False alarm recognition network model system, false alarm recognition network model training method, and automatic false alarm filtering method

Abstract

A false alarm recognition network model system, a false alarm recognition network model training method, and an automatic false alarm filtering method. The system comprises a false alarm recognition network model, and relates to the technical field of intelligent operation and maintenance of rail transit; the false alarm recognition network model comprises an input layer, a hidden layer and an output layer; the input layer is used for receiving quantized alarm conditions; the hidden layer is an IPSO-LSTM neural network after IPSO optimization; and the output layer is used for outputting a result obtained by training by means of the IPSO-LSTM neural network. The present application further provides a device and a readable storage medium. According to the false alarm recognition network model system and the false alarm recognition network model training method of the present application, the IPSO-LSTM neural network after IPSO optimization is constructed to serve as the false alarm recognition network model to recognize alarm information, in the automatic false alarm filtering method, the false alarm recognition network model is utilized to recognize the alarm information and automatically filter a false alarm, the recognition / filtering process is rapid, and the result is accurate, thereby significantly improving the credibility of the result of an early warning system, and facilitating the normal exertion of the function of the early warning system.

Description

False alarm identification network model system, training method, and false alarm automatic filtering method

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on January 12, 2024, with application number 202410050444.X and invention name “False Alarm Identification Network Model System, Training Method, and False Alarm Automatic Filtering Method”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of intelligent operation and maintenance technology for rail transit, and in particular to a false alarm identification network model system, a training method, and an automatic false alarm filtering method. Background Art

[0003] Currently, the large volume of early warning alarm data in the full-lifecycle intelligent operation and maintenance systems used in subways, coupled with factors such as line factors, can lead to false alarms. These false alarms directly result in ineffective repairs, increasing product re-inspection pass rates, reducing availability and increasing field maintenance costs. Furthermore, a high false alarm rate in early warning systems significantly reduces the credibility of their results. In practice, due to a lack of available solutions and experience for addressing false alarms in intelligent operation and maintenance early warning systems, contractors have been forced to significantly reduce the functionality of early warning systems in the later stages of development, setting very high warning fault thresholds or directly blocking fault results in the software. While these erroneous practices can reduce the false alarm rate in intelligent operation and maintenance early warning systems, they also severely limit their effectiveness.

[0004] How to achieve rapid and accurate identification and filtering of false alarms is a technical problem that needs to be solved urgently by those skilled in the art.

[0005] Summary of the Invention

[0006] To solve the above technical problems, the purpose of the present invention is to provide a false alarm identification network model system, a training method, and a false alarm automatic filtering method for quickly and accurately identifying / filtering false alarms.

[0007] The technical solutions provided by the present invention are as follows:

[0008] A false alarm identification network model system includes a false alarm identification network model, wherein the false alarm identification network model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive quantized alarm conditions, the hidden layer is an IPSO-LSTM neural network optimized by IPSO, and the output layer is used to output results trained by the IPSO-LSTM neural network.

[0009] Preferably, the number of hidden layers is 2.

[0010] Preferably, the IPSO-LSTM neural network after IPSO optimization is obtained by optimizing the number of iterations, learning rate and number of neurons in the hidden layer of LSTM using IPSO.

[0011] A false alarm recognition network model training method, comprising:

[0012] S1. Obtain various quantitative alarm conditions;

[0013] S2. Various quantified alarm conditions are formed into a training set as input vectors, and the LSTM neural network is optimized using the IPSO algorithm to form an optimized IPSO-LSTM neural network;

[0014] S3. The training set is fed into the optimized IPSO-LSTM neural network for training, and the test set formed by various quantified alarm conditions is fed into the trained IPSO-LSTM neural network to obtain output parameters.

[0015] S4. Analyze the output parameters;

[0016] S5. If there is an increase in alarm conditions, the increased alarm conditions are quantified to increase the number of training samples, and steps S2 to S4 are repeated to continuously train the network.

[0017] Preferably, in step S1, the quantitative representation of the alarm situation includes: a normal alarm is marked as 1, a false alarm is marked as 0XXX, and 0XXX is a classification number starting from 0001.

[0018] Preferably, in step S2, optimizing the LSTM neural network using the IPSO algorithm includes: optimizing the number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer using IPSO to obtain the optimal number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer.

[0019] Preferably, in step S2, the use of IPSO to optimize the number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer to obtain the optimal number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer includes:

[0020] S21, determine the fitness threshold ω, the maximum allowed number of iterations G max , c1=c2=1.495, search range[-x max ,x max ]、Maximum speed v max , and determine the number of particles according to the network specifications, where (x i=(θ,β,α1,α2), where θ represents the number of LSTM iterations, β represents the learning rate, α1 represents the number of neurons in the first hidden layer, and α2 represents the number of neurons in the second hidden layer);

[0021] S22. Randomly generate a finite number of individuals x according to the size of the particle group. i A whole population is formed, where different individuals have different weights in a set of neural networks;

[0022] S23, calculating the fitness value of each particle;

[0023] S24. Compare fitness and determine the current best fitness L of each particle best And the global best fitness P best And determine the inertia weight w;

[0024] S25, if P best <ω or the number of runs is greater than G max , then stop, and the position of the particle is what we want;

[0025] S26, update the position and velocity of each particle and consider whether they are within the specified range. If v ij (t+1)<-v max , then v ij (t+1)=-v max If v ij (t+1)>v max , then v ij (t+1)=v max If x ij (t+1)<-x max , then x ij (t+1)=-x max If x ij (t+1)>x max , then x ij (t+1)=x max ;

[0026] S27. Return to step S23.

[0027] A false alarm automatic filtering method includes: obtaining alarm information; bringing the alarm information into the trained false alarm recognition network model of the false alarm recognition network model system as described above, identifying whether it is a false alarm, and filtering out the alarm information if it is a false alarm.

[0028] A device comprises: a memory for storing a computer program; and a processor for executing the computer program. The computer program includes the false alarm identification network model system described above, or the processor implements the false alarm identification network model training method described above, or the false alarm automatic filtering method described above when executing the computer program.

[0029] A readable storage medium stores a computer program, wherein the computer program includes the false alarm identification network model system described above, or when executed by a processor, implements the false alarm identification network model training method described above, or the false alarm automatic filtering method described above.

[0030] Compared with the existing technology, the false alarm identification network model system and training method of the present invention identify alarm information by constructing the IPSO-LSTM neural network optimized by IPSO as the false alarm identification network model. The false alarm automatic filtering method uses the false alarm identification network model to identify alarm information and automatically filters false alarms. The identification / filtering process is fast and the results are accurate, which can significantly improve the credibility of the early warning system results and is conducive to the normal functioning of the early warning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] FIG1 is a schematic diagram of the neuron structure of an LSTM neural network in a false alarm identification network model system according to an embodiment of the present invention;

[0033] FIG2 is a schematic diagram of a particle encoding format of an IPSO algorithm in the false alarm identification network model system shown in FIG1 ;

[0034] FIG3 is a flow chart of the IPSO algorithm in the false alarm identification network model system shown in FIG1 ;

[0035] FIG4 is a flow chart of a false alarm identification network model training method according to an embodiment of the present invention shown in FIG1 ;

[0036] FIG5 is a flow chart of the method for automatically filtering false alarms according to the embodiment of the present invention shown in FIG1 . DETAILED DESCRIPTION

[0037] In order to help those skilled in the art better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of this application.

[0038] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described below are merely illustrative. For example, the division of units and modules is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or modules can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0039] In addition, all functional units in the embodiments of the present application may be integrated into one processor, or each unit may be a separate device, or two or more units may be integrated into one device; each functional unit in the embodiments of the present application may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0040] Those skilled in the art will understand that all or part of the steps of the following method embodiments can be implemented by program instructions and related hardware. The aforementioned program instructions can be stored in a computer-readable storage medium. When the program instructions are executed, the steps of the following method embodiments are executed; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, and other media that can store program codes.

[0041] It should be understood that the use of "system," "device," "unit," and / or "module" in this application is merely a method for distinguishing different units, elements, components, parts, or assemblies at different levels. However, if other words can achieve the same purpose, they can be replaced by other expressions.

[0042] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout the description of this application, "plurality" or "several" means two or more, unless otherwise specifically defined.

[0043] If a flow chart is used in this application, the flow chart is used to illustrate the operations performed by the system according to the embodiments of the application. It should be understood that the previous or subsequent operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. At the same time, other operations can be added to these processes, or one or more operations can be removed from these processes.

[0044] It should also be noted that, in this document, terms such as "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that an article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such article or device. In the absence of further limitations, elements defined by the phrase "comprises a ..." do not exclude the presence of other identical elements in the article or device comprising the above elements.

[0045] An embodiment of the present invention provides a false alarm identification network model system, including a false alarm identification network model, wherein the false alarm identification network model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive quantized alarm conditions, the hidden layer is an LSTM (long short-term memory) neural network (hereinafter referred to as an IPSO-LSTM neural network) optimized by an improved particle swarm optimization algorithm (IPSO), and the output layer is used to output results trained by the IPSO-LSTM neural network.

[0046] In this embodiment, the number of hidden layers is 2.

[0047] As shown in Figure 1, the LSTM neural network uses two gates to control the content of the cell state c, one is the forget gate and the other is the input gate.

[0048] The forget gate formula is: f t =σ(W f ·[h t-1 ,x t ]+b f ); (1)

[0049] Where W f is the weight matrix of the forget gate; h t-1 is the implicit state at time t-1; xt is the input vector at time t; b f is the bias term of the forget gate; σ is the sigmoid function.

[0050] The input gate formula is: i t =σ(W i ·[h t-1 ,x t ]+b i ); (2)

[0051] Where W i is the weight matrix of the forget gate; b i It is the bias term of the forget gate; it is used to describe the unit state at the current moment.

[0052] The output gate is responsible for controlling the influence of long-term memory on the current output.

[0053] The output gate formula is: t =σ(W o ·[h t-1 ,x t ]+b o ); (3)

[0054] The final output of LSTM is determined by the output gate and the unit state. The formula is:

[0055] In the formula, the symbol Represents element-wise multiplication.

[0056] The implementation steps of the IPSO algorithm are as follows:

[0057] The particle swarm optimization algorithm was implemented in MATLAB. The particles in the swarm and their velocities are encoded using real numbers. The particle encoding format is shown in Figure 2. Here, dimSize represents the parameter dimension. The particle swarm encoding format is shown in Equation 5, where popSize represents the total swarm size.

[0058] In the MATLAB environment, after the particle swarm optimization algorithm main program runs, it returns the optimal solution and its corresponding fitness. Furthermore, to evaluate algorithm performance, it also returns tracking information for each generation. This tracking information is encoded using the structure shown in Equation 6, where iterMax represents the maximum number of iterations and T represents the number of iterations.

[0059] The steps of the global particle swarm optimization algorithm are shown in Figure 3.

[0060] In this embodiment, the IPSO-LSTM neural network after IPSO optimization is obtained by optimizing the number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer using IPSO.

[0061] Although the LSTM model has obvious advantages, its parameters are difficult to determine, and the selection of different parameters has a great impact on the prediction results. Therefore, this embodiment uses IPSO to optimize the number of LSTM iterations, learning rate, and number of neurons in the hidden layer to make up for the shortcomings of LSTM with the help of the IPSO algorithm.

[0062] As shown in FIG4 , this embodiment further provides a false alarm recognition network model training method, comprising the following steps:

[0063] S1. Obtain various quantitative alarm conditions;

[0064] S2. Various quantified alarm conditions are formed into a training set as input vectors, and the LSTM neural network is optimized using the IPSO algorithm to form an optimized IPSO-LSTM neural network;

[0065] S3. The training set is fed into the optimized IPSO-LSTM neural network for training, and the test set formed by various quantified alarm conditions is fed into the trained IPSO-LSTM neural network to obtain output parameters.

[0066] S4. Analyze the output parameters;

[0067] S5. If there is an increase in alarm conditions, the increased alarm conditions are quantified to increase the number of training samples, and steps S2 to S4 are repeated to continuously train the network.

[0068] In this embodiment, in the acquisition of various types of quantitative alarm conditions, the quantitative representation of the alarm conditions includes: a normal alarm is marked as 1, a false alarm is marked as 0XXX, and 0XXX is a classification number starting from 0001.

[0069] For example, there are 6 alarms in the system. After manual judgment, 1 is a normal alarm (marked as 1) and 4 are false alarms. The specific false alarms are as follows:

[0070] 1. Vehicle fault - false reporting of fault level - falsely reporting a minor fault as a major fault and marking the alarm as (0001);

[0071] 2. Vehicle fault - Fault code false reporting - There is no fault at this location but a door fault is reported, and the alarm is marked as (0002);

[0072] 3. Vehicle fault - Fault system false alarm - The PIS system has no fault but reports a PIS system display fault, and the alarm is marked as (0003);

[0073] 4. Vehicle fault - if the fault lasts for a short time and no alarm is reported afterwards, the alarm will be marked as (0004);

[0074] 5. Vehicle failure - the time difference between the two failures is very short - mark this alarm as (0005).

[0075] In this embodiment, in step S2, optimizing the LSTM neural network using the IPSO algorithm includes: optimizing the number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer using IPSO to obtain the optimal number of LSTM iterations, the learning rate, and the number of neurons in the hidden layer.

[0076] Specifically, step S2 includes the following steps:

[0077] S21, determine the fitness threshold ω, the maximum allowed number of iterations G max , c1=c2=1.495, search range[-x max ,x max ]、Maximum speed v max , and determine the number of particles according to the network specifications, where (x i =(θ,β,α1,α2), where θ represents the number of LSTM iterations, β represents the learning rate, α1 represents the number of neurons in the first hidden layer, and α2 represents the number of neurons in the second hidden layer);

[0078] S22. Randomly generate a finite number of individuals x according to the size of the particle group. i A whole population is formed, where different individuals have different weights in a set of neural networks;

[0079] S23, calculating the fitness value of each particle;

[0080] S24. Compare fitness and determine the current best fitness L of each particle best And the global best fitness P best And determine the inertia weight w;

[0081] S25, if P best <ω or the number of runs is greater than G max , then stop, and the position of the particle is what we want;

[0082] S26, update the position and velocity of each particle and consider whether they are within the specified range. If v ij (t+1)<-v max , then v ij (t+1)=-vmax If v ij (t+1)>v max , then v ij (t+1)=v max If x ij (t+1)<-x max , then x ij (t+1)=-x max If x ij (t+1)>x max , then x ij (t+1)=x max ;

[0083] S27. Return to step S23.

[0084] As shown in FIG5 , this embodiment further provides a method for automatically filtering false alarms, including: when the early warning system alarms, obtaining alarm information; bringing the alarm information into the trained false alarm recognition network model of the false alarm recognition network model system as described above for analysis to identify whether it is a false alarm, and if so, filtering out the alarm information.

[0085] This embodiment also provides a device, including: a memory for storing a computer program; and a processor for executing the computer program. The computer program includes the false alarm identification network model system described above, or the processor implements the false alarm identification network model training method described above, or the false alarm automatic filtering method described above when executing the computer program.

[0086] This embodiment also provides a readable storage medium, on which a computer program is stored. The computer program includes the false alarm identification network model system described above, or, when executed by a processor, implements the false alarm identification network model training method described above, or the false alarm automatic filtering method described above.

[0087] This embodiment numerically quantifies various types of evidence, specifically alarms. The set of alarms increases over time as the system runs and as human intervention is performed. Each time an alarm is added, the entire set is fed into the IPSO-LSTM neural network for training. This dynamic training process improves the accuracy of false alarm detection as the training set grows. With sufficient training samples, the intelligent operation and maintenance early warning system can automatically filter out false alarms.

[0088] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A false alarm recognition network model system, characterized in that, It includes a false alarm recognition network model, which includes an input layer, a hidden layer, and an output layer. The input layer is used to receive the quantified alarm situation. The hidden layer is an IPSO-optimized IPSO-LSTM neural network. The output layer is used to output the result after being trained by the IPSO-LSTM neural network.

2. The false alarm recognition network model system according to claim 1, wherein The number of hidden layers is 2.

3. The false alarm recognition network model system according to claim 1, wherein The IPSO-optimized IPSO-LSTM neural network is obtained by using IPSO to optimize the number of iterations, learning rate, and the number of neurons in the hidden layer of LSTM.

4. A method for training a false alarm recognition network model, characterized in that, It includes: S1. Obtain various quantified alarm situations; S2. Form a training set from various quantified alarm situations as input vectors, and use the IPSO algorithm to optimize the LSTM neural network to form an optimized IPSO-LSTM neural network; S3. Bring the training set into the optimized IPSO-LSTM neural network for training, and bring the test set formed by various quantified alarm situations into the trained IPSO-LSTM neural network to obtain output parameters; S4. Analyze the output parameters; S5. If there is an increase in alarm situations, quantify the increased alarm situations so that the training samples increase, and then repeat steps S2 to S4 to continuously train the network.

5. The method for training a false alarm recognition network model according to claim 4, wherein In step S1, the quantified representation of the alarm situation includes: normal alarm is marked as 1, false alarm is marked as 0XXX, where 0XXX is the classification number, starting from 0001 and numbered sequentially.

6. The false alarm recognition network model training method according to claim 4, wherein, In step S2, using the IPSO algorithm to optimize the LSTM neural network includes: using IPSO to optimize the number of iterations, learning rate, and the number of neurons in the hidden layer of LSTM to obtain the optimal number of iterations, learning rate, and the number of neurons in the hidden layer of LSTM.

7. The method for training a false alarm recognition network model according to claim 6, wherein In step S2, the process of using IPSO to optimize the number of iterations, learning rate, and the number of neurons in the hidden layer of LSTM to obtain the optimal number of iterations, learning rate, and the number of neurons in the hidden layer of LSTM includes: S21. Determine the fitness threshold ω and the maximum allowable number of iterative steps G max , c1 = c2 = 1.495, search range [-x max , x max , maximum speed v max , and determine the number of particles according to the network specifications, where (x i = (θ, β, α1, α2), θ represents the number of iterations of LSTM, β represents the learning rate, α1 represents the number of neurons in the first hidden layer, and α2 represents the number of neurons in the second hidden layer); S22. Randomly generate a finite number of individuals x according to the size of the particle group i to form the entire population, where different individuals have different weights in a set of neural networks; S23. Calculate the fitness value of each particle; S24. Compare the fitness values to determine the current best fitness value L of each particle best and the global best fitness value P best and determine the inertia weight w; S25. If P best <ω or the number of running times is greater than G max , then stop, and the position of this particle is the one sought; S26. Update the positions and velocities of each particle and consider whether they are within the defined range. If v ij (t + 1) < -v max , then v ij (t + 1) = -v max ; if v ij (t + 1) > v max , then v ij (t + 1) = v max ; if x ij (t + 1) < -x max , Then x ij (t + 1)= -x max ; If x ij (t + 1)>x max , then x ij (t + 1)=x max ; S27. Return to step S23.

8. An automatic false alarm filtering method, characterized in that It includes: Obtain the alarm information; bring the alarm information into the false alarm recognition network model in the trained false alarm recognition network model system as claimed in any one of claims 1 to 3 to identify whether it is a false alarm. If so, filter out the alarm information.

9. A device, characterized in that, It includes: A memory for storing computer programs; A processor for executing the computer program, where the computer program includes the false alarm recognition network model system as claimed in any one of claims 1 to 3, or when the processor executes the computer program, it implements the false alarm recognition network model training method as claimed in any one of claims 4 to 7, or the false alarm automatic filtering method as claimed in claim 8.

10. A readable storage medium, characterized in that: The computer program is stored on the readable storage medium, and the computer program includes the false alarm recognition network model system as claimed in any one of claims 1 to 3, or when executed by the processor, it implements the false alarm recognition network model training method as claimed in any one of claims 4 to 7, or the false alarm automatic filtering method as claimed in claim 8.

Images