, the protection scope of the present invention should be limited by the attached claims.

Claims

1. A signal machine fault monitoring method based on Weibull distribution, wherein the signal machine comprises a plurality of signal machine subsystems, characterized in that: The following steps are included: S1, respectively collecting historical fault data of a plurality of signal subsystems, and determining the fault distribution type and distribution parameter of each signal subsystem, so as to obtain the number of safe operation days of each signal subsystem; S2, using a PSO algorithm to optimize a neural network model, and using the neural network model to predict the failure rate of each of the signal subsystems; S3, taking the safe operation days and failure rate of each signal subsystem as accompanying variables, constructing a risk model for each signal subsystem to predict possible failures of each signal subsystem.

2. The signal fault monitoring method according to claim 1, characterized in that: The step S1 comprises: S1.1, respectively collecting historical fault data of each signal subsystem; S1.2, analyzing the historical fault data of each of the signal subsystems by survival analysis method, and determining the fault distribution model and distribution parameters corresponding to the historical fault data distribution of each of the signal subsystems; S1.3, obtaining the fault distribution model and distribution parameters of the historical fault data distribution of each of the signal subsystems, and obtaining the time point of the fault occurrence of each of the signal subsystems through the interval time between the fault occurrence time points.

3. The signal fault monitoring method according to claim 2, characterized in that: The fault distribution model includes: a two-parameter Weibull distribution model and a three-parameter Weibull distribution model; Wherein, the two-parameter Weibull distribution model is characterized by a two-parameter fault distribution function, a two-parameter reliability function, a two-parameter probability density function and a two-parameter failure rate function; The three-parameter Weibull distribution model is characterized by a three-parameter failure distribution function, a three-parameter reliability function, a three-parameter probability density function and a three-parameter failure rate function.

4. The signal fault monitoring method according to claim 2, characterized in that: The step S1.2 comprises: Import the historical fault data of each signal subsystem respectively through the survival statistics tool; Using the maximum likelihood estimation method to fit the historical fault data of each signal subsystem respectively, so as to determine the fault distribution model to which each signal subsystem belongs; By constructing the likelihood function, the parameters in the distribution model are estimated for each signal subsystem that determines the fault distribution model.

5. The signal fault monitoring method according to claim 1, characterized in that: The step S2 comprises: S2.1, using PSO algorithm to set the optimal parameters of BP neural network; S2.2, respectively screening the historical fault data of each of the signal subsystems to obtain training data samples and test data samples corresponding to each of the signal subsystems; S2.3, using the optimal parameters obtained by PSO algorithm training and the training data samples of each signal subsystem as input to train the BP neural network, and obtaining the BP neural network prediction model corresponding to each signal subsystem; S2.4, respectively input the test data samples of each signal subsystem into the corresponding BP neural network prediction model, and respectively obtain the PSO-BP prediction model of each signal subsystem to respectively predict the failure rate of each signal subsystem.

6. The signal fault monitoring method according to claim 5, characterized in that: The setting of the optimal parameters includes: Initialize the number of particles and maximum speed for the PSO algorithm; Fitness evaluation, use the particle position as the weight and threshold of the BP neural network to perform a complete training, and calculate the fitness f of each particle; According to the fitness value f of each particle, update the speed and position of the particle; The weight and threshold corresponding to the particle with the best fitness are selected as the optimal parameters.

7. The signal fault monitoring method according to claim 6, characterized in that: The calculation formula of the fitness function is: Where N is the number of samples, y i is the expected output of the ith sample, is the actual output value of the i-th sample.

8. The signal fault monitoring method according to claim 6, characterized in that: The velocity change of the particle conforms to the following velocity change formula: In the formula, represents the flying speed of the i-th particle in the k+1th generation; w represents the inertia weight; represents the flying speed of the i-th particle in the k-th generation; c1 and c2 represent learning factors; r1 and r2 represent random distribution numbers; represents the best position of the i-th particle to the k-th generation; represents the position of the i-th particle in the k-th generation; Represents the best position of the particle swarm to the kth generation.

9. The signal fault monitoring method according to claim 8, characterized in that: The position change of the particle conforms to the following position change formula: In the formula, represents the position of the i-th particle in the k+1th generation; represents the position of the i-th particle in the k-th generation.

10. The signal fault monitoring method according to claim 2, characterized in that: The step S3 comprises: S3.1, taking the number of safe operation days and failure rate of each signal subsystem as accompanying variables, risk modeling is performed on each signal subsystem; S3.2, verification and optimization, compares the actual failure data with the predicted results after risk modeling.

11. The signal fault monitoring method according to claim 10, characterized in that: The risk function formed by the risk modeling is: Where β is the distribution shape parameter, η is the distribution scale parameter, γ is the distribution location parameter, t is the time point when the fault occurs, and X(t) is the accompanying variable.

12. The signal fault monitoring method according to claim 11, characterized in that: The risk function includes upper and lower control limits; Wherein, the upper control limit is a fault threshold, and when the upper control limit is exceeded, it indicates that the current signal subsystem is unavailable; The lower control limit is a preventive threshold. When it is lower than the lower control limit, it indicates that potential failures begin to occur in the current signal subsystem.

13. The signal fault monitoring method according to claim 1, characterized in that: The multiple signal machine subsystems are: signal light source system, signal control and drive system, mechanical structure system, power supply system, monitoring and maintenance auxiliary system.

14. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the signal fault monitoring method based on Weibull distribution as described in any one of claims 1 to 13.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one program code, and the program code is loaded and executed by a processor to implement the signal fault monitoring method based on Weibull distribution as described in any one of claims 1 to 13.

Images