Railway train operation safety risk early warning interval determination method

Abstract

The invention relates to a railway traffic safety risk early warning interval determination method, which comprises the following steps of: selecting two pieces of feature data from a data set for clustering, and representing the position of each clustering center as a chromosome in a feature space based on a floating-point number encoding mode of the clustering center; each sample is randomly distributed to a certain category to form initial clustering division, then chromosome codes of initial individuals are calculated according to the clustering center of each category, and a first-generation population is constructed; constructing a fitness function; genetic operators are selected for crossover and mutation, and the crossover rate and the mutation rate are adjusted in a self-adaptive mode; after the K mean value operation is carried out, if a preset number of times is reached or exceeded, iteration is terminated; and finally determining a plurality of risk early warning intervals. The risk early warning index historical data is automatically divided and classified, the risk early warning interval is determined according to the boundary of each type, objectivity and dynamics are achieved, and the method can be closer to the real situation.

Description

Technical Field

[0001] The present invention relates to the field of railway safety warning, and in particular to a method for determining a railway driving safety risk warning interval. Background Art

[0002] Risk warning is an important part of accident prevention. It refers to the use of relevant theoretical and technical means to evaluate the degree to which the indicators deviate from the warning limit according to the changing trend of risk warning indicators, so as to take preventive measures in advance. Strengthening the research on railway traffic safety risk warning is of great theoretical and practical significance for scientifically formulating accident control goals and prevention and control decisions, and ensuring railway traffic safety.

[0003] The core content of railway traffic safety risk warning is the determination of risk warning interval; at present, most of the research on risk warning focuses on the establishment of risk warning system, and there is less research on risk warning interval. At present, the risk warning interval is still determined by empirical standards, industry average standards and macroeconomic standards, which are subjective or not dynamic; for my country's high-speed rail system, due to the large regional differences among various railway bureaus, industry standards and macroeconomic standards cannot really be applied to the determination of risk warning intervals for my country's railways. Therefore, in today's my country, which is in a period of rapid railway development, with the railway running speed getting faster and faster and the safety management model constantly changing, how to determine the division of risk warning intervals is a problem that needs to be considered at present. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art, provide a method for determining a railway driving safety risk warning interval, and solve the deficiencies of the prior art.

[0005] The object of the present invention is achieved by the following technical solution: a method for determining a railway driving safety risk warning interval, the method comprising: Step 1: Select two feature data from the railway traffic safety risk warning data set for clustering, and represent the position of each cluster center as a chromosome in the feature space based on the floating point number encoding method of the cluster center; Step 2: Randomly assign each sample to a certain category to form an initial cluster division, and then calculate the chromosome code of the initial individual according to the cluster center of each category, generate multiple initial individuals in total, and construct the first generation population; Step 3: Construct fitness function f =b / (1+E) is used to evaluate the possibility of individuals in the group approaching the optimal solution during the optimization process, b is a constant, and E is the negative value of the silhouette coefficient obtained by K-means clustering; Step 4: Select genetic operators for crossover and mutation, and adaptively adjust the crossover rate and mutation rate; Step 5: After performing the K-means operation, if the current iteration number does not reach the preset maximum number of iterations, the iteration will continue; if it reaches or exceeds the preset number of iterations, the iteration will terminate; Step 6: Finally determine multiple risk warning intervals.

[0006] The step one comprises: Select two feature data from the railway traffic safety risk warning data set, divide the two feature data into three categories: poor, medium and good, and each category is a cluster center; Initialize the cluster center and use the position of the initial cluster center as the code of the first chromosome; The position of each cluster center is represented by coordinates consisting of two floating-point numbers, and the chromosome of the cluster center is encoded as the corresponding coordinates, and each chromosome will contain the coordinates of all cluster centers.

[0007] The second step comprises: Determine the number of chromosomes in the population, and randomly generate gene values ​​for each chromosome, which represent the coordinates of the cluster center; Initialize chromosomes, each chromosome contains a list or array of two floating point numbers; Create an initial population and repeatedly initialize chromosomes until the desired population size is reached.

[0008] The step three comprises: Select two feature data from the railway safety risk warning data set and standardize them; The silhouette coefficient is used to measure the relative position of a sample in its cluster and other clusters. The range of the silhouette coefficient is [-1, 1]. Values ​​close to 1 indicate good clustering effects, and values ​​close to -1 indicate poor clustering effects. f =b / (1+E) to construct the fitness function. Finally, the fitness of each chromosome in each population is calculated by the evalCluster function. This function takes the cluster center of the chromosome as input and returns its fitness value.

[0009] The fourth step comprises: Selection operator: Set the population size to 50, use the roulette wheel selection method, and hope to select 30 chromosomes to enter the next generation. Calculate the fitness of each chromosome, assign a selection probability to each chromosome according to the fitness, and for each new chromosome position, use a random number generator to select the corresponding chromosome according to the probability. The selected chromosome is used for subsequent crossover and mutation operations to generate a new generation of population; Crossover process: select two parent chromosomes, each chromosome consists of a set of cluster centers, and the cluster centers correspond to the two-dimensional coordinates of the two selected feature data. Cross the two parent chromosomes and calculate the offspring chromosomes according to the position of the crossover point. Since the position of the crossover point is random, all offspring chromosomes are weighted combinations of the two parent chromosomes. Add the new offspring chromosomes to the population. Mutation process: randomly select a chromosome from the current population, randomly select one or more sites on the chromosome for mutation, each site represents a specific feature, and use the uniform mutation operator to perform mutation operations on the selected mutation sites; Adaptively adjust the crossover rate and mutation rate: calculate the maximum fitness for the current population f max and the average fitness f avg , for each chromosome in the population, calculate the crossover rate of the chromosome and mutation rate , perform crossover and mutation operations on the chromosome according to the calculated crossover rate and mutation rate.

[0010] The K-means operation includes: In each iteration, the K-means algorithm is used to cluster the two feature data selected from the railway traffic safety risk warning dataset based on the chromosomes of the individuals in the current population, and the data points are divided into K clusters; The silhouette coefficient is used to evaluate the quality of the clustering results. Based on the results of K-means clustering, each chromosome is updated to reflect the new cluster center. In each iteration, the K-means operation is repeated, the chromosome is updated according to the clustering results, and then the fitness evaluation is performed, and the crossover and mutation operations are selected.

[0011] The step six comprises: finally dividing the two feature data selected in the railway traffic safety risk warning data set into three categories: good, medium and poor according to steps one to five, using the upper boundaries of the three categories as the dividing points of the four warning intervals, and determining the four risk warning intervals of each indicator.

[0012] The present invention has the following advantages: a method for determining a railway traffic safety risk warning interval, which automatically divides and classifies historical data of risk warning indicators, determines the risk warning interval according to the boundaries of each category, is objective and dynamic, and is closer to the actual situation; a K-means clustering algorithm based on a genetic algorithm can overcome the local disadvantages of the K-means algorithm, and based on the fitness function, it implements iterative processing of individual structural reorganization within the group by applying genetic operations to group individuals. In this process, group individuals are optimized generation by generation and gradually approach the optimal solution, and ultimately the global optimization of the genetic algorithm is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the risk warning interval of the present invention; Figure 2 It represents the monthly staff violation rate w1 and staff health rate w2 of a railway bureau; Figure 3 A cluster visualization of railway personnel safety risks. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided below in conjunction with the drawings is not intended to limit the scope of protection of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application. The present invention is further described below in conjunction with the drawings.

[0015] The present invention specifically relates to a method for determining railway traffic safety risk warning interval based on genetic algorithm K-means clustering, which is used to define the interval of railway traffic safety risk warning, and aims to solve the problem that the traditional K-means algorithm is prone to fall into local optimality, and specifically includes the following contents: (1) Chromosome encoding; The floating point encoding method based on cluster centers involves representing the position of each cluster center as a chromosome in the feature space. For example, two features are selected in the railway traffic safety risk warning data set: the violation rate of personnel (w1) and the health rate of personnel (w2) to form data points in the two-dimensional space. The steps to encode the centers of different categories in the form of chromosomes are as follows:

[0016] 1) Determine the number of clusters; First, determine the number of categories to be divided, that is, the number of cluster centers. For example, in the process of determining the railway driving safety risk warning interval, the data is divided into three categories: poor, medium, and good, then three cluster centers are required.

[0017] 2) Initialize cluster centers; At the beginning of the genetic algorithm, points in the data set are randomly selected or the positions of the cluster centers are initialized by other methods. The positions of these initial cluster centers will serve as the encoding of the first generation of chromosomes.

[0018] 3) Floating point encoding; The location of each cluster center is represented by two floating point numbers, namely the coordinates of the violation rate (w1) and the health rate (w2). For example, if the coordinates of the first cluster center are (0.2, 0.9), the chromosome encoding of this center is [0.2, 0.9].

[0019] 4) Create chromosomes; Each chromosome will contain the coordinates of all cluster centers. If there are 3 cluster centers, and the coordinates of the cluster centers are: [0.2, 0.9], [0.3, 0.85], [0.5, 0.75], the chromosome can be recorded as [[0.2, 0.9], [0.3,0.85], [0.5, 0.75]].

[0020] (2) Generation of the initial group; 1) Determine the population size; First, decide on the number of individuals (chromosomes) in the population. This number is a parameter that can be set based on the size and complexity of the problem. For example, as a rule of thumb, the initial population size is set to 50, which provides enough diversity without making the computational process too complex.

[0021] 2) Randomly initialize individuals; For each individual, its gene value is randomly generated, which represents the coordinates of the cluster center, that is, the coordinates of the personnel violation rate (w1) and the personnel health rate (w2).

[0022] 3) Create individuals; In Python, use creator.create to define the individual class, and use the ClusterIndividual class to initialize the individual. Each individual is a list or array containing two floating-point numbers.

[0023] 4) Constructing population; In Python, use toolbox.register("population", tools.initRepeat,list, toolbox.individual) to create the initial population. This function will repeatedly initialize individuals until the desired population size is reached.

[0024] (3) Selection of fitness function; First, the data of personnel violation rate (w1) and personnel health rate (w2) are extracted from the railway traffic safety risk warning data set, and standardized to make it suitable for cluster analysis. Then, based on the results of K-means clustering to evaluate the fitness function, the present invention uses the silhouette coefficient to measure the relative position of the sample in its cluster and other clusters. The range of the silhouette coefficient is [-1, 1]. Values ​​close to 1 indicate good clustering effect, and values ​​close to -1 indicate poor clustering effect. Then, use the formula f = b / (1+ E) to construct the fitness function, formula f = b / (1+ E), E is the negative value of the silhouette coefficient obtained by K-means clustering (because the original silhouette coefficient is positive, its negative value is taken to convert to fitness), b is a positive constant (for example, 1), which is used to adjust the scale of the fitness function. Finally, for each individual in the population, in Python, use the evalCluster function to calculate its fitness. This function takes the cluster center of the individual as input and returns its fitness value.

[0025] (4) Genetic operators; 1) Select an operator; Assume that the population size is 50, a roulette wheel selection method is used, and 30 individuals are desired to be selected for the next generation. First, the fitness of each individual is calculated. Then, a selection probability is assigned to each individual according to the fitness. For example, if the sum of the fitness of the individuals is S, and the fitness of individual i is fitness_i, then the selection probability of individual i is fitness_i / S. For the position of each new individual, a random number generator is used to select the corresponding individual according to the probability. The selected individuals will be used in subsequent crossover and mutation operations to generate a new generation of populations. In the process of determining the warning interval for railway traffic safety risks, the selection operator ensures that the algorithm can identify which combinations of cluster centers better divide the risk level, and by retaining these excellent solutions, the algorithm can gradually adjust the warning interval to meet the actual needs of the railway system.

[0026] 2) Crossover process; First, according to the result of the selection operator, two parent individuals with higher fitness are selected from the current population. These individuals represent better cluster centers. Then, one or more crossover points are randomly determined on the chromosomes of the parent individuals. The chromosome here is a sequence consisting of the coordinates of the cluster centers. Next, the genetic information of the two parent individuals is exchanged at the crossover point, thereby generating two new offspring individuals.

[0027] The specific crossover steps are as follows: First, select two parent individuals, each of which consists of a set of cluster centers corresponding to the two-dimensional coordinates of the violation rate (w1) and the health rate (w2). Then, for the two parent individuals and Use formula (3) to perform crossover, where a represents the position of the crossover point. Next, calculate the chromosomes of the offspring individuals based on the value of a. Since a is random, the chromosomes of the offspring individuals will be a weighted combination of the two parent chromosomes. For example, the chromosomes of the parent individuals are as follows: , , randomly select a=0.7, then the newly generated offspring individuals

[0028] Since each coordinate of the offspring individual is obtained by weighted average of the parent coordinates, the coordinates of the first cluster center are as follows: Finally, the newly generated offspring individual Offspring is added to the population, ready for the next step of fitness evaluation.

[0029] 3) Variation process; First, randomly select an individual from the current population, and randomly select one or more sites on its chromosome for mutation. Each site represents a specific feature, such as violation rate (w1) or health rate (w2). Then, use the uniform mutation operator to apply the mutation operation to the selected mutation site. The operation process is as follows: for each selected mutation site, randomly select a value from the legal value range of the gene position to replace the original gene value. The mutated gene value must remain within the legal range, for example, the violation rate and health rate range is [0, 1]. Assume that the chromosome of the individual , randomly select the violation rate (w1) of the first cluster center as the mutation site. The legal value range of the violation rate (w1) is [0, 0.6], and randomly select a new violation rate value is a random number uniformly distributed in the range [0, 0.6]. Replace the original value w 1 i 1, then update the individual chromosome to I new=[( w 1 i 1_ new , w 2 i 1),( w 1 i 2, w 2 i 2)].

[0030] 4) Adaptively adjust the crossover rate and mutation rate; First, for the current population, calculate the maximum fitness f max , the average fitness f avg , then, for each individual in the population, use the formula Calculate the crossover rate of this individual and mutation rate .in, Indicates the fitness value of the individual to be mutated in the population; Indicates the larger fitness value of the two individuals to be crossed. This method sets the corresponding crossover and mutation probabilities for individuals with higher fitness. On the contrary, individuals with lower fitness are given higher crossover and mutation probabilities. Through this adaptive mechanism, the optimal crossover rate can be dynamically adjusted for each solution. and mutation rate , thereby promoting the effective convergence of the algorithm while maintaining population diversity.

[0031] Finally, according to the calculated crossover rate and mutation rate, crossover and mutation operations are performed on individuals. Individuals with higher fitness will have lower mutation rates and higher crossover rates, while individuals with lower fitness will have the opposite. This strategy makes individuals with higher fitness more likely to retain their excellent characteristics, while individuals with lower fitness have a higher probability of mutation, which helps the algorithm to jump out of the local optimum and explore new solution space.

[0032] 5) K-means operation; First, in each iteration of the genetic algorithm, based on the chromosomes of individuals in the current population (representing potential clustering centers), the K-means algorithm is used to cluster the two feature data selected in the railway traffic safety risk warning data set: the violation rate of personnel (w1) and the health rate of personnel (w2), and the data points are divided into K clusters (here K=3, that is, the risk level is good, medium, and poor). Then, the silhouette coefficient is used to evaluate the quality of the clustering results. Next, according to the results of K-means clustering, the chromosome of each individual is updated to reflect the new clustering center. The above process is an iterative optimization process. In each iteration, the K-means operation needs to be repeated, the individual chromosomes need to be updated according to the clustering results, and then the fitness evaluation is performed, and the crossover and mutation operations are selected. Through the K-means operation, the genetic algorithm can use the local search advantage of K-means clustering to quickly converge to a high-quality solution while ensuring the global search capability, so as to more effectively determine the railway traffic safety risk warning interval.

[0033] 6) Iteration termination condition; First, before the algorithm starts, the maximum number of iterations T is set, which is the maximum number of generations allowed by the algorithm to prevent infinite loops. Then, at the beginning of the algorithm, the iteration counter gen is initialized to 0. Secondly, in each iteration, selection, crossover, mutation, and K-means operations are performed, and then the iteration counter gen is updated. Finally, check whether the termination condition of the iteration is met. In the process of determining the railway driving safety risk warning interval, the maximum number of iterations T is set to 100, and a simple fitness stability criterion is used, that is, if the fitness change of 10 consecutive stable generations stable_count is less than the fitness change threshold epsilon, the algorithm is considered to have converged. Here, the initial value of stable_count is set to 0, .

[0034] (5) Determination of safety risk warning interval; like Figure 1 As shown in the figure, the monthly risk warning indicator values ​​of railway enterprises are used as data sets, and each indicator set is standardized. The genetic algorithm-based k-means clustering method is used. According to the above steps (1)-(4), the two feature data selected in the railway traffic safety risk warning data set: personnel violation rate (w1) and personnel health rate (w2) can be finally divided into three categories: good, medium, and poor. The upper boundaries of the three categories are used as the dividing points of the four warning intervals to determine the four risk warning intervals of each indicator.

[0035] The present invention is based on the railway traffic macro-safety risk early warning indicator system, and the personnel safety risk early warning indicator is composed of the personnel violation rate and the personnel health rate. w 1. Personnel health rate w 2 forms a two-dimensional array, and the result is as follows Figure 2 shown.

[0036] The method of the present invention is used to determine the early warning interval of railway personnel safety risks. In jupyter notebook, the data in Table 1 are subjected to a genetic algorithm k-means cluster analysis using Python language and divided into three categories, such as Figure 3 As shown in the figure, cyan represents the first category (poor category), purple represents the second category (medium category), and yellow represents the third category (good category).

[0037] Based on the genetic algorithm k-means clustering, the three cluster centers are determined as [0.06033333 0.96666667], [0.23232558 0.91255814], and [0.4785 0.8625]; the first, second, and third category boundaries are represented in the figure by the coordinates of the farthest points of each category of cluster centers, which are [0.14, 0.94], [0.35, 0.89], and [0.86, 0.84], respectively. Therefore, the three warning intervals of railway personnel safety risks are shown in Table 1.

[0038] Table 1. Three early warning intervals for railway personnel safety risks

[0039] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used for various other combinations, modifications and improvements, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for determining a railway traffic safety risk warning interval, characterized in that: The method comprises: Step 1: Select two feature data from the railway traffic safety risk warning data set for clustering, and represent the position of each cluster center as a chromosome in the feature space based on the floating point number encoding method of the cluster center; Step 2: Randomly assign each sample to a certain category to form an initial cluster division, and then calculate the chromosome code of the initial individual according to the cluster center of each category, generate multiple initial individuals in total, and construct the first generation population; Step 3: Construct fitness function f =b / (1+E) is used to evaluate the possibility of individuals in the group approaching the optimal solution during the optimization process, b is a constant, and E is the negative value of the silhouette coefficient obtained by K-means clustering; Step 4: Select genetic operators for crossover and mutation, and adaptively adjust the crossover rate and mutation rate; Step 5: After performing the K-means operation, if the current iteration number does not reach the preset maximum number of iterations, the iteration will continue; if it reaches or exceeds the preset number of iterations, the iteration will terminate; Step 6: Finally determine multiple risk warning intervals.

2. A method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The step one comprises: Select two feature data from the railway traffic safety risk warning data set, divide the two feature data into three categories: poor, medium and good, and each category is a cluster center; Initialize the cluster center and use the position of the initial cluster center as the code of the first chromosome; The position of each cluster center is represented by coordinates consisting of two floating-point numbers, and the chromosome of the cluster center is encoded as the corresponding coordinates, and each chromosome will contain the coordinates of all cluster centers.

3. The method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The second step comprises: Determine the number of chromosomes in the population, and randomly generate gene values ​​for each chromosome, which represent the coordinates of the cluster center; Initialize chromosomes, each chromosome contains a list or array of two floating point numbers; Create an initial population and repeatedly initialize chromosomes until the desired population size is reached.

4. The method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The step three comprises: Select two feature data from the railway safety risk warning data set and standardize them; The silhouette coefficient is used to measure the relative position of a sample in its cluster and other clusters. The range of the silhouette coefficient is [-1, 1]. Values ​​close to 1 indicate good clustering effects, and values ​​close to -1 indicate poor clustering effects. f =b / (1+E) to construct the fitness function. Finally, the fitness of each chromosome in each population is calculated by the evalCluster function. This function takes the cluster center of the chromosome as input and returns its fitness value.

5. The method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The fourth step comprises: Selection operator: Set the population size to 50, use the roulette wheel selection method, and hope to select 30 chromosomes to enter the next generation. Calculate the fitness of each chromosome, assign a selection probability to each chromosome according to the fitness, and for each new chromosome position, use a random number generator to select the corresponding chromosome according to the probability. The selected chromosome is used for subsequent crossover and mutation operations to generate a new generation of population; Crossover process: select two parent chromosomes, each chromosome consists of a set of cluster centers, and the cluster centers correspond to the two-dimensional coordinates of the two selected feature data. Cross the two parent chromosomes and calculate the offspring chromosomes according to the position of the crossover point. Since the position of the crossover point is random, all offspring chromosomes are weighted combinations of the two parent chromosomes. Add the new offspring chromosomes to the population. Mutation process: randomly select a chromosome from the current population, randomly select one or more sites on the chromosome for mutation, each site represents a specific feature, and use the uniform mutation operator to perform mutation operations on the selected mutation sites; Adaptively adjust the crossover rate and mutation rate: calculate the maximum fitness for the current population f max and the average fitness f avg , for each chromosome in the population, calculate the crossover rate of the chromosome and mutation rate , perform crossover and mutation operations on the chromosome according to the calculated crossover rate and mutation rate.

6. A method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The K-means operation includes: In each iteration, the K-means algorithm is used to cluster the two feature data selected from the railway traffic safety risk warning dataset based on the chromosomes of the individuals in the current population, and the data points are divided into K clusters; The silhouette coefficient is used to evaluate the quality of the clustering results. Based on the results of K-means clustering, each chromosome is updated to reflect the new cluster center. In each iteration, the K-means operation is repeated, the chromosome is updated according to the clustering results, and then the fitness evaluation is performed, and the crossover and mutation operations are selected.

7. A method for determining a railway traffic safety risk warning interval according to claim 1, characterized in that: The step six comprises: finally dividing the two feature data selected in the railway traffic safety risk warning data set into three categories: good, medium and poor according to steps one to five, using the upper boundaries of the three categories as the dividing points of the four warning intervals, and determining the four risk warning intervals of each indicator.

Images