Subway signal fault diagnosis and prediction system based on artificial intelligence

Abstract

The invention discloses a subway signal fault diagnosis and prediction system based on artificial intelligence, and belongs to the technical field of urban rail transit communication. Comprising a data acquisition unit used for acquiring the depth of a subway layer, a ground weather data set above the subway layer, subway layer feature data and subway people flow data, a basic signal analysis unit used for generating a subway signal theoretical score, and a staying people flow analysis unit used for generating staying people flow influence factors. The comprehensive analysis unit is used for generating a subway signal comprehensive score; the diagnosis and prediction unit is used for performing subway signal fault diagnosis and prediction according to the subway signal comprehensive score; according to the method, active prediction of fault risks is realized, the accuracy and timeliness of subway signal fault prediction are improved, and comprehensive analysis of multi-dimensional dynamic factors is realized, so that the allocation efficiency of network resources is optimized.

Description

Technical Field

[0001] This invention belongs to the field of urban rail transit communication technology, and in particular relates to an artificial intelligence-based subway signal fault diagnosis and prediction system. Background Technology

[0002] With the rapid development of urban rail transit, subways have become a major mode of public transportation. The quality of mobile communication networks (hereinafter referred to as "subway signals") within subway stations and tunnels is crucial. When subway signals malfunction or become congested, it can lead to problems such as interrupted calls, high network latency, and video buffering, severely impacting the user experience.

[0003] In current technologies, monitoring of subway signals is mostly based on performance alarms from network equipment (such as base stations) themselves (e.g., CPU load, memory usage, bit error rate, etc.), which is a passive and lagging monitoring method. Maintenance personnel typically intervene only after a fault has occurred and affected users. Furthermore, subway signal quality is affected by complex factors: firstly, the physical environment, such as tunnel depth and surface weather (rain, snow, thunderstorms), can attenuate wireless signal strength; secondly, user behavior, such as the massive instantaneous traffic during rush hour and transfers at popular stations, can overwhelm network capacity.

[0004] Therefore, there is an urgent need for a method that can proactively predict signal failures, taking into account both environmental and human factors, to provide early warnings of subway network congestion, thereby providing decision support for network optimization and dynamic resource allocation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based subway signal fault diagnosis and prediction system, which solves the aforementioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a subway signal fault diagnosis and prediction system based on artificial intelligence, specifically comprising: The data acquisition unit is used to acquire the depth of the subway level, the collection of ground weather data above the subway, subway level characteristic data, and subway passenger flow data; among which, the subway passenger flow data includes the current number of people in the subway and the historical subway passenger flow. The basic signal analysis unit is used to generate a theoretical score for the subway signal based on the depth of the subway level and the collection of ground weather data above the subway. The pedestrian flow analysis unit is used to establish a pedestrian flow analysis model and generate pedestrian flow influencing factors based on subway level characteristic data and real-time pedestrian flow. The comprehensive analysis unit is used to generate a comprehensive subway signal score based on the influencing factors of passenger flow and the theoretical score of subway signals. The diagnostic prediction unit is used to diagnose and predict subway signal faults based on the comprehensive score of subway signals.

[0007] Based on the above technical solutions, the present invention also provides the following optional technical solutions: Further technical solutions: The specific method for generating the theoretical score of the subway signal system includes: Through the formula:

[0008] Generate subway signal theoretical score ; In the formula, This represents the normalized value of the i-th weather data point in the surface weather data set. This represents the weighting coefficient of the i-th weather data in the ground weather data set, where m is the number of weather data.

[0009] Further technical solution: The dwelling flow analysis unit specifically includes: The preliminary prediction module is used to generate preliminary predictions of passenger flow based on historical subway passenger flow data. The current status analysis module is used to generate real-time correction factors based on the current number of people remaining in the subway. The final dwell time passenger flow prediction module is used to generate the final dwell time passenger flow prediction value based on the preliminary dwell time passenger flow prediction value, real-time correction factor and subway level characteristic data; The module for generating factors influencing dwell traffic flow is used to establish a traffic flow analysis model based on the final predicted dwell traffic flow and generate factors influencing dwell traffic flow.

[0010] Further technical solution: The method for generating the preliminary prediction value of the dwelling flow specifically includes: Through the formula:

[0011] Generate preliminary forecasts of visitor flow. ; In the formula, This represents the smoothing weighting coefficient. This indicates the recent historical visitor flow. This represents the periodic historical visitor flow.

[0012] Further technical solution: The method for generating the real-time correction factor specifically includes: Through the formula:

[0013] Generate real-time correction factor ; In the formula, This indicates the current number of people remaining in the subway. This represents the comfortable capacity threshold for the number of people inside a subway station. This represents the attenuation rate correction coefficient.

[0014] Further technical solution: The method for generating the final predicted visitor flow specifically includes: Through the formula:

[0015] Generate final dwell time prediction values ; In the formula, This represents a preliminary forecast of the number of people staying at the location. This represents the real-time correction factor. This indicates the value of the subway transfer station sign. This represents the value of the popular website identifier. This indicates the holiday indicator value. , , All are weighting coefficients, and .

[0016] Further technical solution: The expression of the pedestrian flow analysis model is specifically as follows:

[0017] In the expression, This represents the predicted final visitor flow. This represents the baseline threshold for network capacity, expressed in units of human figures. This represents the growth impact coefficient.

[0018] Further technical solutions: The method for generating the comprehensive subway signal score specifically includes: Through the formula:

[0019] Generate a comprehensive score for subway signals ; In the formula, This represents the theoretical score for subway signaling. This represents the factors influencing the number of people staying at a particular location. This represents the confidence factor.

[0020] Further technical solution: The specific method for generating the confidence factor includes: Through the formula:

[0021] Generate confidence factor ; In the formula, This represents the normalized value of the historical prediction accuracy. This represents the sensitivity coefficient.

[0022] Further technical solutions: The specific methods for diagnosing and predicting subway signal faults include: The comprehensive score of the subway signal system is compared with the preset score threshold. If the overall score of the subway signal is greater than or equal to the preset score threshold, it is predicted that there is no risk of subway signal failure. If the overall score of the subway signal is less than the preset score threshold, it is predicted that there is a risk of subway signal failure.

[0023] This invention provides an artificial intelligence-based subway signal fault diagnosis and prediction system, which has the following advantages compared with the prior art: This invention integrates the dynamic interaction between environmental and human factors through the synergistic effect of a data acquisition unit, a basic signal analysis unit, a passenger flow analysis unit, a comprehensive analysis unit, and a diagnostic prediction unit. It generates a comprehensive score for subway signals by using multi-dimensional data modeling and analysis, thereby enabling proactive prediction of fault risks, improving the accuracy and timeliness of subway signal fault prediction, and realizing comprehensive analysis of multi-dimensional dynamic factors, thus optimizing the allocation efficiency of network resources. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the structure of a subway signal fault diagnosis and prediction system based on artificial intelligence provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0026] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0027] Please see Figure 1 The present invention provides an artificial intelligence-based subway signal fault diagnosis and prediction system, which specifically includes: The data acquisition unit 10 is used to acquire the depth of the subway level, the collection of ground weather data above the subway, subway level characteristic data, and subway passenger flow data; among which, the subway passenger flow data includes the current number of people in the subway and the historical subway passenger flow. The basic signal analysis unit 20 is used to generate a theoretical score for the subway signal based on the depth of the subway level and the collection of ground weather data above the subway. The dwelling flow analysis unit 30 is used to establish a dwelling flow analysis model and generate dwelling flow influencing factors based on the subway floor characteristic data and real-time dwelling flow. The comprehensive analysis unit 40 is used to generate a comprehensive subway signal score based on the influencing factors of passenger flow and the theoretical score of subway signals. The diagnostic prediction unit 50 is used to perform subway signal fault diagnosis and prediction based on the comprehensive subway signal score. The depth of the subway level refers to the vertical distance between the tunnel structure and the ground. This can be achieved using a laser rangefinder or engineering drawings. This parameter directly affects the calculation of electromagnetic wave propagation loss. The ground weather data collection above the subway includes meteorological indicators such as rainfall and lightning intensity, which can be obtained in real time through the meteorological bureau's API interface to quantify the degree of weather interference with wireless signals. The subway layer feature data includes station type (whether it is a transfer station) and popular station type, which can be extracted from the subway operation management system and used to assess the passenger flow of the station; The current number of people remaining in the subway passenger flow data can be counted by combining in-station cameras with image recognition technology, while the historical passenger flow can be obtained from the ticketing system database. Together, they reflect the real-time and periodic passenger flow patterns.

[0028] Specifically, the data acquisition unit simultaneously collects environmental parameters and dynamic pedestrian flow data, overcoming the limitations of a single data source. The basic signal analysis unit establishes a theoretical attenuation model for signal propagation through weighted calculations of depth and weather data; for example, increased depth leads to exponential attenuation of signal strength, and heavy rain exacerbates signal scattering. The pedestrian flow analysis unit combines site structural characteristics, uses historical data to predict baseline pedestrian flow, and then corrects the predicted values ​​based on real-time occupancy, ultimately generating a pedestrian flow impact factor reflecting network pressure. The comprehensive analysis unit dynamically superimposes theoretical attenuation with pedestrian load; for example, during morning and evening peak hours, the weight of the pedestrian flow impact factor increases, and the comprehensive score decreases significantly. The diagnostic prediction unit judges the signal status through threshold comparison; when the score falls below a preset threshold, an early warning is triggered, guiding the dynamic expansion of network resources.

[0029] Compared to existing technologies, current solutions only monitor equipment performance parameters and cannot predict the impact of environmental and pedestrian flow changes on signal quality. This invention constructs a two-factor evaluation model for signal quality through the collaborative analysis of environmental parameters and pedestrian flow data.

[0030] Through the above technical solution, this invention achieves dynamic assessment and fault early warning of subway signal status. During morning rush hour at transfer stations, the system can identify the combined effect of surging passenger flow and signal attenuation in advance, triggering network resource allocation commands to prevent users from experiencing call interruptions or video buffering. In emergency situations such as tunnel seepage caused by heavy rain, the system quickly calculates the signal attenuation rate using weather data and depth parameters, guiding maintenance personnel to prioritize reinforcing network equipment in affected areas.

[0031] Preferably, the present invention further proposes a method for generating the subway signal theoretical score, specifically including: Through the formula:

[0032] Generate subway signal theoretical score ; In the formula, This represents the normalized value of the i-th weather data point in the surface weather data set. This represents the weighting coefficient of the i-th weather data in the surface weather data set, where m is the number of weather data. Among them, the normalized value This refers to converting weather data with different dimensions into dimensionless values ​​with a unified dimension. This can be achieved using the maximum-minimum normalization method, making heterogeneous data such as rainfall, snow cover, and lightning intensity comparable. The weighting coefficient refers to the contribution of different weather types to the impact of subway signal attenuation, which can be determined through regression analysis of historical signal attenuation data and weather factors.

[0033] Specifically, after normalization, surface weather data eliminates dimensional differences, allowing for direct weighted summation of multi-source data. The weighting coefficients reflect the varying impacts of different weather conditions on signal attenuation; for example, the high weighting coefficient for thunderstorms demonstrates their significant interference with signal strength. The summation term represents the total signal attenuation under the combined effects of multiple weather factors. Its reciprocal form ensures that the theoretical score decreases non-linearly as severe weather factors increase, consistent with the physical law that multiple factors working together in real-world scenarios lead to a sharp deterioration in signal strength. This scoring result provides an environmental assessment basis for subsequent comprehensive analysis of pedestrian flow factors.

[0034] Compared to existing technologies, which rely on passive monitoring based on network device performance alarms and cannot quantify the impact of weather factors on signal quality, this invention establishes a mathematical quantification model to convert weather data into calculable scoring indicators, enabling proactive prediction of environmental factors. Existing technologies do not consider the varying impacts of different weather data on signal attenuation, while this invention, through weighted coefficient allocation, accurately reflects the differences in the contribution of weather events such as thunderstorms and rainfall.

[0035] Through the above technical solution, the present invention solves the problem that the existing technology cannot quantify the impact of weather factors on subway signals. By normalization processing and weight allocation, the present invention achieves the fusion calculation of multiple weather data, generates a theoretical score that can quantify and evaluate signal strength, provides an environmental factor analysis basis for subsequent fault prediction, and supports dynamic allocation decision-making of network resources.

[0036] Preferably, the present invention further proposes that the dwelling flow analysis unit 30 specifically includes: The preliminary prediction module is used to generate preliminary predictions of passenger flow based on historical subway passenger flow data. The current status analysis module is used to generate real-time correction factors based on the current number of people remaining in the subway. The final dwell time passenger flow prediction module is used to generate the final dwell time passenger flow prediction value based on the preliminary dwell time passenger flow prediction value, real-time correction factor and subway level characteristic data; The module for generating factors influencing dwell traffic flow is used to establish a traffic flow analysis model based on the final predicted dwell traffic flow and generate factors influencing dwell traffic flow. Among them, historical subway passenger flow refers to the statistical amount of the number of people staying in the subway station in the past time period. Specifically, it can be stored in a time series database and divided by time window to reflect the periodic pattern and short-term trend of passenger flow. The current number of people remaining in the subway refers to the real-time data on the number of people in the station, which can be obtained using infrared sensors or video recognition technology to capture sudden changes in passenger flow. Metro level feature data refers to station attribute data that includes transfer station signs, popular station signs, and holiday signs. Specifically, it can be represented by binary encoding or categorical variables to distinguish the different impacts of different station types on passenger flow distribution. The real-time correction factor refers to the adjustment parameter that performs a non-linear mapping based on the ratio of the current number of people remaining to the comfort capacity threshold using a logistic function. Specifically, it can be implemented using the sigmoid function to suppress the deviation of the predicted value when the capacity threshold is exceeded. The crowd flow analysis model is a quantitative model that maps the final predicted value to the network carrying capacity benchmark threshold using an exponential function. Specifically, the normalized predicted value can be used as the input variable to characterize the dynamic impact of crowd overload on signal attenuation.

[0037] Specifically, the preliminary prediction module integrates recent and periodic historical data, using smoothing weighting coefficients to balance short-term fluctuations and long-term patterns to generate a basic prediction value. The current state analysis module inputs the real-time number of people remaining and a preset capacity threshold into a sigmoid function to generate a correction factor reflecting instantaneous passenger flow pressure. The final prediction module calculates the prediction results by weighting and overlaying the basic prediction value, correction factor, and station characteristic data, and adjusting the prediction results based on the spatial attributes of transfer stations, popular stations, and holidays. The passenger flow impact factor generation module inputs the final prediction value and the network capacity benchmark threshold into an exponential function to generate an impact factor that grows non-linearly with the degree of passenger flow overload, providing a dynamic correction basis for subsequent signal scoring.

[0038] Compared to existing technologies, current solutions rely solely on passive alarm data from network devices, failing to predict signal quality degradation caused by sudden surges in pedestrian traffic. This invention establishes a dynamic correction mechanism combining historical and real-time data, incorporating site spatial attribute characteristics to achieve multi-dimensional modeling of pedestrian flow trends. For example, when the number of people currently present exceeds a comfortable capacity threshold, a real-time correction factor automatically reduces the reliability of the predicted value, avoiding prediction bias caused by sudden surges in pedestrian traffic. Simultaneously, the exponential function mapping model accurately reflects the nonlinear impact of pedestrian overload on signal attenuation, exhibiting higher predictive sensitivity compared to traditional linear models.

[0039] Through the above technical solution, this invention solves the problem in existing technologies that cannot proactively predict the decline in subway signal quality caused by dynamic changes in passenger flow. By integrating historical trends, real-time status, and station characteristic data, it generates dynamically corrected passenger flow prediction values ​​and establishes a nonlinear influencing factor model to accurately quantify the potential impact of passenger flow distribution on signal quality. For example, during morning and evening rush hours, the system can automatically adjust the prediction values ​​based on the real-time number of passengers, identify key nodes by combining transfer station signs, and provide early warnings of potential network congestion, providing data support for the dynamic allocation of signal resources.

[0040] Preferably, the present invention further proposes a method for generating the preliminary predicted value of the dwelling flow, specifically including: Through the formula:

[0041] Generate preliminary forecasts of visitor flow. ; In the formula, This represents the smoothing weighting coefficient. This indicates the recent historical visitor flow. This indicates the periodic historical visitor flow, where periodicity refers to the corresponding time in a time cycle, such as this Friday and last Friday. Among them, the smoothing weight coefficient It refers to the coefficient used to adjust the contribution ratio of recent data and periodic data in the prediction model. Specifically, it can be implemented using dynamic adjustment algorithms, such as setting different weight values ​​based on differences in site type or time period. Its role is to balance the impact of sudden events and long-term patterns on traffic flow. Recent historical visitor traffic This refers to historical pedestrian traffic data within a recent time period. Specifically, a sliding window algorithm can be used to collect data from the last three days or one week, which is used to capture short-term pedestrian traffic fluctuations. Periodic historical visitor flow This refers to historical data that has the same periodicity as the predicted time point. Specifically, time series matching algorithms can be used to extract data from the same day of the previous week or the same day of the same month. Its purpose is to reflect the flow patterns of people at fixed time points.

[0042] Specifically, a linear combination formula is used to fuse recent data with periodic data. During the calculation, the smoothing weight coefficient is dynamically configured to adapt to the prediction needs of different scenarios. For example, at transfer stations, the weight coefficient can be set to a higher value to enhance the responsiveness to recent sudden changes in passenger flow; at ordinary stations, the weight coefficient can be set to a lower value to strengthen the stability of periodic patterns. Through an adjustable weighting mechanism, this formula ensures that the predicted values ​​can reflect both short-term fluctuations caused by temporary activities or sudden events and the periodic patterns inherited from historical data from the same period, thereby avoiding prediction bias caused by a single data source.

[0043] Compared to existing technologies, traditional methods rely solely on historical data from a single dimension for prediction, such as using data from the most recent period or averages over a fixed period, which cannot cope with the complex spatiotemporal variations in pedestrian traffic. This invention solves the problem of low prediction accuracy from a single data source by constructing a two-dimensional data fusion model that collaboratively analyzes short-term dynamics and long-term patterns. Furthermore, the dynamic configurability of weighting coefficients allows for adaptation to different site characteristics and operational scenarios.

[0044] Through the above technical solution, the present invention can effectively predict the dynamic trend of passenger flow in subway stations. By integrating the dual characteristics of recent and periodic data, it improves the accuracy and robustness of the prediction results, provides reliable passenger flow prediction data support for subsequent signal fault diagnosis, and thus identifies the risk of network capacity overload in advance and triggers the early warning mechanism.

[0045] Preferably, the present invention further proposes a method for generating the real-time correction factor, specifically including: Through the formula:

[0046] Generate real-time correction factor ; In the formula, This indicates the current number of people remaining in the subway. This represents the comfortable capacity threshold for the number of people inside a subway station. This represents the attenuation rate correction factor; Among them, the current number of subway passengers This refers to the number of people in the station that is collected in real time through sensors or video recognition technology. Specifically, it can be achieved by using an infrared counter or a smart camera combined with image analysis algorithms to reflect the instantaneous state of the current network load. Comfortable capacity threshold for people in subway stations It refers to the critical number of people preset based on the site's spatial layout and the network equipment's carrying capacity. Specifically, it can be determined by multiplying the site's design capacity by a safety factor, and is used to measure the potential pressure of people flow on network performance. Attenuation rate correction factor It refers to the parameter that controls the rate of change of the correction factor as the number of people deviates from the threshold. Specifically, it can be dynamically adjusted by fitting historical data or using machine learning models to regulate the system's sensitivity to sudden changes in population flow.

[0047] Specifically, the model uses the ratio of the current subway passenger volume to the comfort capacity threshold as an input variable through a logistic function model. When the actual passenger volume approaches the threshold, the normalized difference term triggers a nonlinear response of the exponential function, causing the correction factor to decline smoothly as the number of passengers increases. For example, when the actual passenger volume exceeds the threshold, the positive term in the denominator of the exponential term causes the correction factor to decrease rapidly, thereby suppressing excessive growth in the predicted value. By dynamically linking real-time passenger flow with the preset capacity, this model avoids the predictive abrupt change problem of traditional linear correction methods under critical conditions. Furthermore, the normalization process ensures the comparability of calculation results for stations of different sizes.

[0048] Compared to existing technologies, traditional methods typically employ fixed threshold alarms or adjust predicted values ​​based on linear proportions, failing to accurately reflect the dynamic nonlinear impact of pedestrian flow on network capacity. For example, existing technologies may simply generate a binary alarm signal by comparing the current number of people with a threshold, while this invention, through a continuously corrective factor output by a logic function, can reflect both the degree of pedestrian flow deviation and quantify its gradual impact on network performance.

[0049] Through the above technical solution, this invention achieves dynamic quantitative assessment of real-time passenger flow in subway stations, solving the problem that traditional passive monitoring methods cannot capture the impact of sudden changes in passenger flow on network capacity. It effectively avoids drastic fluctuations in predicted values ​​at critical points, providing an accurate basis for dynamic allocation of network resources. Furthermore, the introduction of normalization processing and adjustable attenuation rates allows the solution to adapt to the needs of stations of different sizes and diverse operational scenarios, ultimately improving the timeliness and accuracy of signal fault prediction.

[0050] Preferably, the present invention further proposes a method for generating the predicted final dwell time flow value, specifically including: Through the formula:

[0051] Generate final dwell time prediction values ; In the formula, This represents a preliminary forecast of the number of people staying at the location. This represents the real-time correction factor. This indicates the value of the subway transfer station sign. This represents the value of the popular website identifier. This indicates the holiday indicator value. , , All are weighting coefficients, and ; Among them, the value of subway transfer station signs This refers to the identifier of whether a station is a transfer hub. Specifically, it can be implemented using binary variables, such as marking a transfer station as 1 and a non-transfer station as 0, to characterize the agglomeration effect of transfer behavior on passenger flow. Popular Sites Flag Values It refers to the indicator of the commercial activity around the site, which can be implemented using continuous variables, such as assigning values ​​according to the level of customer traffic in the business district, to reflect the attraction of commercial activities to people. Holiday indicator value This refers to the classification identifier of the date attribute, which can be implemented using classification variables. For example, statutory holidays can be marked as 1 and weekdays as 0, in order to capture the characteristics of peak travel during holidays. Weighting coefficient , , This refers to the contribution allocation parameters of different scenario factors. Specifically, it can be achieved using normalization constraints, such as by constraining the total weight to 1, to avoid a single scenario factor excessively affecting the prediction results.

[0052] Specifically, a preliminary predicted value is first generated based on historical passenger flow data, which incorporates recent and periodic patterns of passenger flow changes. Then, a real-time correction factor is used to dynamically adjust the current number of passengers remaining in the subway. When the real-time number approaches the capacity threshold, the correction factor significantly reduces the predicted value to reflect network load pressure. Furthermore, features from three scenarios—transfer stations, popular stations, and holidays—are overlaid. The transfer station indicator strengthens the passenger flow aggregation effect at hub stations, the popular station indicator quantifies the attraction of commercial activities to passenger flow, and the holiday indicator identifies peak travel demand on special dates. Weighting coefficients are then used to... , , The impact of the three types of scenarios is allocated, and the contribution ratio of different scenarios to the prediction of pedestrian flow is dynamically adjusted while ensuring that the total weight is 1. Finally, the predicted values ​​of comprehensive environmental status, real-time capacity and scenario characteristics are generated.

[0053] Compared to existing technologies, traditional methods rely solely on equipment performance indicators or static historical data for prediction, failing to capture dynamic human factors such as sudden surges in passenger flow at transfer stations and abrupt changes in travel patterns during holidays. This invention introduces scenario-specific values ​​and a weighting mechanism to encode dynamic factors such as transfer behavior, commercial activities, and holiday characteristics into quantifiable parameters, enabling the prediction model to adapt to changes in passenger flow under different operational scenarios.

[0054] Through the above technical solution, this invention solves the problem of delayed signal fault warnings caused by neglecting dynamic human factors in traditional prediction methods. By integrating multi-dimensional scene marker values ​​and using a weighted constraint mechanism, it achieves accurate modeling of complex scenarios such as passenger flow gathering at transfer hubs, peak congestion in commercial areas, and surges in travel during holidays. This improves the timeliness and accuracy of passenger flow prediction, providing reliable data support for capacity warnings and resource scheduling in subway signaling networks.

[0055] Preferably, the present invention further proposes the following expression for the crowd flow analysis model:

[0056] In the expression, This represents the predicted final visitor flow. This represents the baseline threshold for network capacity, expressed in units of human figures. This represents the growth impact coefficient; Among them, the predicted final dwell time flow It refers to the predicted pedestrian flow value calculated based on historical data and real-time pedestrian flow status. Specifically, it can be generated by combining a time series prediction model with a real-time correction factor to reflect the actual pedestrian flow that may occur in future periods. Baseline threshold for network carrying capacity It refers to the maximum number of people that the network can carry, determined based on the base station hardware configuration, channel bandwidth, and signal coverage. It can be specifically set through network stress testing or historical operation data analysis, and is used to characterize the physical network capacity limit of a specific site. Growth impact coefficient It refers to the adjustment parameter that controls the relationship between the degree of overcrowding of people and the attenuation of signal quality. Specifically, it can be obtained by training historical fault data through machine learning algorithms and used to dynamically adjust the model's response strength to changes in people flow.

[0057] Specifically, the deviation between predicted passenger flow and the network's baseline threshold is converted into an exponentially growing impact factor. When the predicted passenger flow is below the baseline threshold, the impact factor approaches 1, indicating that the passenger load is within the network's normal carrying capacity. When the predicted passenger flow exceeds the baseline threshold, the rapid decay of the exponential term causes the denominator to decrease sharply, resulting in a sharp increase in the impact factor value towards 2, accurately reflecting the exponential negative impact of passenger overload on signal quality. By adjusting the slope of the exponential term, the growing impact coefficient can adapt to the sensitivity differences of network equipment at different sites. For example, a larger coefficient is set at transfer hubs to enhance early warning response, while a smaller coefficient is set at ordinary sites to avoid false alarms.

[0058] Compared to existing technologies, traditional methods employ fixed threshold alarm mechanisms, triggering warnings only when real-time pedestrian traffic exceeds a preset threshold. This fails to quantify the nonlinear relationship between overload and signal quality degradation. Existing linear regression models simply correlate pedestrian traffic with signal quality, ignoring the diminishing marginal effect of network capacity as overload increases. This invention, by introducing a logistic function with saturation characteristics, establishes a dynamic mapping relationship between the proportion of pedestrian traffic overload and the degree of signal quality degradation. This maintains baseline assessment when not overloaded and accurately reflects the accelerating deterioration trend of network performance during overload phases.

[0059] Through the above technical solution, this invention achieves dynamic quantitative assessment of network carrying capacity pressure under scenarios of sudden surges in pedestrian flow, solving the problem that traditional methods cannot accurately characterize the nonlinear relationship between pedestrian flow and signal quality. By combining a benchmark threshold and a growth coefficient, the assessment sensitivity can be automatically adjusted for different site network conditions, providing differentiated judgment criteria for signal fault prediction. Based on the nonlinear transformation mechanism of the exponential function, an early warning signal can be generated in advance when pedestrian flow approaches a critical value, providing a window of time for dynamic allocation of network resources.

[0060] Preferably, the present invention further proposes a method for generating the comprehensive subway signal score, specifically including: Through the formula:

[0061] Generate a comprehensive score for subway signals ; In the formula, This represents the theoretical score for subway signaling. This represents the factors influencing the number of people staying at a particular location. This represents the confidence factor; Among them, the theoretical score of subway signaling It refers to the theoretical signal quality score calculated based on the depth of the subway level and ground weather data. Specifically, it can be achieved by using normalized weather data and the weighted reciprocal of the weight coefficients to quantify the attenuation effect of environmental factors on signal strength. Factors affecting the number of people staying It refers to the dynamic impact coefficient generated based on the pedestrian flow prediction model. Specifically, it can be calculated through the logical function relationship between the predicted final pedestrian flow and the network carrying capacity benchmark threshold. It is used to reflect the negative offsetting effect of pedestrian overload on signal quality. Confidence factor It refers to the correction coefficient generated based on the historical prediction accuracy. Specifically, it can be calculated through a logistic function of normalized historical accuracy and sensitivity coefficient, and is used to enhance the credibility of the comprehensive score.

[0062] Specifically, by inversely correlating the theoretical score of subway signaling with the passenger flow influencing factor, the combined impact of environmental attenuation and passenger flow pressure on signal quality is reflected. For example, when the predicted passenger flow exceeds the network capacity benchmark threshold, the influencing factor increases, leading to a decrease in the overall score, thus providing an early warning of signal quality risks. Simultaneously, the confidence factor positively weights the score based on historical prediction accuracy; a higher historical accuracy increases the credibility of the overall score. This scheme establishes a multi-dimensional scoring model by integrating theoretical environmental attenuation values, real-time passenger flow pressure, and historical prediction reliability. This ensures that signal quality assessment encompasses both dynamic changes in the physical environment and user behavior trends, and can dynamically adjust the weights of the results based on the credibility of historical data.

[0063] Compared to existing technologies, current methods rely solely on passive monitoring of network device performance indicators, failing to anticipate signal quality degradation caused by sudden environmental or pedestrian changes. This invention, however, proactively models the coupling relationship between environmental and pedestrian factors, incorporating historical prediction reliability corrections to achieve dynamic prediction of signal failure risks. For example, in scenarios involving heavy rain combined with peak-hour pedestrian surges, existing technologies can only trigger alarms after base station load anomalies, while this invention can identify signal attenuation trends in advance through comprehensive scoring, providing an early warning window for resource allocation.

[0064] Through the above technical solution, this invention solves the problem of delayed fault prediction caused by relying solely on equipment performance data in existing technologies. By jointly modeling the environment, passenger flow, and historical reliability, it achieves a dynamic and comprehensive assessment of subway signal quality. For example, in high-passenger-flow scenarios at transfer stations during holidays, the system can promptly reflect network capacity pressure through passenger flow influencing factors, correct signal attenuation expectations by combining theoretical scores calculated with weather data, and adjust the score reliability based on historical prediction accuracy. Thus, it can identify signal fault risks and trigger optimization measures in advance before equipment performance becomes abnormal.

[0065] Preferably, the present invention further proposes a method for generating the confidence factor that specifically includes: Through the formula:

[0066] Generate confidence factor ; In the formula, This represents the normalized value of the historical prediction accuracy. This represents the sensitivity coefficient; Among them, the normalized value of historical prediction accuracy This refers to the standardized value obtained by comparing historical prediction results with actual failure occurrences. Specifically, the maximum-minimum normalization method can be used to map the original accuracy rate to the interval between 0 and 1, thereby eliminating the impact of differences in the units of accuracy across different time spans on the calculation. The sensitivity coefficient is a parameter used to adjust the rate of change of the historical prediction accuracy with the confidence factor. Specifically, the sensitivity of the confidence factor to the increase of the accuracy rate can be controlled by adjusting the curvature of the exponential function. For example, when the sensitivity coefficient is large, a small increase in the accuracy rate can significantly increase the confidence factor.

[0067] Specifically, this technical solution uses a logistic function to non-linearly correlate historical prediction accuracy with the confidence factor. The historical prediction accuracy, after normalization, is input into the formula, causing the confidence factor to increase slowly when the accuracy is low, rise rapidly in the middle range, and flatten out near the upper limit. A sensitivity coefficient controls the steepness of the curve; for example, when the system requires high prediction stability, increasing the sensitivity coefficient allows changes in accuracy to be more sensitively reflected in the adjustment of the confidence factor. This generates a dynamically changing confidence factor that can adaptively correct the reliability of the comprehensive subway signal scoring.

[0068] In some specific implementations, the historical prediction accuracy can be calculated using a sliding window mechanism, for example, using the prediction results of the past 30 days as the statistical period and updating the accuracy data daily. The sensitivity coefficient can be dynamically configured according to network operation and maintenance needs. For example, a higher sensitivity coefficient can be used during peak hours to enhance the response speed of confidence adjustment, while the coefficient value can be appropriately reduced during off-peak hours to reduce fluctuations.

[0069] Compared to existing technologies, current solutions typically use fixed thresholds or linear weighting to assess the reliability of prediction results, which cannot effectively address the scoring instability caused by fluctuations in historical prediction accuracy. This invention, however, achieves nonlinear adaptive adjustment of the confidence factor as historical accuracy changes through a combination of logistic functions and sensitivity coefficients. This avoids oversensitivity in low-accuracy phases while ensuring rapid convergence in high-accuracy phases.

[0070] Through the above technical solution, this invention effectively solves the problem of unstable confidence level in the comprehensive score caused by fluctuations in the reliability of historical prediction results in subway signal fault prediction. The dynamically generated confidence factor can objectively reflect the cumulative effect of historical prediction accuracy, enabling the comprehensive signal score to automatically correct confidence deviations during calculation, thereby improving the reliability of fault warning results.

[0071] Preferably, the present invention further proposes a method for diagnosing and predicting subway signal faults, specifically including: The comprehensive score of the subway signal system is compared with the preset score threshold. If the overall score of the subway signal is greater than or equal to the preset score threshold, it is predicted that there is no risk of subway signal failure. If the overall score of the subway signal is less than the preset score threshold, it is predicted that there is a risk of subway signal failure.

[0072] This invention also proposes an artificial intelligence-based method for diagnosing and predicting subway signaling faults, applied to the aforementioned artificial intelligence-based subway signaling fault diagnosis and prediction system, specifically including: Acquire the depth of the subway level, the collection of surface weather data above the subway, subway level characteristic data, and subway passenger flow data; among which, the subway passenger flow data includes the current number of people in the subway and the historical subway passenger flow. A theoretical score for the subway signal is generated based on the depth of the subway level and the collection of ground weather data above the subway. Based on the characteristic data of the subway level and the real-time passenger flow, a passenger flow analysis model is established to generate the influencing factors of passenger flow. A comprehensive subway signal score is generated based on the influencing factors of passenger flow and the theoretical score of subway signals. Based on the comprehensive score of subway signaling, subway signaling fault diagnosis and prediction are carried out.

[0073] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A subway signal fault diagnosis and prediction system based on artificial intelligence, characterized in that, The system specifically includes: The data acquisition unit is used to acquire the depth of the subway level, the collection of ground weather data above the subway, subway level characteristic data, and subway passenger flow data; among which, the subway passenger flow data includes the current number of people in the subway and the historical subway passenger flow. The basic signal analysis unit is used to generate a theoretical score for the subway signal based on the depth of the subway level and the collection of ground weather data above the subway. The pedestrian flow analysis unit is used to establish a pedestrian flow analysis model and generate pedestrian flow influencing factors based on subway level characteristic data and real-time pedestrian flow. The comprehensive analysis unit is used to generate a comprehensive subway signal score based on the influencing factors of passenger flow and the theoretical score of subway signals. The diagnostic prediction unit is used to diagnose and predict subway signal faults based on the comprehensive subway signal score. The dwelling flow analysis unit specifically includes: The preliminary prediction module is used to generate preliminary predictions of passenger flow based on historical subway passenger flow data. The current status analysis module is used to generate real-time correction factors based on the current number of people remaining in the subway. The final dwell time passenger flow prediction module is used to generate the final dwell time passenger flow prediction value based on the preliminary dwell time passenger flow prediction value, real-time correction factor and subway level characteristic data; The module for generating factors influencing dwell traffic flow is used to establish a traffic flow analysis model based on the final predicted dwell traffic flow and generate factors influencing dwell traffic flow.

2. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for generating the subway signal theory score include: Through the formula: Generate subway signal theoretical score ; In the formula, This represents the normalized value of the i-th weather data point in the surface weather data set. This represents the weighting coefficient of the i-th weather data in the ground weather data set, where m is the number of weather data.

3. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for generating the preliminary predicted value of the number of people staying include: Through the formula: Generate preliminary forecasts of visitor flow. ; In the formula, This represents the smoothing weighting coefficient. This indicates the recent historical visitor flow. This represents the periodic historical visitor flow.

4. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for generating the real-time correction factor include: Through the formula: Generate real-time correction factor ; In the formula, This indicates the current number of people remaining in the subway. This represents the comfortable capacity threshold for the number of people inside a subway station. This represents the attenuation rate correction coefficient.

5. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for generating the final predicted passenger flow include: Through the formula: Generate final dwell time prediction values ; In the formula, This represents a preliminary forecast of the number of people staying at the site. This represents the real-time correction factor. This indicates the value of the subway transfer station sign. This represents the value of the popular website indicator. This indicates the holiday indicator value. , , All are weighting coefficients, and .

6. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific expression of the crowd flow analysis model is as follows: In the expression, This represents the predicted final visitor flow. This represents the baseline threshold for network capacity, expressed in units of human figures. This represents the growth impact coefficient.

7. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for generating the comprehensive subway signal score include: Through the formula: Generate a comprehensive score for subway signals ; In the formula, This represents the theoretical score for subway signaling. This represents the factors influencing the number of people staying at a particular location. This represents the confidence factor.

8. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 7, characterized in that, The confidence factor is generated in the following ways: Through the formula: Generate confidence factor ; In the formula, This represents the normalized value of the historical prediction accuracy. This represents the sensitivity coefficient.

9. The subway signal fault diagnosis and prediction system based on artificial intelligence according to claim 1, characterized in that, The specific methods for diagnosing and predicting subway signal faults include: The comprehensive score of the subway signal system is compared with the preset score threshold. If the overall score of the subway signal is greater than or equal to the preset score threshold, it is predicted that there is no risk of subway signal failure. If the overall score of the subway signal is less than the preset score threshold, it is predicted that there is a risk of subway signal failure.

Images