A Method and Equipment for Urban Rail Transit Train Condition Assessment Based on TOPSIS-Markov Joint Model

Abstract

This invention discloses a method and device for urban rail train condition assessment based on a TOPSIS-Markov joint model. The method includes data acquisition and processing of access subsystems, evaluation of the state of each subsystem based on the TOPSIS comprehensive evaluation method, correction of the state transition probability matrix based on the TOPSIS comprehensive evaluation method, real-time subsystem state assessment, and construction of the overall vehicle state transition matrix based on the Markov model to obtain the overall vehicle state assessment. This invention aims to effectively reduce urban rail train operation and maintenance costs, improve equipment reliability, extend train service life, and ensure train operation safety.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit technology, specifically to a method and device for evaluating the condition of urban rail trains based on the TOPSIS-Markov joint model. Background Technology

[0002] With the rapid development of urban rail transit, the number of rail transit vehicles continues to grow. As an indispensable means of transportation in modern society, the operational safety and reliability of trains have become a core concern for the industry. However, traditional train condition monitoring mainly relies on manual inspections, which suffers from low efficiency, a high risk of missed inspections, and high workload. Therefore, there is an urgent need for a method that can acquire real-time or near-real-time information on the abnormal status of subsystems and the overall operation of trains based on multi-dimensional detection data during train operation. This would effectively reduce the cost of manual inspections, extend the service life of trains, and ensure operational safety. Summary of the Invention

[0003] The present invention proposes a method and device for evaluating the condition of urban rail trains based on the TOPSIS-Markov joint model, which can at least solve one of the technical problems in the background art.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for evaluating the state of urban rail transit trains based on a TOPSIS-Markov joint model includes the following steps:

[0006] S1. Collect real-time operating data, historical fault data, and historical maintenance data of each train and its subsystems, and process the data.

[0007] S2. Based on historical fault data and maintenance cost data of each subsystem, extract equipment operating cost and benefit-type indicator data, and combine them with TOPSIS comprehensive evaluation to obtain discrete state labels and proximity of each subsystem. ;

[0008] S3. Based on the TOPSIS comprehensive evaluation in S2, obtain the discrete state labels and proximity of each subsystem. The state transition matrix of each subsystem is corrected, and the health index weight, time decay weight, and state label correction factor are extracted to measure the future health status of each subsystem.

[0009] S4. The overall vehicle evaluation status is output based on the transition probabilities of each subsystem combined with the Markov chain.

[0010] Furthermore, the acquisition and processing of subsystem data:

[0011] Real-time operating data, historical fault data, and historical maintenance data of the bogie system's vibration / temperature / geometric characteristics / dynamic indicators, traction system's current characteristics / temperature characteristics / insulation characteristics / efficiency indicators, braking system's hydraulic characteristics / wear indicators / response performance / temperature field, and auxiliary power supply system's output quality / battery health / temperature characteristics / efficiency indicators are acquired through on-site installed sensors and image acquisition equipment.

[0012] The next step is to preprocess the collected real-time operation data and historical fault data, including text, image, video and audio data, to extract the data features of the equipment during operation, including time domain / frequency domain features, and to standardize the feature data.

[0013] Further, based on historical fault data and maintenance cost data of each subsystem, equipment operating cost and benefit-type indicator data are extracted.

[0014] Evaluate the state of each subsystem based on the TOPSIS method.

[0015] Based on the above, the indicator-type and benefit-type data in the fault data and maintenance cost data are standardized, and the weighted normalization matrix of the original data matrix is ​​combined with the positive and negative ideal solutions to calculate the calculation of each subsystem. The steps for determining proximity are as follows: Step 1: Extract n evaluation index data from m samples and set up a decision matrix. The normalized matrix is The weighted normalization matrix is ,in It's the weight.

[0016] in Sample size Number of evaluation indicators :No. One sample / observation, :No. One sample / observation, :No. The first sample The actual measured value of each indicator : Normalized index value :No. The weight of each indicator, : Relative health contribution value after indicator importance :No. The weight of each indicator.

[0017] Step 2: Determine the application of positive and negative ideal solution indices

[0018] For efficiency-type indicators: the ideal solution Negative ideal solution ;

[0019] For cost-related indicators: the ideal solution Negative ideal solution ;

[0020] Step 3: Calculate the distance from each sample to the positive ideal PIS and the negative ideal NIS, as follows:

[0021] To the ideal distance: Distance to negative ideal: .

[0022] in: Euclidean distance calculation.

[0023] Step 4: Based on the methods in Steps 1 to 3 above, calculate the relative proximity of each subsystem in turn.

[0024]

[0025] in: To the ideal distance, The distance to the negative ideal. This indicates that the state is close to optimal. 0 indicates that the state is close to the worst.

[0026] Step 5: For each subsystem The membership degrees of the four states are calculated using the trapezoidal membership function. , , , After normalization, the probability distribution is obtained. .

[0027] Let the membership function threshold be... =0.8, =0.5, =0.3, and the state probability distribution is as follows:

[0028] (healthy)

[0029] (good)

[0030] (Warning)

[0031] (Fault)

[0032] Normalized probability distribution:

[0033]

[0034] The state transition probability matrix of each subsystem is constructed based on the discrete state labels and C_i and the closeness of the health status of each subsystem obtained by the TOPSIS comprehensive evaluation in the second step. The state transition matrix of each subsystem is corrected, and the health index weight, time decay weight, and state label correction factor are extracted to predict the future health status of each subsystem.

[0035] During operation, the train bogie system is subjected to load conditions considering mechanical vibration and geometric deformation characteristics (such as axle temperature trends and wheelset wear). Among these, mechanical fatigue wear is the dominant state factor. Based on the discrete state labels and health index values ​​obtained from the TOPSIS comprehensive evaluation, the bogie state transition probability formula is as follows:

[0036]

[0037] : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t;

[0038] :Transition probabilities normalized based on fatigue and wear; State transition sensitivity coefficient;

[0039] Cumulative fatigue level; Health index; Load sensitivity coefficient; Load factor;

[0040] The train traction system needs to pay attention to IGBT switching loss rate, motor three-phase imbalance, power device degradation, and the probability of equipment state transition under different loads and weather conditions. Based on the TOPSIS comprehensive evaluation, the state label and health index value are extracted and the state probability transition formula is output as follows.

[0041]

[0042] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized transition probability. : Health correction function, where Indicates the health decay coefficient. It indicates a health index. Weather factors. Loading factor. : Switching loss function, where Loss sensitivity coefficient For the rate of loss exceeding the standard, This represents the material sensitivity coefficient. Health amplification factor. Three-phase unbalance function, where For system sensitivity, The thermal effect follows a quadratic law. Health vulnerability.

[0043] The train braking system uses brake pad thickness and brake pressure response delay parameters as the basis for condition assessment. Brake pad thickness is one of the important indicators for condition classification. When brake pad wear reaches a certain threshold, it will directly lead to a degrade of the braking system condition. However, rain and snow, mileage, and brake pad temperature are all factors that contribute to brake pad wear during train operation. The following is a state probability transition formula derived by combining the TOPSIS comprehensive evaluation to extract state labels and health index values:

[0044]

[0045] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized transition probability; :state The wear sensitivity coefficient, Current brake pad thickness state Ideal thickness Temperature compensation factor (deterioration accelerates at high temperatures).

[0046] The train auxiliary power supply system mainly monitors output voltage ripple and capacitor ESR changes. Power module reliability is also a key indicator. Based on the TOPSIS comprehensive evaluation, state labels and health indices are extracted, and the state probability transition formula is as follows:

[0047]

[0048] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized basic transition probability; :Basic factors, among which basic factors (Ripple Factor) (ESR factor) (Temperature factor) (Aging factors); Forced jump variable (increases the probability of "mutation failure" when health deteriorates), where Health deviation function when hour Time-suppressed jump, when hour Promotes leaps in performance. Kronecker function When =1, the target state is S4. =0 indicates other states.

[0049] The ripple factor included in the basic factors of the auxiliary power supply system as described above ESR factor Temperature factor aging factors The calculation method is as follows: Ripple factor: In the formula Peak ripple voltage, Indicates the rated output voltage. This represents the ripple sensitivity coefficient.

[0050] ESR factor: The denominator of the health index in the formula reflects that "the worse the health status, the greater the impact on ESR". The EST sensitivity coefficient This is the current ESR value. This is the initial ESR value.

[0051] Temperature factor: In the formula Component junction temperature, Reference temperature This is the temperature coefficient. The formula quantifies the halving of lifespan for every 20-degree increase in temperature.

[0052] Aging factors: = ( In the formula Indicates running time (hours) Indicates the design life. This represents the aging coefficient.

[0053] Construct the vehicle state transition matrix;

[0054] The transition states of each subsystem are obtained using... Construct the transition matrix of the entire vehicle. Define the bogie system state matrix as follows: The state transition matrix of the traction system is The state transition matrix of the braking system is The auxiliary power supply system is The state transition matrix of the whole vehicle system product;

[0055]

[0056] The order of status coding is defined as: bogie, traction, braking, auxiliary power supply;

[0057] The first step is to calculate the two subsystems. Product, such as first calculating the joint matrix of the bogie and traction system: ;

[0058] The second step involves comparing the above results with the transition matrix of the braking system. product: The third step is to perform a transfer matrix interaction with the auxiliary power supply system. product: ; each step The rules for calculating the product are as follows: For two matrices and their product It is Block matrix:

[0059]

[0060] Therefore, for the four subsystems, each with four states, the size of the vehicle's state transition matrix is: In the vehicle state transition matrix, each element represents the probability of transitioning from one combination state to another.

[0061] Based on the above iterative calculation using the transition matrix of the subsystem, the state distribution of the entire vehicle can be represented as the state distribution of each subsystem. Product, calculate the vehicle state transition matrix: Let but It is The elements of the matrix can be determined as follows: Represent the overall vehicle state using a quadruple (i, j, k, l), where i, j, k, and l are integers from 0 to 3 (corresponding to 4 states). Map the quadruple to row or column indices, usually in lexicographical order (i.e., I changes the fastest, then k, then j, and i changes the slowest).

[0062] index Therefore, in the transition matrix, the transition probability from state (i, j, k, l) to state (i', j', k', l') is: .

[0063] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0064] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0065] As described above, the urban rail train state evaluation method based on the TOPSIS-Markov joint model of the present invention includes the acquisition and processing of data from access subsystems, evaluation of the state of each subsystem based on the TOPSIS comprehensive evaluation method, correction of the state transition probability matrix based on the TOPSIS comprehensive evaluation method, real-time state evaluation of subsystems, construction of the whole vehicle state transition matrix based on the Markov model, and obtaining the whole vehicle state evaluation. The present invention aims to effectively reduce the operation and maintenance costs of urban rail trains, improve equipment reliability, extend train service life, and ensure train operation safety.

[0066] Specifically, traditional train operation and maintenance relies on threshold alarms or simple weighting / scoring of multiple indicators. With fixed indicator weights, it cannot scientifically handle conflicts and differences in dimensions between indicators, nor can it reflect the transition of critical failure modes during equipment aging. The output is typically a binary judgment of "normal / abnormal" or a single, definitive status label, losing information about the uncertainties in the evaluation process. The original Markov model uses a static transition matrix, which cannot respond to changes in load, environment, etc., making it difficult to assess the systemic risk of subsystem coupling failures to the entire vehicle.

[0067] This invention is based on the TOPSIS-Markov joint model. By incorporating multi-source heterogeneous data fusion technology, it introduces the multi-attribute decision framework of TOPSIS, eliminates dimensions through vector normalization, defines benefit / cost indicators and positive and negative ideal solutions, unifies the optimization direction with clear physical meaning, and outputs a comprehensive health index, thereby achieving lossless compression and scientific aggregation of multi-dimensional information.

[0068] The real-time health index output by TOPSIS By dynamically adjusting the Markov transition matrix with discrete state labels as correction factors, fragmented alarm signals are transformed into a global, interpretable quantitative indicator of "health," enabling the evaluation model to have "lifespan awareness" capabilities. The evaluation results are more aligned with actual engineering applications, eliminating missed and false alarms during operation and maintenance. This upgrades equipment health assessment from "deterministic" diagnosis to "probabilistic prediction," transforming the prediction model from "average patterns" to "individualized real-time patterns." This significantly improves the accuracy and sensitivity of short-term predictions, achieving a leap from train equipment health management to system-level risk control. It can calculate the overall risk probability of "train downtime due to failure of any critical subsystem," achieving true dynamic prediction. This fundamentally shifts from experience-based "timely maintenance" and "retroactive repair" to data- and model-based "predictive intervention." While ensuring safety, it provides a solid technical path to optimize operation and maintenance costs, quantifying the probability of different risk levels and providing more refined basis for operation and maintenance decisions. Attached Figure Description

[0069] Figure 1 This is a data processing flowchart of an embodiment of the present invention;

[0070] Figure 2 This is a flowchart illustrating the process of moving objects according to an embodiment of the present invention;

[0071] Figure 3 This is a flowchart illustrating the system framework of an embodiment of the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0073] like Figure 1 , Figure 2 , Figure 3 As shown, an embodiment of the present invention is a train health status assessment method based on a TOPSIS-Markov joint model, comprising the following steps:

[0074] Step 1: Acquire data from each train's subsystems: Collect real-time operating data, historical fault data, and historical maintenance data for the bogie system (vibration / temperature / geometric characteristics / dynamic indicators), traction system (current characteristics / temperature characteristics / insulation characteristics / efficiency indicators), braking system (hydraulic characteristics / wear indicators / response performance / temperature field), and auxiliary power supply system (output quality / battery health / temperature characteristics / efficiency indicators).

[0075] Step 2: Construct a status evaluation system for each subsystem of the train based on the TOPSIS comprehensive evaluation method. The implementation method is as follows:

[0076] The collected real-time operating data and historical fault data, including text, image, video and audio data, are preprocessed to extract data features of equipment failures during operation, including time domain / frequency domain features, and the feature data is standardized.

[0077] Further, historical fault data and maintenance cost data of each subsystem are used to extract equipment operating cost and benefit-type indicators.

[0078] See attached data processing flowchart. Figure 1 .

[0079] The next step involves standardizing the indicator and benefit data in the dataset, and then calculating the values ​​of each subsystem based on the weighted normalization matrix of the original data matrix and the positive and negative ideal solutions. Proximity. The steps are as follows: First, extract n evaluation index data from m samples, and set the decision matrix. The normalized matrix is The weighted normalization matrix is ,in It's the weight.

[0080] in Sample size Number of evaluation indicators :No. One sample / observation, :No. One sample / observation, :No. The first sample The actual measured value of each indicator : Normalized index value :No. The weight of each indicator, : Relative health contribution value after indicator importance :No. The weight of each indicator.

[0081] The application of positive and negative ideal solution indices is determined as follows: calculate the distance of each sample to the positive ideal PIS and the negative ideal NIS, and the distance to the positive ideal: Distance to negative ideal: .

[0082] For efficiency-type indicators: the ideal solution Negative ideal solution ;

[0083] For cost-related indicators: the ideal solution Negative ideal solution ;

[0084] The distances from each sample to the positive ideal PIS and the negative ideal NIS are calculated as follows:

[0085] To the ideal distance: Distance to negative ideal: ;

[0086] in: It is a calculation of Euclidean distance;

[0087] Step 4: Based on the methods in Steps 1 to 3 above, calculate the relative proximity of each subsystem in turn.

[0088]

[0089] in: To the ideal distance, The distance to the negative ideal. This indicates that the state is close to optimal. 0 indicates that the state is close to the worst;

[0090] Calculate the relative proximity of each subsystem to the system The method is as follows: For each subsystem The membership degrees of the four states are calculated using the trapezoidal membership function. , , , After normalization, the probability distribution is obtained. Let the threshold values ​​for the membership functions of each subsystem be... =0.8, =0.5, =0.3, and the state probability distribution of each subsystem is as follows:

[0091] (healthy)

[0092] (good)

[0093] (Warning)

[0094] (Fault)

[0095] Normalized probability distribution:

[0096] ;

[0097] Finally, the subsystem state transition probability matrix is ​​constructed: based on the discrete state labels and C of each subsystem obtained from steps one to five. iThe state transition matrix of each subsystem is corrected based on the closeness of the health status, and the health index weight, time decay weight, and state label correction factor are extracted to predict the future health status of each subsystem.

[0098] See attached flowchart for subsystem state transition. Figure 2 .

[0099] Step 3: Based on the probability distribution P of each subsystem calculated by the trapezoidal membership function in Step 2, extract the discrete state label, consider the impact of the whole vehicle and each subsystem on the state transition under different operating environments and health states, and apply the corresponding correction factor to correct the state transition matrix for each possible state as the transition starting point.

[0100] The corrected formula is as follows: 1. Train bogie system: During operation, the train bogie system considers mechanical vibration and geometric deformation characteristics (such as axle temperature trend and wheelset wear) and load conditions, among which mechanical fatigue wear is the dominant condition factor:

[0101]

[0102] : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t;

[0103] :Transition probabilities normalized based on fatigue and wear; State transition sensitivity coefficient;

[0104] Cumulative fatigue level; Health index; Load sensitivity coefficient; Load factor;

[0105] 2. Train traction system: The train traction system needs to pay attention to IGBT switching loss rate, motor three-phase imbalance, power device degradation, and the probability of equipment state transition under different loads and weather conditions.

[0106]

[0107] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized transition probability. : Health correction function, where Indicates the health decay coefficient. It indicates a health index. Weather factors. Loading factor. : Switching loss function, where Loss sensitivity coefficient For the rate of loss exceeding the standard, This represents the material sensitivity coefficient. Health amplification factor. Three-phase unbalance function, where For system sensitivity, The thermal effect follows a quadratic law. Health vulnerability.

[0108] 3. Train Braking System: The condition assessment of the train braking system is based on brake pad thickness and brake pressure response delay parameters. Brake pad thickness is one of the important indicators for condition classification. When brake pad wear reaches a certain threshold, it will directly lead to a degrade of the braking system condition. However, rain and snow, mileage, and brake pad temperature are all factors that contribute to brake pad wear during train operation.

[0109]

[0110] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized transition probability; :state The wear sensitivity coefficient, Current brake pad thickness state Ideal thickness Temperature compensation factor (deterioration accelerates at high temperatures).

[0111] 4. Train Auxiliary Power Supply System: The train auxiliary power supply system mainly monitors output voltage ripple and capacitor ESR changes. Power module reliability is also a key indicator. Based on the TOPSIS comprehensive evaluation, status labels and health indices are extracted, and the state probability transition formula is as follows:

[0112]

[0113] : Represents the transition probability matrix after multi-factor correction, which represents the transition probability from state i to j at time t. : Normalized basic transition probability; :Basic factors, among which basic factors (Ripple Factor) (ESR factor) (Temperature factor) (Aging factors); Forced jump variable (increases the probability of "mutation failure" when health deteriorates), where Health deviation function when hour Time-suppressed jump, when hour Promotes leaps in performance. Kronecker function When =1, the target state is S4. =0 indicates other states.

[0114] The ripple factor included in the basic factors of the auxiliary power supply system. ESR factor Temperature factor aging factors The calculation method is as follows: Ripple factor: In the formula Peak ripple voltage, Indicates the rated output voltage. This represents the ripple sensitivity coefficient.

[0115] ESR factor: The denominator of the health index in the formula reflects that "the worse the health status, the greater the impact on ESR". The EST sensitivity coefficient This is the current ESR value. This is the initial ESR value.

[0116] Temperature factor: In the formula Component junction temperature, Reference temperature This is the temperature coefficient. The formula quantifies the halving of lifespan for every 20-degree increase in temperature.

[0117] Aging factors: = ( In the formula Indicates running time (hours) Indicates the design life. This represents the aging coefficient.

[0118] Step 4: Construct the vehicle state transition matrix based on the states of each subsystem obtained in Step 3. The specific method is as follows:

[0119] The transition states of each subsystem are obtained using... Construct the transition matrix of the entire vehicle. Define the bogie system state matrix as follows: The state transition matrix of the traction system is The state transition matrix of the braking system is ,

[0120] Auxiliary power supply system is The state transition matrix of the whole vehicle system product;

[0121]

[0122] The order of status coding is defined as: bogie, traction, braking, auxiliary power supply;

[0123] The first step is to calculate the two subsystems. Product, such as first calculating the joint matrix of the bogie and traction system: ;

[0124] The second step involves comparing the above results with the transition matrix of the braking system. product: The third step is to perform a transfer matrix interaction with the auxiliary power supply system. product: ; each step The rules for calculating the product are as follows: For two matrices and their product It is Block matrix:

[0125]

[0126] Therefore, for the four subsystems, each with four states, the size of the vehicle's state transition matrix is: In the vehicle state transition matrix, each element represents the probability of transitioning from one combination state to another.

[0127] Based on the above iterative calculation using the transition matrix of the subsystem, the state distribution of the entire vehicle can be represented as the state distribution of each subsystem. Product, calculate the vehicle state transition matrix: Let but It is The elements of the matrix can be determined as follows: Represent the overall vehicle state using a quadruple (i, j, k, l), where i, j, k, and l are integers from 0 to 3 (corresponding to 4 states). Map the quadruple to row or column indices, usually in lexicographical order (i.e., I changes the fastest, then k, then j, and i changes the slowest).

[0128] index Therefore, in the transition matrix, the transition probability from state (i, j, k, l) to state (i', j', k', l') is: .

[0129] The following is an example illustrating the application of train system data during the operation of a Type B train at a specific site:

[0130] 1. Obtain the status of the train bogie system

[0131] Step 1: Obtain sample data of vibration acceleration, axle temperature, and wheel flange wear of the train bogie at four different times (T1-T4), as shown in Table 1 below:

[0132] Table 1

[0133]

[0134] Indicator types: The first three are cost-based (the smaller the better), and the last one, stress safety system, is benefit-based (the larger the better).

[0135] Step 2: Data normalization, processed using the vector normalization formula. Taking the vibration index at time T4 as an example: r_vibration = 0.10 / sqrt(0.08² + 0.12² + 0.15² + 0.10²) ≈ 0.10 / 0.236 ≈ 0.424

[0136] The same calculations were performed on all indicators and times to obtain the normalized matrix R, as shown in Table 2 below:

[0137] Table 2

[0138]

[0139] Step 3: Construct the weighted normalization matrix (v)

[0140] Five indicators were assigned weights: vibration (0.30), shaft temperature (0.20), wear (0.15), load (0.15), stress (0.20), and load (0.15). The total weight was 1.

[0141] Calculate the weighted values ​​of T4: v_vibration = 0.424 * 0.30 ≈ 0.127; v_load = 0.459 * 0.15 ≈ 0.069

[0142] Following this pattern, the weighted normalization matrix V is obtained, as shown in Table 3 below:

[0143] Table 3

[0144]

[0145] Step 4: Based on the index type, select from the four samples of the weighting matrix V respectively: Positive Ideal (PIS): take the minimum value for cost-type indices (vibration, shaft temperature, wear, load); take the maximum value for benefit-type indices (stress).

[0146] PIS = [ min(v1), min(v2), min(v3), max(v4), min(v5) ] = [0.102, 0.084, 0.026, 0.102, 0.055]

[0147] Negative Ideal Solution (NIS): Cost-related indicators take the maximum value; benefit-related indicators take the minimum value.

[0148] NIS = [ max(v1), max(v2), max(v3), min(v4), max(v5) ] = [0.191, 0.103, 0.115, 0.070, 0.087]

[0149] Step 5: Calculate the distance and comprehensive health index (C_i), calculate the Euclidean distance between the current time T4 and the two ideal solutions, and the distance D to the positive ideal. + :sqrt[(0.127-0.102)² + (0.087-0.084)² + (0.038-0.026)²+ (0.093-0.102)² + (0.069-0.055)²] ≈ 0.034

[0150] Distance D to the negative ideal - :sqrt[(0.127-0.191)² + (0.087-0.103)² + (0.038-0.115)² + (0.093-0.070)² + (0.069-0.087)²] ≈ 0.110

[0151] Calculate the comprehensive health index C_i: C_i = D - / (D + + D - = 0.110 / (0.034 + 0.110) ≈ 0.764

[0152] Step 6: State Mapping and Decision Making. Input the health index C_i = 0.764 into the trapezoidal membership function (thresholds θ1=0.8, θ2=0.5, θ3=0.3) and calculate the state probabilities: Membership degrees: mu(S1)=0.820, mu(S2)=0.180, mu(S3)=0, mu(S4)=0

[0153] Probability distribution: P = [0.820, 0.180, 0, 0] (normalized).

[0154] According to the maximum probability method, the current state is S1 (healthy). Considering the addition of load, the health index (0.764) has decreased slightly, and the probability of being in state S2 has increased from 8.5% to 18%. This indicates that the health state based on the model is still "healthy", but it has shown a more obvious transition to the "good" state.

[0155] Step 7: Dynamically adjust the transition probability based on the Markov model

[0156] Based on the calculation results from steps 1 to 6, the current discrete state S1 has a health index CI of 0.764 and a load rate of 75%, based on the formula... The corrected state transition matrix is ​​as follows:

[0157] First, output the key parameters and assumptions, as shown in Table 4 below:

[0158] Table 4

[0159]

[0160] Based on historical data and engineering experience, the state transition sensitivity coefficient (αij) matrix is ​​set as follows:

[0161] α = [

[0162] [0.10, 0.15, 0.25, 0.40], # The sensitivity to transition to a worse state starting from S1 increases.

[0163] [0.15, 0.20, 0.30, 0.50], # Starting from S2

[0164] [0.20, 0.25, 0.35, 0.60], # Starting from S3

[0165] [0.25, 0.30, 0.40, 0.70]# Starting from S4 ]

[0167] The basic transition probability matrix (P_base) based on the static Markov model is as follows:

[0168] P_base = [

[0169] [0.85, 0.12, 0.02, 0.01], # S1 -> [S1, S2, S3, S4]

[0170] [0.10, 0.75, 0.12, 0.03], # S2 ->...

[0171] [0.05, 0.15, 0.70, 0.10], # S3 ->...

[0172] [0.02, 0.08, 0.20, 0.70]# S4 ->... ]

[0174] Further adjustments are made based on fatigue and health correction terms and load correction terms. Since we are currently in state S1, we only need to correct the first row of the positive moments (i.e., the probability of transitioning from S1 to S1, S2, S3, S4), while keeping the others unchanged.

[0175] 1) First, calculate the common parameters:

[0176] (1 - Ci) = 1 - 0.764 = 0.236

[0177] ΔF × (1 - Ci) = 0.25 × 0.236 = 0.059

[0178] F2 = 1 + β × load_factor = 1 + 0.05 × 0.75 = 1.0375

[0179] 2) Calculate the fatigue correction terms for each transfer in the first row (S1).

[0180] F1_1 (S1->S1) = 1 + 0.10 × 0.059 = 1.0059

[0181] F1_2 (S1->S2) = 1 + 0.15 × 0.059 = 1.00885

[0182] F1_3 (S1->S3) = 1 + 0.25 × 0.059 = 1.01475

[0183] F1_4 (S1->S4) = 1 + 0.40 × 0.059 = 1.0236

[0184] 3) Calculate the total correction factor F_total_j for each transition in the first row.

[0185] F_total_1 = 1.0059 × 1.0375 ≈ 1.0436

[0186] F_total_2 = 1.00885 × 1.0375 ≈ 1.0467

[0187] F_total_3 = 1.01475 × 1.0375 ≈ 1.0528

[0188] F_total_4 = 1.0236 × 1.0375 ≈ 1.0620

[0189] 4) Apply the correction factor and calculate the corrected probability P_corrected[0][j]

[0190] P_corr_1 = 0.85 × 1.0436 ≈ 0.8871

[0191] P_corr_2 = 0.12 × 1.0467 ≈ 0.1256

[0192] P_corr_3 = 0.02 × 1.0528 ≈ 0.0211

[0193] P_corr_4 = 0.01 × 1.0620 ≈ 0.0106

[0194] 5) Normalize the corrected rows.

[0195] The corrected sum of the first row is: 0.8871 + 0.1256 + 0.0211 + 0.0106 = 1.0444

[0196] P_11 = 0.8871 / 1.0444 ≈ 0.8494

[0197] P_12 = 0.1256 / 1.0444 ≈ 0.1202

[0198] P_13 = 0.0211 / 1.0444 ≈ 0.0202

[0199] P_14 = 0.0106 / 1.0444 ≈ 0.0102

[0200] The corrected P_corrected transition probability matrix is ​​shown in Table 5 below:

[0201] Table 5

[0202]

[0203] After probability redistribution, compared with the basic matrix, the probability of remaining in S1 decreases from 0.85 to 0.8494, while the probability of shifting to S2, S3, and S4 increases slightly.

[0204] 2. Similarly, based on historical data from each subsystem, the current states of the train auxiliary power supply system, train traction system, and train braking system are obtained through TOPSIS comprehensive evaluation and state mapping, as shown in the following tables:

[0205] Table 6 Train Traction System:

[0206]

[0207] Table 7 Train Braking System:

[0208]

[0209] Table 8 Train Auxiliary Power Supply System:

[0210]

[0211] 3. Kronecker product operation of state transition probabilities of the whole vehicle system

[0212] Based on the above calculation results, the probability distribution vectors of the current state of the four subsystems have been determined:

[0213] Bogie system: =[0.82, 0.18, 0.00, 0.00];

[0214] Traction system: =[0.95, 0.05, 0.00, 0.00];

[0215] Braking system: =[0.70, 0.25, 0.00, 0.00];

[0216] Auxiliary power supply system: =[0.98, 0.02, 0.00, 0.00];

[0217] Among the four subsystems, the bogie system has an 82% probability of being in state S1 (healthy) and an 18% probability of being in state S2 (good); the traction system has a 95% probability of being in state S1 and a 5% probability of being in state S2; the braking system, considering the wear and deterioration of the brake pads, has a 70% probability of being in state S1, a 25% probability of being in state S2, and a 5% probability of being in state S3; the auxiliary power supply system has a 98% probability of being in state S1 and a 2% probability of being in state S2.

[0218] The probability that all four subsystems are simultaneously in optimal state (S1) is calculated using the multiplication formula for the probability of independent events as follows:

[0219] The probability of a vehicle being in "completely healthy" (S1) is the probability of that the vehicle is in "completely healthy". The probability is 53.5%;

[0220] The probability of S2 for the whole vehicle is: all subsystems must be in a state no worse than S2, and at least one of them must be S2, while there cannot be S3 or S4. Subtract the portion that is "all S1". The probability is 41.5%.

[0221] The probability of the entire vehicle reaching S3 status is as follows: All subsystems must be at least at state S3, with at least one subsystem at S3, and no subsystems can reach S4. Currently, only the braking system has an S3 state (probability 0.05), and other systems have no S3 / S4 states. The probability is 5%.

[0222] Vehicle S4 probability: This means that at least one subsystem is in a fault state (S4). Based on the current data, the probability is 0.

[0223] Based on the above reasoning and calculations, there is a 53.5% probability that the vehicle system is in "complete health" and there is no risk of immediate failure (S4) in the current state. There is a 41.5% probability that it is in a slightly deteriorated state and a 5% probability that it will reach the warning state.

[0224] In summary, this invention covers the feasibility of using a TOPSIS-Markov joint model to achieve urban rail transit train condition assessment and prediction. This synthesis method successfully integrates discrete health assessments at the component level (subsystem) into a continuous probabilistic description of the system (vehicle), providing an overall health score and achieving a leap from static assessment to dynamic prediction, thus possessing value for field applications.

[0225] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0226] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0227] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the urban rail train state evaluation methods based on the TOPSIS-Markov joint model in the above embodiments.

[0228] It is understood that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0229] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0230] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0231] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0232] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the state of urban rail transit trains based on a TOPSIS-Markov joint model, characterized in that, Perform the following steps using a computer device. S1. Collect and process real-time operating data, historical fault data, and historical maintenance data of each train and its subsystems. Specifically, this includes acquiring real-time operating data, historical fault data, and historical maintenance data of the bogie system's vibration / temperature / geometric characteristics / dynamic indicators, traction system's current characteristics / temperature characteristics / insulation characteristics / efficiency indicators, braking system's hydraulic characteristics / wear indicators / response performance / temperature field, and auxiliary power supply system's output quality / battery health / temperature characteristics / efficiency indicators through field sensors and image acquisition equipment. S2. Based on historical fault data and maintenance cost data of each subsystem, extract equipment operating cost and benefit-type indicator data, and combine them with TOPSIS comprehensive evaluation to obtain discrete state labels and proximity of each subsystem. ; Among them, the closeness to each subsystem The membership degrees of the four states are calculated using the trapezoidal membership function. , , , After normalization, the probability distribution is obtained. ; Let the membership function threshold be... =0.8, =0.5, =0.3, and the state probability distribution is as follows: healthy, good, Warning, Fault, Normalized probability distribution: ； S3. Based on the TOPSIS comprehensive evaluation in S2, obtain the discrete state labels and proximity of each subsystem. The state transition matrix of each subsystem is corrected, and the health index weight, time decay weight, and state label correction factor are extracted to measure the transition state of each subsystem, i.e., the future health state. S4. Based on the transition states of each subsystem measured in S3, combine them with a Markov chain to output the overall vehicle evaluation state; specifically, the transition states of each subsystem are used... Construct the transition matrix of the entire vehicle. Define the bogie system state matrix as follows: The state transition matrix of the traction system is The state transition matrix of the braking system is The auxiliary power supply system is The state transition matrix of the whole vehicle system product; The order of defining the status codes is: bogie, traction, braking, auxiliary power supply.

2. The urban rail transit train state assessment method based on the TOPSIS-Markov joint model according to claim 1, characterized in that: In S1, data processing is performed, including classifying and preprocessing the collected real-time operating data and historical fault data, including text, image, video, and audio data, extracting data features of equipment faults during operation, including time domain / frequency domain features, and standardizing the feature data.

3. The urban rail transit train condition assessment method based on the TOPSIS-Markov joint model according to claim 2, characterized in that: S2 includes, Step 1: Extract n evaluation index data from m samples, and set up the decision matrix. The normalized matrix is The weighted normalization matrix is ,in It is weight; Sample size Number of evaluation indicators :No. One sample / observation, :No. One sample / observation, :No. The first sample The actual measured value of each indicator : Normalized index value :No. The weight of each indicator, : Relative health contribution value after indicator importance :No. The weight of each indicator; Step 2: Determine the application of positive and negative ideal solution indices For efficiency-type indicators: the ideal solution Negative ideal solution ; For cost-related indicators: the ideal solution Negative ideal solution ; Step 3: Calculate the distance from each sample to the positive ideal PIS and the negative ideal NIS, as follows: To the ideal distance: Distance to negative ideal: ; in: The distance is calculated using Euclidean distance; Step 4: Based on the methods in Steps 1 to 3 above, calculate the relative proximity of each subsystem in turn. in: To the ideal distance, The distance to the negative ideal. This indicates that the state is close to optimal. 0 indicates that the state is close to the worst.

4. The urban rail transit train state assessment method based on the TOPSIS-Markov joint model according to claim 3, characterized in that: S3 includes, During operation, the train bogie system is subjected to mechanical vibration and geometric deformation loads, with mechanical fatigue wear being the dominant state factor. Based on the TOPSIS comprehensive evaluation, the system's discrete state labels and health index values ​​are used to output the bogie state transition probability formula as follows: : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t; :Transition probabilities normalized based on fatigue and wear; State transition sensitivity coefficient; Cumulative fatigue level; Health index; Load sensitivity coefficient; : Load factor.

5. The urban rail transit train condition assessment method based on the TOPSIS-Markov joint model according to claim 3, characterized in that: S3 includes, The train traction system focuses on IGBT switching loss rate, motor three-phase imbalance, power device degradation, and considers the equipment state transition probability under different loads and weather conditions. Based on the TOPSIS comprehensive evaluation, the state label and health index value are extracted and the state probability transition formula is output as follows. : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t; : Normalized transition probability; : Health correction function, where Indicates the health decay coefficient. Indicates health index; Weather factors; Loading factor; : Switching loss function, where Loss sensitivity coefficient For the rate of loss exceeding the standard, The material sensitivity coefficient, Health amplification factor; Three-phase unbalance function, where For system sensitivity, The thermal effect follows the quadratic law. Health vulnerability.

6. The urban rail transit train state assessment method based on the TOPSIS-Markov joint model according to claim 3, characterized in that: S3 includes, The train braking system uses brake pad thickness and brake pressure response delay parameters as the basis for condition assessment. Brake pad thickness is one of the important indicators for condition classification. When brake pad wear reaches a certain threshold, it will directly lead to a degrade of the braking system condition. However, rain and snow, mileage, and brake pad temperature are all factors that contribute to brake pad wear during train operation. The following is a state probability transition formula derived by combining the TOPSIS comprehensive evaluation to extract state labels and health index values: : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t; : Normalized transition probability; :state The wear sensitivity coefficient, Current brake pad thickness state Ideal thickness Temperature compensation factor.

7. The urban rail transit train state assessment method based on the TOPSIS-Markov joint model according to claim 3, characterized in that: S3 includes, The train auxiliary power supply system monitors output voltage ripple and capacitor ESR changes. Power module reliability is also a key indicator. Based on the TOPSIS comprehensive evaluation, state labels and health indices are extracted, and the state probability transition formula is as follows: : Represents the transition probability matrix after multi-factor correction, and the transition probability from state i to j at time t; : Normalized basic transition probability; :Basic factors, among which basic factors Ripple factor ESR factor Temperature factor Aging factors; Forced jump variables increase the probability of "mutation failure" when health deteriorates. For health deviation function, when hour Time-suppressed jump, when hour Promote jumps; For the Kronecker function, When =1, the target state is S4. When =0, it represents other states; The ripple factor included in the basic factors of the auxiliary power supply system as described above ESR factor Temperature factor aging factors The calculation method is as follows: Ripple factor: In the formula Peak ripple voltage, Indicates the rated output voltage. Indicates the ripple sensitivity coefficient; ESR factor: The denominator of the health index in the formula reflects that "the worse the health status, the greater the impact on ESR". The EST sensitivity coefficient This is the current ESR value. This is the initial ESR value; Temperature factor: In the formula Component junction temperature, Reference temperature is the temperature coefficient; where the lifespan is halved for every 20-degree increase in temperature in this formula. Aging factors: = ( In the formula The running time is expressed in hours. Indicates the design life. This represents the aging coefficient.

8. The urban rail transit train condition assessment method based on the TOPSIS-Markov joint model according to claim 1, characterized in that: S4 also includes, The first step is to calculate the two subsystems. The product, i.e., first calculate the joint matrix of the bogie and traction system: ; The second step involves comparing the above results with the transition matrix of the braking system. product: The third step is to perform a transfer matrix interaction with the auxiliary power supply system. product: ; each step The product calculation rules are as follows: For two matrices and their product It is Block matrix: Therefore, for the four subsystems, each with four states, the size of the vehicle's state transition matrix is: In the vehicle state transition matrix, each element represents the probability of transitioning from one combination state to another. Based on the above iterative calculation using the transition matrix of the subsystem, the state distribution of the entire vehicle is represented by the state distribution of each subsystem. Product, calculate the vehicle state transition matrix: Let but It is The elements of the matrix are determined as follows: the vehicle state is represented by a quadruple (i, j, k, l), where i, j, k, l are integers from 0 to 3, corresponding to four states. The quadruple is then mapped to either a row index or a column index. index Therefore, in the transition matrix, the transition probability from state (i, j, k, l) to state (i', j', k', l') is: .

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

Images